mirror/freebsd
1
0
Fork 0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-18 10:35:55 +00:00
Code Issues Releases Activity
90a93ec643
freebsd/sys/i386/isa/wd.c

2441 lines
65 KiB
C
Raw Normal View History

Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*-
* Copyright (c) 1990 The Regents of the University of California.
* All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* William Jolitz.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Berkeley and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
* from: @(#)wd.c 7.2 (Berkeley) 5/9/91
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
* $Id: wd.c,v 1.139 1997/09/20 07:41:43 dyson Exp $
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* TODO:
* o Bump error count after timeout.
* o Satisfy ATA timing in all cases.
* o Finish merging berry/sos timeout code (bump error count...).
* o Merge/fix TIH/NetBSD bad144 code.
* o Don't use polling except for initialization. Need to
* reorganize the state machine. Then "extra" interrupts
* shouldn't happen (except maybe one for initialization).
* o Fix disklabel, boot and driver inconsistencies with
* bad144 in standard versions.
* o Support extended DOS partitions.
* o Support swapping to DOS partitions.
* o Handle bad sectors, clustering, disklabelling, DOS
* partitions and swapping driver-independently. Use
* i386/dkbad.c for bad sectors. Swapping will need new
* driver entries for polled reinit and polled write).
*/
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#include "wd.h"
Make files get the NWDC definition from "wdc.h" instead of "wd.h". This way, if using wd and/or wcd devices, the wdc controller code will get compiled.
1995-09-07 08:20:18 +00:00
#ifdef NWDC
#undef NWDC
#endif
#include "wdc.h"
Move CMD640 option from kernel Makefile into opt_wd.h Submitted by: Wolfgang Helbig <helbig@mx.ba-stuttgart.de>
1997-04-28 19:26:18 +00:00
#include "opt_wd.h"
Bannish ATAPI and ATAPI_STATIC #defines to opt_atapi.h.
1996-09-06 23:32:55 +00:00
#include "opt_atapi.h"
Make files get the NWDC definition from "wdc.h" instead of "wd.h". This way, if using wd and/or wcd devices, the wdc controller code will get compiled.
1995-09-07 08:20:18 +00:00
#if NWDC > 0
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
#include "pci.h"
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <sys/param.h>
#include <sys/dkbad.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/conf.h>
#include <sys/disklabel.h>
Declare all the args of wdclose() and wdioctl(). Cosmetic.
1995-02-26 01:15:30 +00:00
#include <sys/diskslice.h>
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <sys/buf.h>
#include <sys/malloc.h>
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
#ifdef DEVFS
#include <sys/devfsext.h>
#endif /*DEVFS*/
Improve the case for ST506 disks a bit: Pick up Bios-geometry. That was the good news. The bad news is that bad144 is a proper mess, and I don't have time to fix it now, so you will probably not be able to use it anyway. Sorry guys, go out and buy a 100Mb IDE drive and a paddleboard :-( If somebody wants to pick up on this: bad144 needs to learn how to stay inside our slice of the disk. That's the trick.
1994-11-18 11:27:41 +00:00
#include <machine/bootinfo.h>
Declare all the args of wdclose() and wdioctl(). Cosmetic.
1995-02-26 01:15:30 +00:00
#include <machine/clock.h>
Made wcstart() non-static again. It is called from atapi.c. The modularization of the wd/wcd/atapi driver is ugly. Include cons.h from a less bogus place. Removed an ARGSUSED. Unused args are normal for devswitch functions and lint was informed about them for about 5 functions out of 1000. Lint should be informed about them, if at all, in some other way.
1995-10-29 17:34:17 +00:00
#include <machine/cons.h>
Add and move declarations to fix all of the warnings from `gcc -Wimplicit' (except in netccitt, netiso and netns) and most of the warnings from `gcc -Wnested-externs'. Fix all the bugs found. There were no serious ones.
1995-03-16 18:17:34 +00:00
#include <machine/md_var.h>
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <i386/isa/isa.h>
#include <i386/isa/isa_device.h>
#include <i386/isa/wdreg.h>
#include <sys/syslog.h>
Add interface to new device configuration table. Also implement transfer statistics for iostat, vmstat, and systat.
1994-10-16 03:50:36 +00:00
#include <sys/dkstat.h>
Change all #includes to follow the current Berkeley style. Some of these ``changes'' are actually not changes at all, but CVS sometimes has trouble telling the difference. This also includes support for second-directory compiles. This is not quite complete yet, as `config' doesn't yet do the right thing. You can still make it work trivially, however, by doing the following: rm /sys/compile mkdir /usr/obj/sys/compile ln -s M-. /sys/compile cd /sys/i386/conf config MYKERNEL cd ../../compile/MYKERNEL ln -s /sys @ rm machine ln -s @/i386/include machine make depend make
1994-08-13 03:50:34 +00:00
#include <vm/vm.h>
Untangled the vm.h include file spaghetti.
1995-12-07 12:48:31 +00:00
#include <vm/vm_prot.h>
#include <vm/pmap.h>
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
#ifdef ATAPI
#include <i386/isa/atapi.h>
#endif
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
#include <i386/isa/wdc_p.h>
#endif /*CMD640*/
Made wcstart() non-static again. It is called from atapi.c. The modularization of the wd/wcd/atapi driver is ugly. Include cons.h from a less bogus place. Removed an ARGSUSED. Unused args are normal for devswitch functions and lint was informed about them for about 5 functions out of 1000. Lint should be informed about them, if at all, in some other way.
1995-10-29 17:34:17 +00:00
extern void wdstart(int ctrlr);
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
#define TIMEOUT 10000
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#define RETRIES 5 /* number of retries before giving up */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#define RECOVERYTIME 500000 /* usec for controller to recover after err */
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
#define MAXTRANSFER 255 /* max size of transfer in sectors */
Detect the "EOF" condition. Specifically, end of partition. Submitted by: Bruce Evans
1994-08-08 13:53:55 +00:00
/* correct max is 256 but some controllers */
/* can't handle that in all cases */
Implemented the multi-sector I/O enable and the 32 bit I/O probe through the use of the config file flags as opposed to the option "NSECS_MULTI". "NSECS_MULTI" has been removed from the driver. The new capability allows boot-time modification of the config.
1995-04-24 04:32:31 +00:00
#define WDOPT_32BIT 0x8000
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
#define WDOPT_SLEEPHACK 0x4000
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
#define WDOPT_DMA 0x2000
Add yet another kludge to this driver. Man page update to follow.
1996-07-21 09:28:50 +00:00
#define WDOPT_FORCEHD(x) (((x)&0x0f00)>>8)
Implemented the multi-sector I/O enable and the 32 bit I/O probe through the use of the config file flags as opposed to the option "NSECS_MULTI". "NSECS_MULTI" has been removed from the driver. The new capability allows boot-time modification of the config.
1995-04-24 04:32:31 +00:00
#define WDOPT_MULTIMASK 0x00ff
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/*
* This biotab field doubles as a field for the physical unit number on
* the controller.
*/
#define id_physid id_scsiid
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* Drive states. Used to initialize drive.
*/
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#define CLOSED 0 /* disk is closed. */
#define WANTOPEN 1 /* open requested, not started */
#define RECAL 2 /* doing restore */
#define OPEN 3 /* done with open */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#define PRIMARY 0
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
/*
* Disk geometry. A small part of struct disklabel.
* XXX disklabel.5 contains an old clone of disklabel.h.
*/
struct diskgeom {
u_long d_secsize; /* # of bytes per sector */
u_long d_nsectors; /* # of data sectors per track */
u_long d_ntracks; /* # of tracks per cylinder */
u_long d_ncylinders; /* # of data cylinders per unit */
u_long d_secpercyl; /* # of data sectors per cylinder */
u_long d_secperunit; /* # of data sectors per unit */
u_long d_precompcyl; /* XXX always 0 */
};
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* The structure of a disk drive.
*/
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
struct disk {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
long dk_bc; /* byte count left */
short dk_skip; /* blocks already transferred */
Changes to support 1Tb filesizes. Pages are now named by an (object,index) pair instead of (object,offset) pair.
1995-12-11 04:58:34 +00:00
int dk_ctrlr; /* physical controller number */
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
int dk_ctrlr_cmd640;/* controller number for CMD640 quirk */
#endif
Changes to support 1Tb filesizes. Pages are now named by an (object,index) pair instead of (object,offset) pair.
1995-12-11 04:58:34 +00:00
int dk_unit; /* physical unit number */
int dk_lunit; /* logical unit number */
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
int dk_interface; /* interface (two ctrlrs per interface) */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
char dk_state; /* control state */
u_char dk_status; /* copy of status reg. */
u_char dk_error; /* copy of error reg. */
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
u_char dk_timeout; /* countdown to next timeout */
Fixed pessimized (short) i/o port type.
1996-11-11 15:57:40 +00:00
int dk_port; /* i/o port base */
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
int dk_altport; /* altstatus port base */
First attempt at creating devfs entries for sliced devices. Doesn't quite work yet, so the heart of it is disabled. Added bdev and cdev args to dsopen(). drivers: Fixed device names, links, minor numbers and modes. wd.c: Started actually supporting devfs. diskslice.h: Added devfs tokens to structs (currently 576 of them per disk! :-(). subr_diskslice.c: Create devfs entries in dsopen() and (unsuccessfully) attempt to make them go away at the right times. DEVFS is #undefed at the start so that this shouldn't cause problems.
1996-01-27 04:18:15 +00:00
#ifdef DEVFS
void *dk_bdev; /* devfs token for whole disk */
void *dk_cdev; /* devfs token for raw whole disk */
#endif
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
u_long cfg_flags; /* configured characteristics */
Changes to support the "new" bad144. Adds a "BADSCAN" ioctl and support. Reviewed by: Bruce Evans Submitted by: Poul-Henning Kamp
1995-05-16 07:52:17 +00:00
short dk_flags; /* drive characteristics found */
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#define DKFL_SINGLE 0x00004 /* sector at a time mode */
#define DKFL_ERROR 0x00008 /* processing a disk error */
#define DKFL_LABELLING 0x00080 /* readdisklabel() in progress */
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
#define DKFL_32BIT 0x00100 /* use 32-bit i/o mode */
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
#define DKFL_MULTI 0x00200 /* use multi-i/o mode */
Changes to support the "new" bad144. Adds a "BADSCAN" ioctl and support. Reviewed by: Bruce Evans Submitted by: Poul-Henning Kamp
1995-05-16 07:52:17 +00:00
#define DKFL_BADSCAN 0x00400 /* report all errors */
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
#define DKFL_USEDMA 0x00800 /* use DMA for data transfers */
#define DKFL_DMA 0x01000 /* using DMA on this transfer-- DKFL_SINGLE
* overrides this
*/
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
struct wdparams dk_params; /* ESDI/IDE drive/controller parameters */
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
int dk_dkunit; /* disk stats unit number */
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
int dk_multi; /* multi transfers */
int dk_currentiosize; /* current io size */
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
struct diskgeom dk_dd; /* device configuration data */
struct diskslices *dk_slices; /* virtual drives */
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
void *dk_dmacookie; /* handle for DMA services */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
};
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
#define WD_COUNT_RETRIES
static int wdtest = 0;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
static struct disk *wddrives[NWD]; /* table of units */
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
static struct buf_queue_head drive_queue[NWD]; /* head of queue per drive */
static struct {
int b_active;
} wdutab[NWD];
/*
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
static struct buf wdtab[NWDC];
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
*/
static struct {
struct buf_queue_head controller_queue;
int b_errcnt;
int b_active;
} wdtab[NWDC];
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
struct wddma wddma[NWDC];
Implement a prototype interface to bus-master IDE DMA on the Triton chipset. This does not attempt to do anything special with the timing on the hope that the BIOS will have done the right thing already. The actual interface from the wd driver to the new facility is not implemented yet (this commit being an attempt at prodding someone else to do it because looking at the wd driver always confuses the h*** out of me).
1996-01-28 22:16:20 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#ifdef notyet
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
static struct buf rwdbuf[NWD]; /* buffers for raw IO */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
static int wdprobe(struct isa_device *dvp);
static int wdattach(struct isa_device *dvp);
static void wdustart(struct disk *du);
static int wdcontrol(struct buf *bp);
static int wdcommand(struct disk *du, u_int cylinder, u_int head,
u_int sector, u_int count, u_int command);
static int wdsetctlr(struct disk *du);
Clean up -Wunused warnings. Reviewed by: bde
1996-06-12 05:11:41 +00:00
#if 0
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
static int wdwsetctlr(struct disk *du);
Clean up -Wunused warnings. Reviewed by: bde
1996-06-12 05:11:41 +00:00
#endif
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
static int wdsetmode(int mode, void *wdinfo);
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
static int wdgetctlr(struct disk *du);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
static void wderror(struct buf *bp, struct disk *du, char *mesg);
static void wdflushirq(struct disk *du, int old_ipl);
static int wdreset(struct disk *du);
static void wdsleep(int ctrlr, char *wmesg);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
static void wdstrategy1(struct buf *bp);
Ran ft.c through ident. Added a missing #ifdef INET wrapper in lpt.c Main change: Removed the timeout_func_t casts from timeout calls and correctly defined the timeout routines to conform to the new format. lpt.c doesn't have this change. Reviewed by: Submitted by:
1994-08-23 07:52:29 +00:00
static timeout_t wdtimeout;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
static int wdunwedge(struct disk *du);
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
static int wdwait(struct disk *du, u_char bits_wanted, int timeout);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
struct isa_driver wdcdriver = {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdprobe, wdattach, "wdc",
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
};
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
static d_open_t wdopen;
static d_close_t wdclose;
Removed new alias d_size_t for d_psize_t. Removed old aliases d_rdwr_t and d_ttycv_t for d_read_t/d_write_t and d_devtotty_t. Sorted declarations of switch functions into switch order. Removed duplicated comments and declarations of nonexistent switch functions.
1995-12-10 15:55:34 +00:00
static d_strategy_t wdstrategy;
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
static d_ioctl_t wdioctl;
static d_dump_t wddump;
Removed new alias d_size_t for d_psize_t. Removed old aliases d_rdwr_t and d_ttycv_t for d_read_t/d_write_t and d_devtotty_t. Sorted declarations of switch functions into switch order. Removed duplicated comments and declarations of nonexistent switch functions.
1995-12-10 15:55:34 +00:00
static d_psize_t wdsize;
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
#define CDEV_MAJOR 3
#define BDEV_MAJOR 0
Make a "DWIM" function for adding [bc]devsw entries for bdev drivers. Saves about 280 butes of source per driver, 56 bytes in object size and another 56 bytes moves from data to bss. No functional change intended nor expected. GENERIC should be about one k smaller now :-)
1996-07-23 21:52:43 +00:00
static struct cdevsw wd_cdevsw;
Julian forgot to make the *devsw structures static.
1995-12-08 23:23:00 +00:00
static struct bdevsw wd_bdevsw =
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
{ wdopen, wdclose, wdstrategy, wdioctl, /*0*/
This is the kernel Lite/2 commit. There are some requisite userland changes, so don't expect to be able to run the kernel as-is (very well) without the appropriate Lite/2 userland changes. The system boots and can mount UFS filesystems. Untested: ext2fs, msdosfs, NFS Known problems: Incorrect Berkeley ID strings in some files. Mount_std mounts will not work until the getfsent library routine is changed. Reviewed by: various people Submitted by: Jeffery Hsu <hsu@freebsd.org>
1997-02-10 02:22:35 +00:00
wddump, wdsize, D_DISK, "wd", &wd_cdevsw, -1 };
Pass 3 of the great devsw changes most devsw referenced functions are now static, as they are in the same file as their devsw structure. I've also added DEVFS support for nearly every device in the system, however many of the devices have 'incorrect' names under DEVFS because I couldn't quickly work out the correct naming conventions. (but devfs won't be coming on line for a month or so anyhow so that doesn't matter) If you "OWN" a device which would normally have an entry in /dev then search for the devfs_add_devsw() entries and munge to make them right.. check out similar devices to see what I might have done in them in you can't see what's going on.. for a laugh compare conf.c conf.h defore and after... :) I have not doen DEVFS entries for any DISKSLICE devices yet as that will be a much more complicated job.. (pass 5 :) pass 4 will be to make the devsw tables of type (cdevsw * ) rather than (cdevsw) seems to work here.. complaints to the usual places.. :)
1995-12-08 11:19:42 +00:00
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
static int atapictrlr;
static int eide_quirks;
#endif
/*
* Here we use the pci-subsystem to find out, whether there is
* a cmd640b-chip attached on this pci-bus. This public routine
* will be called by wdc_p.c .
*/
#ifdef CMD640
void
wdc_pci(int quirks)
{
eide_quirks = quirks;
}
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* Probe for controller.
*/
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
wdprobe(struct isa_device *dvp)
{
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
int unit = dvp->id_unit;
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
void *cookie;
int interface;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
struct disk *du;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if (unit >= NWDC)
return (0);
isa.c isa_device.h: declare & define {e,}isa_{in,ex}ternalize(). fd.c: register devices and implement disk stats. wd.c: fix disk stats and call isa_externalize() as appropriate.
1994-10-19 00:08:07 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du = malloc(sizeof *du, M_TEMP, M_NOWAIT);
if (du == NULL)
return (0);
bzero(du, sizeof *du);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du->dk_ctrlr = dvp->id_unit;
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
interface = du->dk_ctrlr / 2;
du->dk_interface = interface;
#if !defined(DISABLE_PCI_IDE) && (NPCI > 0)
if (wddma[interface].wdd_candma) {
du->dk_dmacookie = wddma[interface].wdd_candma(dvp->id_iobase, du->dk_ctrlr);
du->dk_port = dvp->id_iobase;
du->dk_altport = wddma[interface].wdd_altiobase(du->dk_dmacookie);
} else {
du->dk_port = dvp->id_iobase;
du->dk_altport = du->dk_port + wd_ctlr;
}
#else
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_port = dvp->id_iobase;
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
du->dk_altport = du->dk_port + wd_ctlr;
#endif
Add a class field to devconf and mst drivers. For those where it was easy, drivers were also fixed to call dev_attach() during probe rather than attach (in keeping with the new design articulated in a mail message five months ago). For a few that were really easy, correct state tracking was added as well. The `fd' driver was fixed to correctly fill in the description. The CPU identify code was fixed to attach a `cpu' device. The code was also massively reordered to fill in cpu_model with somethingremotely resembling what identifycpu() prints out. A few bytes saved by using %b to format the features list rather than lots of ifs.
1995-04-12 20:48:13 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* check if we have registers that work */
Bring IDE CDROM support up to latest version (1.8a?) from Serge. Submitted by: vak@cronyx.ru
1995-09-30 00:11:19 +00:00
outb(du->dk_port + wd_sdh, WDSD_IBM); /* set unit 0 */
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
outb(du->dk_port + wd_cyl_lo, 0xa5); /* wd_cyl_lo is read/write */
Latest fixes from Serge: I tried to solve the problem of IDE probing compatibility in this version. When compiled without an ATAPI option, the wd driver is fully backward compatible with 2.0.5. With ATAPI option, the wdprobe becomes strictly weaker. That is, if wdprobe works without ATAPI option, it will always work with it too. Another problem was with the CD-ROM drive attached as a slave in the IDE bus, where there is no master. All IDE CD-ROM drives are shipped in slave configuration, and most users just plug them in, never thinking about jumpers. It works fine with ms-dos and ms-windows, and this version of the driver supports it as well. The eject op can now load disks. Just repeat it twice, and the disk will be ejected and then loaded back. The disc cannot be ejected if it is mounted. Submitted by: Serge Vakulenko, <vak@cronyx.ru>
1995-10-14 15:41:10 +00:00
if (inb(du->dk_port + wd_cyl_lo) == 0xff) { /* XXX too weak */
#ifdef ATAPI
/* There is no master, try the ATAPI slave. */
outb(du->dk_port + wd_sdh, WDSD_IBM | 0x10);
outb(du->dk_port + wd_cyl_lo, 0xa5);
if (inb(du->dk_port + wd_cyl_lo) == 0xff)
#endif
goto nodevice;
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Latest fixes from Serge: I tried to solve the problem of IDE probing compatibility in this version. When compiled without an ATAPI option, the wd driver is fully backward compatible with 2.0.5. With ATAPI option, the wdprobe becomes strictly weaker. That is, if wdprobe works without ATAPI option, it will always work with it too. Another problem was with the CD-ROM drive attached as a slave in the IDE bus, where there is no master. All IDE CD-ROM drives are shipped in slave configuration, and most users just plug them in, never thinking about jumpers. It works fine with ms-dos and ms-windows, and this version of the driver supports it as well. The eject op can now load disks. Just repeat it twice, and the disk will be ejected and then loaded back. The disc cannot be ejected if it is mounted. Submitted by: Serge Vakulenko, <vak@cronyx.ru>
1995-10-14 15:41:10 +00:00
if (wdreset(du) == 0)
goto reset_ok;
#ifdef ATAPI
/* test for ATAPI signature */
outb(du->dk_port + wd_sdh, WDSD_IBM); /* master */
if (inb(du->dk_port + wd_cyl_lo) == 0x14 &&
inb(du->dk_port + wd_cyl_hi) == 0xeb)
goto reset_ok;
du->dk_unit = 1;
outb(du->dk_port + wd_sdh, WDSD_IBM | 0x10); /* slave */
if (inb(du->dk_port + wd_cyl_lo) == 0x14 &&
inb(du->dk_port + wd_cyl_hi) == 0xeb)
goto reset_ok;
#endif
DELAY(RECOVERYTIME);
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
if (wdreset(du) != 0) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
goto nodevice;
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
}
Latest fixes from Serge: I tried to solve the problem of IDE probing compatibility in this version. When compiled without an ATAPI option, the wd driver is fully backward compatible with 2.0.5. With ATAPI option, the wdprobe becomes strictly weaker. That is, if wdprobe works without ATAPI option, it will always work with it too. Another problem was with the CD-ROM drive attached as a slave in the IDE bus, where there is no master. All IDE CD-ROM drives are shipped in slave configuration, and most users just plug them in, never thinking about jumpers. It works fine with ms-dos and ms-windows, and this version of the driver supports it as well. The eject op can now load disks. Just repeat it twice, and the disk will be ejected and then loaded back. The disc cannot be ejected if it is mounted. Submitted by: Serge Vakulenko, <vak@cronyx.ru>
1995-10-14 15:41:10 +00:00
reset_ok:
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* execute a controller only command */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (wdcommand(du, 0, 0, 0, 0, WDCC_DIAGNOSE) != 0
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
|| wdwait(du, 0, TIMEOUT) < 0) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
goto nodevice;
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Modify wdprobe() to correctly recognise some 2 drive systems, on which it was failing. Modifications follow ATA, but also allow some weird setups which seem to contradict ATA (aaarrrrgghhh!!)
1994-04-10 11:17:13 +00:00
/*
* drive(s) did not time out during diagnostic :
* Get error status and check that both drives are OK.
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
* Table 9-2 of ATA specs suggests that we must check for
* a value of 0x01
Modify wdprobe() to correctly recognise some 2 drive systems, on which it was failing. Modifications follow ATA, but also allow some weird setups which seem to contradict ATA (aaarrrrgghhh!!)
1994-04-10 11:17:13 +00:00
*
* Strangely, some controllers will return a status of
* 0x81 (drive 0 OK, drive 1 failure), and then when
* the DRV bit is set, return status of 0x01 (OK) for
* drive 2. (This seems to contradict the ATA spec.)
*/
du->dk_error = inb(du->dk_port + wd_error);
if(du->dk_error != 0x01) {
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
if(du->dk_error & 0x80) { /* drive 1 failure */
Modify wdprobe() to correctly recognise some 2 drive systems, on which it was failing. Modifications follow ATA, but also allow some weird setups which seem to contradict ATA (aaarrrrgghhh!!)
1994-04-10 11:17:13 +00:00
/* first set the DRV bit */
u_int sdh;
sdh = inb(du->dk_port+ wd_sdh);
sdh = sdh | 0x10;
outb(du->dk_port+ wd_sdh, sdh);
/* Wait, to make sure drv 1 has completed diags */
if ( wdwait(du, 0, TIMEOUT) < 0)
goto nodevice;
/* Get status for drive 1 */
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
du->dk_error = inb(du->dk_port + wd_error);
Modify wdprobe() to correctly recognise some 2 drive systems, on which it was failing. Modifications follow ATA, but also allow some weird setups which seem to contradict ATA (aaarrrrgghhh!!)
1994-04-10 11:17:13 +00:00
/* printf("Error (drv 1) : %x\n", du->dk_error); */
Apply my small patch to make detection of ATAPI CD-ROMs happen a little more reliably. So far I've received a couple of positive responses and no objections to these changes. There are two one-line changes: - In wdprobe(), when testing the error status of drives, don't unconditionally decide that there is no controller present if we read back a value of 0x81 (drive 0 okay, drive 1 failed) twice in a row. This may be caused by having an ATAPI CD-ROM jumpered as a master on the controller with no slave. - In wdgetctlr(), when checking for a status of WDCS_READY, check the value twice. The first time may be bogus. This stops a phantom wd2 device from being detected when an ATAPI CD-ROM is attached to the secondary controller alone as a slave. (This can cause installation to fail when sysinstall attempts to open the phantom device and wedges the system as a result. This has bitten me a couple of times on some Gateway 2000 machines.)
1996-08-12 00:53:02 +00:00
/*
* Sometimes (apparently mostly with ATAPI
* drives involved) 0x81 really means 0x81
* (drive 0 OK, drive 1 failed).
*/
if(du->dk_error != 0x01 && du->dk_error != 0x81)
Modify wdprobe() to correctly recognise some 2 drive systems, on which it was failing. Modifications follow ATA, but also allow some weird setups which seem to contradict ATA (aaarrrrgghhh!!)
1994-04-10 11:17:13 +00:00
goto nodevice;
} else /* drive 0 fail */
goto nodevice;
}
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
Modify wdprobe() to correctly recognise some 2 drive systems, on which it was failing. Modifications follow ATA, but also allow some weird setups which seem to contradict ATA (aaarrrrgghhh!!)
1994-04-10 11:17:13 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
free(du, M_TEMP);
xxxprobe now returns the correct number of I/O registers used so that the probe output during boot gives the I/O range instead of just the starting address
1993-08-28 03:02:52 +00:00
return (IO_WDCSIZE);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
nodevice:
free(du, M_TEMP);
return (0);
}
/*
* Attach each drive if possible.
*/
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
static int
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
wdattach(struct isa_device *dvp)
{
First attempt at creating devfs entries for sliced devices. Doesn't quite work yet, so the heart of it is disabled. Added bdev and cdev args to dsopen(). drivers: Fixed device names, links, minor numbers and modes. wd.c: Started actually supporting devfs. diskslice.h: Added devfs tokens to structs (currently 576 of them per disk! :-(). subr_diskslice.c: Create devfs entries in dsopen() and (unsuccessfully) attempt to make them go away at the right times. DEVFS is #undefed at the start so that this shouldn't cause problems.
1996-01-27 04:18:15 +00:00
#ifdef DEVFS
int mynor;
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
int unit, lunit;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
struct isa_device *wdup;
struct disk *du;
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
struct wdparams *wp;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (dvp->id_unit >= NWDC)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (0);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
if (eide_quirks & Q_CMD640B) {
if (dvp->id_unit == PRIMARY) {
printf("wdc0: CMD640B workaround enabled\n");
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bufq_init(&wdtab[PRIMARY].controller_queue);
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
}
} else
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bufq_init(&wdtab[dvp->id_unit].controller_queue);
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#else
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bufq_init(&wdtab[dvp->id_unit].controller_queue);
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
Add interface to new device configuration table. Also implement transfer statistics for iostat, vmstat, and systat.
1994-10-16 03:50:36 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
for (wdup = isa_biotab_wdc; wdup->id_driver != 0; wdup++) {
if (wdup->id_iobase != dvp->id_iobase)
continue;
lunit = wdup->id_unit;
if (lunit >= NWD)
continue;
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
unit = wdup->id_physid;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du = malloc(sizeof *du, M_TEMP, M_NOWAIT);
if (du == NULL)
continue;
if (wddrives[lunit] != NULL)
panic("drive attached twice");
wddrives[lunit] = du;
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bufq_init(&drive_queue[lunit]);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
bzero(du, sizeof *du);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du->dk_ctrlr = dvp->id_unit;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
if (eide_quirks & Q_CMD640B) {
du->dk_ctrlr_cmd640 = PRIMARY;
} else {
du->dk_ctrlr_cmd640 = du->dk_ctrlr;
}
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du->dk_unit = unit;
du->dk_lunit = lunit;
du->dk_port = dvp->id_iobase;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Implemented the multi-sector I/O enable and the 32 bit I/O probe through the use of the config file flags as opposed to the option "NSECS_MULTI". "NSECS_MULTI" has been removed from the driver. The new capability allows boot-time modification of the config.
1995-04-24 04:32:31 +00:00
/*
* Use the individual device flags or the controller
* flags.
*/
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
du->cfg_flags = wdup->id_flags |
((dvp->id_flags) >> (16 * unit));
if (wdgetctlr(du) == 0) {
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
/*
* Print out description of drive.
Use the new & improved printf rather than homegrown kludges. Proposed by: bde
1996-01-16 18:13:18 +00:00
* wdp_model may not be null terminated.
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
*/
Use the new & improved printf rather than homegrown kludges. Proposed by: bde
1996-01-16 18:13:18 +00:00
printf("wdc%d: unit %d (wd%d): <%.*s>",
dvp->id_unit, unit, lunit,
sizeof du->dk_params.wdp_model,
du->dk_params.wdp_model);
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
if (du->dk_flags & DKFL_USEDMA)
printf(", DMA");
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (du->dk_flags & DKFL_32BIT)
printf(", 32-bit");
if (du->dk_multi > 1)
printf(", multi-block-%d", du->dk_multi);
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
if (du->cfg_flags & WDOPT_SLEEPHACK)
printf(", sleep-hack");
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
printf("\n");
if (du->dk_params.wdp_heads == 0)
printf("wd%d: size unknown, using %s values\n",
lunit, du->dk_dd.d_secperunit > 17
? "BIOS" : "fake");
printf(
"wd%d: %luMB (%lu sectors), %lu cyls, %lu heads, %lu S/T, %lu B/S\n",
lunit,
du->dk_dd.d_secperunit
Use the same blocks->size-in-MB conversion algorithm as the SCSI code to avoid overflowing an intermediate value for disks > 2^32 bytes large.
1997-02-18 23:31:53 +00:00
/ ((1024L * 1024L) / du->dk_dd.d_secsize),
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
du->dk_dd.d_secperunit,
du->dk_dd.d_ncylinders,
du->dk_dd.d_ntracks,
du->dk_dd.d_nsectors,
du->dk_dd.d_secsize);
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
if (bootverbose) {
wp = &du->dk_params;
printf(
"wd%d: ATA INQUIRE valid = %04x, dmamword = %04x, apio = %04x, udma = %04x\n",
du->dk_lunit,
wp->wdp_atavalid,
wp->wdp_dmamword,
wp->wdp_eidepiomodes,
wp->wdp_udmamode);
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
/*
* Start timeout routine for this drive.
* XXX timeout should be per controller.
*/
wdtimeout(du);
First attempt at creating devfs entries for sliced devices. Doesn't quite work yet, so the heart of it is disabled. Added bdev and cdev args to dsopen(). drivers: Fixed device names, links, minor numbers and modes. wd.c: Started actually supporting devfs. diskslice.h: Added devfs tokens to structs (currently 576 of them per disk! :-(). subr_diskslice.c: Create devfs entries in dsopen() and (unsuccessfully) attempt to make them go away at the right times. DEVFS is #undefed at the start so that this shouldn't cause problems.
1996-01-27 04:18:15 +00:00
#ifdef DEVFS
mynor = dkmakeminor(unit, WHOLE_DISK_SLICE, RAW_PART);
Fixed group of disk devices (was wheel or games, now operator). Added scsi control devices. Converted almost everything that I changed to use devfs_add_devswf() and verbose id macros. st.c: Renamed enrst* to erst* since that's what the current name is (enrst seems to be an old name).
1996-03-27 18:50:10 +00:00
du->dk_bdev = devfs_add_devswf(&wd_bdevsw, mynor,
DV_BLK, UID_ROOT,
GID_OPERATOR, 0640,
"wd%d", unit);
du->dk_cdev = devfs_add_devswf(&wd_cdevsw, mynor,
DV_CHR, UID_ROOT,
GID_OPERATOR, 0640,
"rwd%d", unit);
First attempt at creating devfs entries for sliced devices. Doesn't quite work yet, so the heart of it is disabled. Added bdev and cdev args to dsopen(). drivers: Fixed device names, links, minor numbers and modes. wd.c: Started actually supporting devfs. diskslice.h: Added devfs tokens to structs (currently 576 of them per disk! :-(). subr_diskslice.c: Create devfs entries in dsopen() and (unsuccessfully) attempt to make them go away at the right times. DEVFS is #undefed at the start so that this shouldn't cause problems.
1996-01-27 04:18:15 +00:00
#endif
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (dk_ndrive < DK_NDRIVE) {
sprintf(dk_names[dk_ndrive], "wd%d", lunit);
/*
* XXX we don't know the transfer rate of the
* drive. Guess the maximum ISA rate of
* 4MB/sec. `wpms' is words per _second_
* according to iostat.
*/
dk_wpms[dk_ndrive] = 4 * 1024 * 1024 / 2;
du->dk_dkunit = dk_ndrive++;
} else {
du->dk_dkunit = -1;
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
} else {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
free(du, M_TEMP);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wddrives[lunit] = NULL;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
}
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
#ifdef ATAPI
/*
* Probe all free IDE units, searching for ATAPI drives.
*/
for (unit=0; unit<2; ++unit) {
for (lunit=0; lunit<NWD && wddrives[lunit]; ++lunit)
if (wddrives[lunit]->dk_ctrlr == dvp->id_unit &&
wddrives[lunit]->dk_unit == unit)
goto next;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
if (atapi_attach (dvp->id_unit, unit, dvp->id_iobase))
atapictrlr = dvp->id_unit;
#else
Remove devconf, it never grew up to be of any use.
1996-09-06 23:09:20 +00:00
atapi_attach (dvp->id_unit, unit, dvp->id_iobase);
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
Don't depend on gcc's feature of permitting labels that aren't followed by a statement.
1997-07-01 00:22:51 +00:00
next: ;
}
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/*
* Discard any interrupts generated by wdgetctlr(). wdflushirq()
* doesn't work now because the ambient ipl is too high.
*/
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
Fix problem introduced in the CMD640 workaround which could lead to a NULL pointer dereference. PR: kern/3696 Submitted by: Joel.Faedi@esial.u-nancy.fr
1997-05-27 18:28:08 +00:00
if (eide_quirks & Q_CMD640B) {
wdtab[PRIMARY].b_active = 2;
} else {
wdtab[dvp->id_unit].b_active = 2;
}
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#else
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wdtab[dvp->id_unit].b_active = 2;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (1);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/* Read/write routine for a buffer. Finds the proper unit, range checks
* arguments, and schedules the transfer. Does not wait for the transfer
* to complete. Multi-page transfers are supported. All I/O requests must
* be a multiple of a sector in length.
*/
Make the LINT kernel compile with -W -Wreturn-type -Wcomment -Werror, and add same (sans -Werror) to Makefile for future compilations.
1993-11-25 01:38:01 +00:00
void
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
wdstrategy(register struct buf *bp)
{
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
struct disk *du;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
int lunit = dkunit(bp->b_dev);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
int s;
/* valid unit, controller, and request? */
|Both, the wd and the sd driver do not reject invalid request of odd |sizes. They simply pass them to the disks, which usually causes |fatal errors then.
1994-11-03 18:20:15 +00:00
if (lunit >= NWD || bp->b_blkno < 0 || (du = wddrives[lunit]) == NULL
Last commit was bogus. Changed b_bsize to b_bcount.
1994-11-04 05:21:17 +00:00
|| bp->b_bcount % DEV_BSIZE != 0) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
bp->b_error = EINVAL;
bp->b_flags |= B_ERROR;
goto done;
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/*
Removed all references to b_cylinder (aka b_cylin). It was evil and hasn't been used for a year or two since disksort() started sorting on b_pblkno.
1996-12-01 16:34:41 +00:00
* Do bounds checking, adjust transfer, and set b_pblkno.
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
*/
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (dscheck(bp, du->dk_slices) <= 0)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
goto done;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
/*
* Check for *any* block on this transfer being on the bad block list
* if it is, then flag the block as a transfer that requires
* bad block handling. Also, used as a hint for low level disksort
* clustering code to keep from coalescing a bad transfer into
* a normal transfer. Single block transfers for a large number of
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
* blocks associated with a cluster I/O are undesirable.
*
* XXX the old disksort() doesn't look at B_BAD. Coalescing _is_
* desirable. We should split the results at bad blocks just
* like we should split them at MAXTRANSFER boundaries.
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
*/
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (dsgetbad(bp->b_dev, du->dk_slices) != NULL) {
long *badsect = dsgetbad(bp->b_dev, du->dk_slices)->bi_bad;
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
int i;
int nsecs = howmany(bp->b_bcount, DEV_BSIZE);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
/* XXX pblkno is too physical. */
daddr_t nspblkno = bp->b_pblkno
- du->dk_slices->dss_slices[dkslice(bp->b_dev)].ds_offset;
int blkend = nspblkno + nsecs;
for (i = 0; badsect[i] != -1 && badsect[i] < blkend; i++) {
if (badsect[i] >= nspblkno) {
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
bp->b_flags |= B_BAD;
break;
}
}
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* queue transfer on drive, activate drive and controller if idle */
s = splbio();
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bufqdisksort(&drive_queue[lunit], bp);
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
if (wdutab[lunit].b_active == 0)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wdustart(du); /* start drive */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/* Pick up changes made by readdisklabel(). */
if (du->dk_flags & DKFL_LABELLING && du->dk_state > RECAL) {
wdsleep(du->dk_ctrlr, "wdlab");
du->dk_state = WANTOPEN;
}
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
if (wdtab[du->dk_ctrlr_cmd640].b_active == 0)
#else
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (wdtab[du->dk_ctrlr].b_active == 0)
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wdstart(du->dk_ctrlr); /* start controller */
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (du->dk_dkunit >= 0) {
/*
* XXX perhaps we should only count successful transfers.
*/
dk_xfer[du->dk_dkunit]++;
/*
* XXX we can't count seeks correctly but we can do better
* than this. E.g., assume that the geometry is correct
* and count 1 seek if the starting cylinder of this i/o
* differs from the starting cylinder of the previous i/o,
* or count 1 seek if the starting bn of this i/o doesn't
* immediately follow the ending bn of the previos i/o.
*/
dk_seek[du->dk_dkunit]++;
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
splx(s);
return;
done:
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
s = splbio();
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* toss transfer, we're done early */
biodone(bp);
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
splx(s);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
static void
wdstrategy1(struct buf *bp)
{
/*
* XXX - do something to make wdstrategy() but not this block while
* we're doing dsinit() and dsioctl().
*/
wdstrategy(bp);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* Routine to queue a command to the controller. The unit's
* request is linked into the active list for the controller.
* If the controller is idle, the transfer is started.
*/
static void
wdustart(register struct disk *du)
{
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
register struct buf *bp;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
int ctrlr = du->dk_ctrlr_cmd640;
#else
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
int ctrlr = du->dk_ctrlr;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* unit already active? */
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
if (wdutab[du->dk_lunit].b_active)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return;
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bp = bufq_first(&drive_queue[du->dk_lunit]);
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
if (bp == NULL) { /* yes, an assign */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return;
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
}
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bufq_remove(&drive_queue[du->dk_lunit], bp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* link onto controller queue */
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bufq_insert_tail(&wdtab[ctrlr].controller_queue, bp);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* mark the drive unit as busy */
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
wdutab[du->dk_lunit].b_active = 1;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/*
* Controller startup routine. This does the calculation, and starts
* a single-sector read or write operation. Called to start a transfer,
* or from the interrupt routine to continue a multi-sector transfer.
* RESTRICTIONS:
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
* 1. The transfer length must be an exact multiple of the sector size.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
void
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdstart(int ctrlr)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
{
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
register struct disk *du;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
register struct buf *bp;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
struct diskgeom *lp; /* XXX sic */
long blknum;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
long secpertrk, secpercyl;
int lunit;
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
int count;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
int ctrlr_atapi;
if (eide_quirks & Q_CMD640B) {
ctrlr = PRIMARY;
ctrlr_atapi = atapictrlr;
} else {
ctrlr_atapi = ctrlr;
}
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Make ATAPI support completely optional - if ATAPI isn't defined, the original code/algorithm is used completely (with one bugfix). This makes it "safe" to bring these changes into -stable.
1995-09-30 15:19:44 +00:00
#ifdef ATAPI
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
if (wdtab[ctrlr].b_active == 2)
wdtab[ctrlr].b_active = 0;
if (wdtab[ctrlr].b_active)
return;
Make ATAPI support completely optional - if ATAPI isn't defined, the original code/algorithm is used completely (with one bugfix). This makes it "safe" to bring these changes into -stable.
1995-09-30 15:19:44 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* is there a drive for the controller to do a transfer with? */
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bp = bufq_first(&wdtab[ctrlr].controller_queue);
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
if (bp == NULL) {
#ifdef ATAPI
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
if (atapi_start && atapi_start (ctrlr_atapi))
wdtab[ctrlr].b_active = 3;
#else
Latest fixes from Serge: I tried to solve the problem of IDE probing compatibility in this version. When compiled without an ATAPI option, the wd driver is fully backward compatible with 2.0.5. With ATAPI option, the wdprobe becomes strictly weaker. That is, if wdprobe works without ATAPI option, it will always work with it too. Another problem was with the CD-ROM drive attached as a slave in the IDE bus, where there is no master. All IDE CD-ROM drives are shipped in slave configuration, and most users just plug them in, never thinking about jumpers. It works fine with ms-dos and ms-windows, and this version of the driver supports it as well. The eject op can now load disks. Just repeat it twice, and the disk will be ejected and then loaded back. The disc cannot be ejected if it is mounted. Submitted by: Serge Vakulenko, <vak@cronyx.ru>
1995-10-14 15:41:10 +00:00
if (atapi_start && atapi_start (ctrlr))
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
/* mark controller active in ATAPI mode */
wdtab[ctrlr].b_active = 3;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
#endif
Correct the location of a "return". This can/will cause system crashes.
1995-09-06 05:06:18 +00:00
return;
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* obtain controller and drive information */
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
lunit = dkunit(bp->b_dev);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du = wddrives[lunit];
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* if not really a transfer, do control operations specially */
if (du->dk_state < OPEN) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (du->dk_state != WANTOPEN)
printf("wd%d: wdstart: weird dk_state %d\n",
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_lunit, du->dk_state);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (wdcontrol(bp) != 0)
printf("wd%d: wdstart: wdcontrol returned nonzero, state = %d\n",
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_lunit, du->dk_state);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return;
}
/* calculate transfer details */
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
blknum = bp->b_pblkno + du->dk_skip;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#ifdef WDDEBUG
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (du->dk_skip == 0)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
printf("wd%d: wdstart: %s %d@%d; map ", lunit,
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
(bp->b_flags & B_READ) ? "read" : "write",
bp->b_bcount, blknum);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
else
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
printf(" %d)%x", du->dk_skip, inb(du->dk_altport));
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif
lp = &du->dk_dd;
secpertrk = lp->d_nsectors;
secpercyl = lp->d_secpercyl;
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
if (du->dk_skip == 0) {
du->dk_bc = bp->b_bcount;
Add semi-bogus initialization of dk_wpms so that systat and friends actually do something useful with the disk stats. (This needs to be determined dynamically, but I don't want to screw with this driver.)
1994-10-16 05:02:37 +00:00
Detect the "EOF" condition. Specifically, end of partition. Submitted by: Bruce Evans
1994-08-08 13:53:55 +00:00
if (bp->b_flags & B_BAD
/*
* XXX handle large transfers inefficiently instead
* of crashing on them.
*/
|| howmany(du->dk_bc, DEV_BSIZE) > MAXTRANSFER)
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
du->dk_flags |= DKFL_SINGLE;
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (du->dk_flags & DKFL_SINGLE
&& dsgetbad(bp->b_dev, du->dk_slices) != NULL) {
/* XXX */
u_long ds_offset =
du->dk_slices->dss_slices[dkslice(bp->b_dev)].ds_offset;
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
blknum = transbad144(dsgetbad(bp->b_dev, du->dk_slices),
blknum - ds_offset) + ds_offset;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wdtab[ctrlr].b_active = 1; /* mark controller active */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* if starting a multisector transfer, or doing single transfers */
if (du->dk_skip == 0 || (du->dk_flags & DKFL_SINGLE)) {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
u_int command;
u_int count;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
long cylin, head, sector;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
cylin = blknum / secpercyl;
head = (blknum % secpercyl) / secpertrk;
sector = blknum % secpertrk;
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
/*
* XXX this looks like an attempt to skip bad sectors
* on write.
*/
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (wdtab[ctrlr].b_errcnt && (bp->b_flags & B_READ) == 0)
du->dk_bc += DEV_BSIZE;
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
count = howmany( du->dk_bc, DEV_BSIZE);
du->dk_flags &= ~DKFL_MULTI;
Reverting wd driver back to version before Bruces new driver until the many bugs can be worked out of it...
1993-09-07 02:08:51 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#ifdef B_FORMAT
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if (bp->b_flags & B_FORMAT) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
command = WDCC_FORMAT;
count = lp->d_nsectors;
sector = lp->d_gap3 - 1; /* + 1 later */
} else
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
{
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
if (du->dk_flags & DKFL_SINGLE) {
command = (bp->b_flags & B_READ)
? WDCC_READ : WDCC_WRITE;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
count = 1;
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
du->dk_currentiosize = 1;
} else {
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
if((du->dk_flags & DKFL_USEDMA) &&
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
wddma[du->dk_interface].wdd_dmaverify(du->dk_dmacookie,
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
(void *)((int)bp->b_un.b_addr +
du->dk_skip * DEV_BSIZE),
du->dk_bc,
bp->b_flags & B_READ)) {
du->dk_flags |= DKFL_DMA;
if( bp->b_flags & B_READ)
command = WDCC_READ_DMA;
else
command = WDCC_WRITE_DMA;
du->dk_currentiosize = count;
} else if( (count > 1) && (du->dk_multi > 1)) {
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
du->dk_flags |= DKFL_MULTI;
if( bp->b_flags & B_READ) {
command = WDCC_READ_MULTI;
} else {
command = WDCC_WRITE_MULTI;
}
du->dk_currentiosize = du->dk_multi;
if( du->dk_currentiosize > count)
du->dk_currentiosize = count;
} else {
if( bp->b_flags & B_READ) {
command = WDCC_READ;
} else {
command = WDCC_WRITE;
}
du->dk_currentiosize = 1;
}
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Reverting wd driver back to version before Bruces new driver until the many bugs can be worked out of it...
1993-09-07 02:08:51 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/*
* XXX this loop may never terminate. The code to handle
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
* counting down of retries and eventually failing the i/o
* is in wdintr() and we can't get there from here.
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
*/
if (wdtest != 0) {
if (--wdtest == 0) {
wdtest = 100;
printf("dummy wdunwedge\n");
wdunwedge(du);
}
}
Add interface to new device configuration table. Also implement transfer statistics for iostat, vmstat, and systat.
1994-10-16 03:50:36 +00:00
if(du->dk_dkunit >= 0) {
dk_busy |= 1 << du->dk_dkunit;
}
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
if ((du->dk_flags & (DKFL_DMA|DKFL_SINGLE)) == DKFL_DMA) {
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
wddma[du->dk_interface].wdd_dmaprep(du->dk_dmacookie,
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
(void *)((int)bp->b_un.b_addr +
du->dk_skip * DEV_BSIZE),
du->dk_bc,
bp->b_flags & B_READ);
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
while (wdcommand(du, cylin, head, sector, count, command)
!= 0) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wderror(bp, du,
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
"wdstart: timeout waiting to give command");
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdunwedge(du);
Date: Fri, 3 Sep 93 01:25:44 +1000 From: bde@kralizec.zeta.org.au (Bruce Evans) This fixes several problems. It has almost everything that is in the FreeBSD version of Aug 22. It is missing the bug for 2 drives and still prints <> around the drive type. It handles BADEND144 less verbosely. It does wdgetctlr() somewhat differently from NetBSD/FreeBSD. Date: Sun Sep 5 06:45:12 PDT 1993 From: rgrimes@cdrom.com Fixed printf's to be correct, since Bruce used the old ones. This driver still prints phantom wd1's on some Maxtor and other IDE drives, I have a patch I am sending to Bruce for checking.
1993-09-05 13:46:17 +00:00
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#ifdef WDDEBUG
printf("cylin %ld head %ld sector %ld addr %x sts %x\n",
cylin, head, sector,
(int)bp->b_un.b_addr + du->dk_skip * DEV_BSIZE,
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
inb(du->dk_altport));
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/*
* Schedule wdtimeout() to wake up after a few seconds. Retrying
* unmarked bad blocks can take 3 seconds! Then it is not good that
* we retry 5 times.
*
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
* On the first try, we give it 10 seconds, for drives that may need
* to spin up.
*
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
* XXX wdtimeout() doesn't increment the error count so we may loop
* forever. More seriously, the loop isn't forever but causes a
* crash.
*
* TODO fix b_resid bug elsewhere (fd.c....). Fix short but positive
* counts being discarded after there is an error (in physio I
* think). Discarding them would be OK if the (special) file offset
* was not advanced.
*/
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
if (wdtab[ctrlr].b_errcnt == 0)
du->dk_timeout = 1 + 10;
else
du->dk_timeout = 1 + 3;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
/* if this is a DMA op, start DMA and go away until it's done. */
if ((du->dk_flags & (DKFL_DMA|DKFL_SINGLE)) == DKFL_DMA) {
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
wddma[du->dk_interface].wdd_dmastart(du->dk_dmacookie);
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
return;
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* If this is a read operation, just go away until it's done. */
if (bp->b_flags & B_READ)
return;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Ready to send data? */
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (wdwait(du, WDCS_READY | WDCS_SEEKCMPLT | WDCS_DRQ, TIMEOUT) < 0) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wderror(bp, du, "wdstart: timeout waiting for DRQ");
/*
* XXX what do we do now? If we've just issued the command,
* then we can treat this failure the same as a command
* failure. But if we are continuing a multi-sector write,
* the command was issued ages ago, so we can't simply
* restart it.
*
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
* XXX we waste a lot of time unnecessarily translating block
* numbers to cylin/head/sector for continued i/o's.
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
*/
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
count = 1;
if( du->dk_flags & DKFL_MULTI) {
count = howmany(du->dk_bc, DEV_BSIZE);
if( count > du->dk_multi)
count = du->dk_multi;
if( du->dk_currentiosize > count)
du->dk_currentiosize = count;
}
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
if (du->dk_flags & DKFL_32BIT)
outsl(du->dk_port + wd_data,
(void *)((int)bp->b_un.b_addr + du->dk_skip * DEV_BSIZE),
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
(count * DEV_BSIZE) / sizeof(long));
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
else
outsw(du->dk_port + wd_data,
(void *)((int)bp->b_un.b_addr + du->dk_skip * DEV_BSIZE),
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
(count * DEV_BSIZE) / sizeof(short));
du->dk_bc -= DEV_BSIZE * count;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (du->dk_dkunit >= 0) {
/*
* `wd's are blocks of 32 16-bit `word's according to
* iostat. dk_wds[] is the one disk i/o statistic that
* we can record correctly.
* XXX perhaps we shouldn't record words for failed
* transfers.
*/
dk_wds[du->dk_dkunit] += (count * DEV_BSIZE) >> 6;
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/* Interrupt routine for the controller. Acknowledge the interrupt, check for
* errors on the current operation, mark it done if necessary, and start
* the next request. Also check for a partially done transfer, and
* continue with the next chunk if so.
*/
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
void
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdintr(int unit)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
{
register struct disk *du;
Changes to support 1Tb filesizes. Pages are now named by an (object,index) pair instead of (object,offset) pair.
1995-12-11 04:58:34 +00:00
register struct buf *bp;
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
int dmastat;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
int ctrlr_atapi;
if (eide_quirks & Q_CMD640B) {
unit = PRIMARY;
ctrlr_atapi = atapictrlr;
} else {
ctrlr_atapi = unit;
}
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if (wdtab[unit].b_active == 2)
return; /* intr in wdflushirq() */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (!wdtab[unit].b_active) {
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
#ifdef WDDEBUG
/*
* These happen mostly because the power-mgt part of the
* bios shuts us down, and we just manage to see the
* interrupt from the "SLEEP" command.
*/
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
printf("wdc%d: extra interrupt\n", unit);
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return;
}
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
#ifdef ATAPI
if (wdtab[unit].b_active == 3) {
/* process an ATAPI interrupt */
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
if (atapi_intr && atapi_intr (ctrlr_atapi))
#else
Latest fixes from Serge: I tried to solve the problem of IDE probing compatibility in this version. When compiled without an ATAPI option, the wd driver is fully backward compatible with 2.0.5. With ATAPI option, the wdprobe becomes strictly weaker. That is, if wdprobe works without ATAPI option, it will always work with it too. Another problem was with the CD-ROM drive attached as a slave in the IDE bus, where there is no master. All IDE CD-ROM drives are shipped in slave configuration, and most users just plug them in, never thinking about jumpers. It works fine with ms-dos and ms-windows, and this version of the driver supports it as well. The eject op can now load disks. Just repeat it twice, and the disk will be ejected and then loaded back. The disc cannot be ejected if it is mounted. Submitted by: Serge Vakulenko, <vak@cronyx.ru>
1995-10-14 15:41:10 +00:00
if (atapi_intr && atapi_intr (unit))
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
/* ATAPI op continues */
return;
/* controller is free, start new op */
wdtab[unit].b_active = 0;
wdstart (unit);
return;
}
#endif
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bp = bufq_first(&wdtab[unit].controller_queue);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
du = wddrives[dkunit(bp->b_dev)];
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
/* finish off DMA */
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
if (du->dk_flags & (DKFL_DMA|DKFL_USEDMA)) {
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
/* XXX SMP boxes sometimes generate an early intr. Why? */
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
if ((wddma[du->dk_interface].wdd_dmastatus(du->dk_dmacookie) & WDDS_INTERRUPT)
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
== 0)
Make the WD code work on my P6DNF running on the SMP code. It appears that there was an "early" interrupt, and this checks for it.
1997-08-04 05:26:49 +00:00
return;
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
dmastat = wddma[du->dk_interface].wdd_dmadone(du->dk_dmacookie);
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
}
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
du->dk_timeout = 0;
/* check drive status/failure */
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (wdwait(du, 0, TIMEOUT) < 0) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wderror(bp, du, "wdintr: timeout waiting for status");
du->dk_status |= WDCS_ERR; /* XXX */
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* is it not a transfer, but a control operation? */
if (du->dk_state < OPEN) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdtab[unit].b_active = 0;
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
switch (wdcontrol(bp)) {
case 0:
return;
case 1:
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdstart(unit);
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
return;
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
case 2:
goto done;
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/* have we an error? */
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
if ((du->dk_status & (WDCS_ERR | WDCS_ECCCOR))
|| (((du->dk_flags & (DKFL_DMA|DKFL_SINGLE)) == DKFL_DMA)
&& dmastat != WDDS_INTERRUPT)) {
unsigned int errstat;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
oops:
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
/*
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
* XXX bogus inb() here
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
*/
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
errstat = inb(du->dk_port + wd_error);
if(((du->dk_flags & (DKFL_DMA|DKFL_SINGLE)) == DKFL_DMA) &&
(errstat & WDERR_ABORT)) {
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
wderror(bp, du, "reverting to PIO mode");
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
du->dk_flags &= ~DKFL_USEDMA;
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
} else if((du->dk_flags & DKFL_MULTI) &&
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
(errstat & WDERR_ABORT)) {
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
wderror(bp, du, "reverting to non-multi sector mode");
du->dk_multi = 1;
}
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
if (!(du->dk_status & (WDCS_ERR | WDCS_ECCCOR)) &&
(((du->dk_flags & (DKFL_DMA|DKFL_SINGLE)) == DKFL_DMA) &&
(dmastat != WDDS_INTERRUPT)))
printf("wd%d: DMA failure, DMA status %b\n",
du->dk_lunit, dmastat, WDDS_BITS);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#ifdef WDDEBUG
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wderror(bp, du, "wdintr");
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if ((du->dk_flags & DKFL_SINGLE) == 0) {
du->dk_flags |= DKFL_ERROR;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
goto outt;
}
#ifdef B_FORMAT
if (bp->b_flags & B_FORMAT) {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
bp->b_error = EIO;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
bp->b_flags |= B_ERROR;
goto done;
}
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
Changes to support the "new" bad144. Adds a "BADSCAN" ioctl and support. Reviewed by: Bruce Evans Submitted by: Poul-Henning Kamp
1995-05-16 07:52:17 +00:00
if (du->dk_flags & DKFL_BADSCAN) {
bp->b_error = EIO;
bp->b_flags |= B_ERROR;
} else if (du->dk_status & WDCS_ERR) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (++wdtab[unit].b_errcnt < RETRIES) {
wdtab[unit].b_active = 0;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
} else {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wderror(bp, du, "hard error");
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
bp->b_error = EIO;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
bp->b_flags |= B_ERROR; /* flag the error */
}
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
} else if (du->dk_status & WDCS_ECCCOR)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wderror(bp, du, "soft ecc");
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/*
* If this was a successful read operation, fetch the data.
*/
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if (((bp->b_flags & (B_READ | B_ERROR)) == B_READ)
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
&& !((du->dk_flags & (DKFL_DMA|DKFL_SINGLE)) == DKFL_DMA)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
&& wdtab[unit].b_active) {
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
int chk, dummy, multisize;
multisize = chk = du->dk_currentiosize * DEV_BSIZE;
if( du->dk_bc < chk) {
chk = du->dk_bc;
if( ((chk + DEV_BSIZE - 1) / DEV_BSIZE) < du->dk_currentiosize) {
du->dk_currentiosize = (chk + DEV_BSIZE - 1) / DEV_BSIZE;
multisize = du->dk_currentiosize * DEV_BSIZE;
}
}
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* ready to receive data? */
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
if ((du->dk_status & (WDCS_READY | WDCS_SEEKCMPLT | WDCS_DRQ))
!= (WDCS_READY | WDCS_SEEKCMPLT | WDCS_DRQ))
wderror(bp, du, "wdintr: read intr arrived early");
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (wdwait(du, WDCS_READY | WDCS_SEEKCMPLT | WDCS_DRQ, TIMEOUT) != 0) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wderror(bp, du, "wdintr: read error detected late");
goto oops;
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* suck in data */
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
if( du->dk_flags & DKFL_32BIT)
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
insl(du->dk_port + wd_data,
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
(void *)((int)bp->b_un.b_addr + du->dk_skip * DEV_BSIZE),
chk / sizeof(long));
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
else
insw(du->dk_port + wd_data,
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
(void *)((int)bp->b_un.b_addr + du->dk_skip * DEV_BSIZE),
chk / sizeof(short));
du->dk_bc -= chk;
/* XXX for obsolete fractional sector reads. */
while (chk < multisize) {
insw(du->dk_port + wd_data, &dummy, 1);
chk += sizeof(short);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (du->dk_dkunit >= 0)
dk_wds[du->dk_dkunit] += chk >> 6;
Add interface to new device configuration table. Also implement transfer statistics for iostat, vmstat, and systat.
1994-10-16 03:50:36 +00:00
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
/* final cleanup on DMA */
if (((bp->b_flags & B_ERROR) == 0)
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
&& ((du->dk_flags & (DKFL_DMA|DKFL_SINGLE)) == DKFL_DMA)
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
&& wdtab[unit].b_active) {
int iosize;
iosize = du->dk_currentiosize * DEV_BSIZE;
du->dk_bc -= iosize;
if (du->dk_dkunit >= 0)
dk_wds[du->dk_dkunit] += iosize >> 6;
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
outt:
if (wdtab[unit].b_active) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if ((bp->b_flags & B_ERROR) == 0) {
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
du->dk_skip += du->dk_currentiosize;/* add to successful sectors */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (wdtab[unit].b_errcnt)
wderror(bp, du, "soft error");
wdtab[unit].b_errcnt = 0;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* see if more to transfer */
if (du->dk_bc > 0 && (du->dk_flags & DKFL_ERROR) == 0) {
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
if( (du->dk_flags & DKFL_SINGLE) ||
((bp->b_flags & B_READ) == 0)) {
wdtab[unit].b_active = 0;
wdstart(unit);
} else {
du->dk_timeout = 1 + 3;
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return; /* next chunk is started */
} else if ((du->dk_flags & (DKFL_SINGLE | DKFL_ERROR))
== DKFL_ERROR) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
du->dk_skip = 0;
du->dk_flags &= ~DKFL_ERROR;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_flags |= DKFL_SINGLE;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdtab[unit].b_active = 0;
wdstart(unit);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return; /* redo xfer sector by sector */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
}
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
done: ;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* done with this transfer, with or without error */
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
du->dk_flags &= ~(DKFL_SINGLE|DKFL_DMA);
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
bufq_remove( &wdtab[unit].controller_queue, bp);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdtab[unit].b_errcnt = 0;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
bp->b_resid = bp->b_bcount - du->dk_skip * DEV_BSIZE;
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
wdutab[du->dk_lunit].b_active = 0;
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
du->dk_skip = 0;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
biodone(bp);
}
Add interface to new device configuration table. Also implement transfer statistics for iostat, vmstat, and systat.
1994-10-16 03:50:36 +00:00
if(du->dk_dkunit >= 0) {
dk_busy &= ~(1 << du->dk_dkunit);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* controller idle */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdtab[unit].b_active = 0;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* anything more on drive queue? */
The big 4.4BSD Lite to FreeBSD 2.0.0 (Development) patch. Reviewed by: Rodney W. Grimes Submitted by: John Dyson and David Greenman
1994-05-25 09:21:21 +00:00
wdustart(du);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* anything more for controller to do? */
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
#ifndef ATAPI
/* This is not valid in ATAPI mode. */
aha1542.c aic6360.c cy.c fd.c ft.c if_ie.c if_wl.c if_zp.c isa.c isa_device.h labpc.c mcd.c ncr5380.c scd.c seagate.c si.c sio.c tw.c ultra14f.c wcd.c wd.c: Update for changes in the callout interface. apic_vector.s icu_vector.s ipl.s ipl_funcs.c: Add CAM software/hardware interrupt support.
1997-09-21 21:41:49 +00:00
if (bufq_first(&wdtab[unit].controller_queue) != NULL)
Bring in Serge Vakulenko's IDE CDROM (ATAPI) driver. A number of people have now indicated to me that it's working more than well enough to bring into -current. Submitted by: Serge Vakulenko <vak@cronyx.ru>
1995-08-18 11:26:35 +00:00
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdstart(unit);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/*
* Initialize a drive.
*/
int
wdopen(dev_t dev, int flags, int fmt, struct proc *p)
{
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
register unsigned int lunit;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
register struct disk *du;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
int error;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
lunit = dkunit(dev);
if (lunit >= NWD || dktype(dev) != 0)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (ENXIO);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du = wddrives[lunit];
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if (du == NULL)
return (ENXIO);
/* Finish flushing IRQs left over from wdattach(). */
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
if (wdtab[du->dk_ctrlr_cmd640].b_active == 2)
wdtab[du->dk_ctrlr_cmd640].b_active = 0;
#else
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if (wdtab[du->dk_ctrlr].b_active == 2)
wdtab[du->dk_ctrlr].b_active = 0;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
Changes to support the "new" bad144. Adds a "BADSCAN" ioctl and support. Reviewed by: Bruce Evans Submitted by: Poul-Henning Kamp
1995-05-16 07:52:17 +00:00
du->dk_flags &= ~DKFL_BADSCAN;
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
/* spin waiting for anybody else reading the disk label */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
while (du->dk_flags & DKFL_LABELLING)
tsleep((caddr_t)&du->dk_flags, PZERO - 1, "wdopen", 1);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#if 1
wdsleep(du->dk_ctrlr, "wdopn1");
du->dk_flags |= DKFL_LABELLING;
du->dk_state = WANTOPEN;
{
struct disklabel label;
bzero(&label, sizeof label);
label.d_secsize = du->dk_dd.d_secsize;
label.d_nsectors = du->dk_dd.d_nsectors;
label.d_ntracks = du->dk_dd.d_ntracks;
label.d_ncylinders = du->dk_dd.d_ncylinders;
label.d_secpercyl = du->dk_dd.d_secpercyl;
label.d_secperunit = du->dk_dd.d_secperunit;
error = dsopen("wd", dev, fmt, &du->dk_slices, &label, wdstrategy1,
First attempt at creating devfs entries for sliced devices. Doesn't quite work yet, so the heart of it is disabled. Added bdev and cdev args to dsopen(). drivers: Fixed device names, links, minor numbers and modes. wd.c: Started actually supporting devfs. diskslice.h: Added devfs tokens to structs (currently 576 of them per disk! :-(). subr_diskslice.c: Create devfs entries in dsopen() and (unsuccessfully) attempt to make them go away at the right times. DEVFS is #undefed at the start so that this shouldn't cause problems.
1996-01-27 04:18:15 +00:00
(ds_setgeom_t *)NULL, &wd_bdevsw, &wd_cdevsw);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
}
du->dk_flags &= ~DKFL_LABELLING;
wdsleep(du->dk_ctrlr, "wdopn2");
return (error);
#else
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if ((du->dk_flags & DKFL_BSDLABEL) == 0) {
Date: Fri, 3 Sep 93 01:25:44 +1000 From: bde@kralizec.zeta.org.au (Bruce Evans) This fixes several problems. It has almost everything that is in the FreeBSD version of Aug 22. It is missing the bug for 2 drives and still prints <> around the drive type. It handles BADEND144 less verbosely. It does wdgetctlr() somewhat differently from NetBSD/FreeBSD. Date: Sun Sep 5 06:45:12 PDT 1993 From: rgrimes@cdrom.com Fixed printf's to be correct, since Bruce used the old ones. This driver still prints phantom wd1's on some Maxtor and other IDE drives, I have a patch I am sending to Bruce for checking.
1993-09-05 13:46:17 +00:00
/*
Whoops, shouldn't really have been here, but what the heck: remove some stale comments.
1996-04-18 21:37:43 +00:00
* wdtab[ctrlr].b_active != 0 implies XXX applicable now ??
Move from the old buf.b_actf to the new TAILQ(buf.b_act).
1996-05-03 14:57:27 +00:00
* drive_queue[lunit].b_act == NULL (?) XXX applicable now ??
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
* so the following guards most things (until the next i/o).
* It doesn't guard against a new i/o starting and being
* affected by the label being changed. Sigh.
Date: Fri, 3 Sep 93 01:25:44 +1000 From: bde@kralizec.zeta.org.au (Bruce Evans) This fixes several problems. It has almost everything that is in the FreeBSD version of Aug 22. It is missing the bug for 2 drives and still prints <> around the drive type. It handles BADEND144 less verbosely. It does wdgetctlr() somewhat differently from NetBSD/FreeBSD. Date: Sun Sep 5 06:45:12 PDT 1993 From: rgrimes@cdrom.com Fixed printf's to be correct, since Bruce used the old ones. This driver still prints phantom wd1's on some Maxtor and other IDE drives, I have a patch I am sending to Bruce for checking.
1993-09-05 13:46:17 +00:00
*/
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdsleep(du->dk_ctrlr, "wdopn1");
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
du->dk_flags |= DKFL_LABELLING;
Reverting wd driver back to version before Bruces new driver until the many bugs can be worked out of it...
1993-09-07 02:08:51 +00:00
du->dk_state = WANTOPEN;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
error = dsinit(dkmodpart(dev, RAW_PART), wdstrategy,
&du->dk_dd, &du->dk_slices);
if (error != 0) {
du->dk_flags &= ~DKFL_LABELLING;
return (error);
}
/* XXX check value returned by wdwsetctlr(). */
wdwsetctlr(du);
if (dkslice(dev) == WHOLE_DISK_SLICE) {
dsopen(dev, fmt, du->dk_slices);
return (0);
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/*
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
* Read label using RAW_PART partition.
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
*
* If the drive has an MBR, then the current geometry (from
* wdgetctlr()) is used to read it; then the BIOS/DOS
* geometry is inferred and used to read the label off the
* 'c' partition. Otherwise the label is read using the
* current geometry. The label gives the final geometry.
* If bad sector handling is enabled, then this geometry
* is used to read the bad sector table. The geometry
* changes occur inside readdisklabel() and are propagated
* to the driver by resetting the state machine.
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
*
* XXX can now handle changes directly since dsinit() doesn't
* do too much.
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
*/
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
msg = correct_readdisklabel(dkmodpart(dev, RAW_PART), wdstrategy,
&du->dk_dd);
/* XXX check value returned by wdwsetctlr(). */
wdwsetctlr(du);
if (msg == NULL && du->dk_dd.d_flags & D_BADSECT)
msg = readbad144(dkmodpart(dev, RAW_PART), wdstrategy,
&du->dk_dd, &du->dk_bad);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du->dk_flags &= ~DKFL_LABELLING;
if (msg != NULL) {
Fixed for DQFLQUIET flag from: Terry Lee <terry@uivlsisd.csl.uiuc.edu> I submitted the part of patch00155 that fixes wd.c not reporting any bad blocks. Unfortunately, the patch created a new problem that makes new installations using that kernel impossible. I have included a patch below that should fix both problems. The fix comes from Tom Ivar Helbekkmo's wd driver (the Norway wd driver). I've also eliminated the use of the DKFL_QUIET flag completely, since it is not used anymore. Fixed wdgetctrl routines for so that it returns a meaniful string for non IDE disk drives. It also nolonger returns success when a drive is not really there.
1993-08-08 07:03:21 +00:00
log(LOG_WARNING, "wd%d: cannot find label (%s)\n",
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
lunit, msg);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (part != RAW_PART)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (EINVAL); /* XXX needs translation */
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
/*
* Soon return. This is how slices without labels
* are allowed. They only work on the raw partition.
*/
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
} else {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
unsigned long newsize, offset, size;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#if 0
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/*
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
* Force RAW_PART partition to be the whole disk.
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
*/
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
offset = du->dk_dd.d_partitions[RAW_PART].p_offset;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if (offset != 0) {
printf(
Change WDRAW back to 3, too early step right now, disklabels needs more work Change hardcoded 'd' in printfs to 'a' + WDPART
1994-08-30 14:26:13 +00:00
"wd%d: changing offset of '%c' partition from %lu to 0\n",
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
du->dk_lunit, 'a' + RAW_PART, offset);
du->dk_dd.d_partitions[RAW_PART].p_offset = 0;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
size = du->dk_dd.d_partitions[RAW_PART].p_size;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
newsize = du->dk_dd.d_secperunit; /* XXX */
if (size != newsize) {
printf(
Change WDRAW back to 3, too early step right now, disklabels needs more work Change hardcoded 'd' in printfs to 'a' + WDPART
1994-08-30 14:26:13 +00:00
"wd%d: changing size of '%c' partition from %lu to %lu\n",
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
du->dk_lunit, 'a' + RAW_PART, size,
newsize);
du->dk_dd.d_partitions[RAW_PART].p_size
= newsize;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/* Pick up changes made by readdisklabel(). */
wdsleep(du->dk_ctrlr, "wdopn2");
du->dk_state = WANTOPEN;
Reverting wd driver back to version before Bruces new driver until the many bugs can be worked out of it...
1993-09-07 02:08:51 +00:00
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/*
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
* Warn if a partion is opened that overlaps another partition which
* is open unless one is the "raw" partition (whole disk).
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
*/
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if ((du->dk_openpart & mask) == 0 && part != RAW_PART) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
int start, end;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
pp = &du->dk_dd.d_partitions[part];
start = pp->p_offset;
end = pp->p_offset + pp->p_size;
for (pp = du->dk_dd.d_partitions;
pp < &du->dk_dd.d_partitions[du->dk_dd.d_npartitions];
pp++) {
if (pp->p_offset + pp->p_size <= start ||
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
pp->p_offset >= end)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
continue;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (pp - du->dk_dd.d_partitions == RAW_PART)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
continue;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if (du->dk_openpart
& (1 << (pp - du->dk_dd.d_partitions)))
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
log(LOG_WARNING,
"wd%d%c: overlaps open partition (%c)\n",
lunit, part + 'a',
pp - du->dk_dd.d_partitions + 'a');
}
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (part >= du->dk_dd.d_npartitions && part != RAW_PART)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (ENXIO);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
dsopen(dev, fmt, du->dk_slices);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (0);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/*
* Implement operations other than read/write.
* Called from wdstart or wdintr during opens and formats.
* Uses finite-state-machine to track progress of operation in progress.
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
* Returns 0 if operation still in progress, 1 if completed, 2 if error.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
static int
wdcontrol(register struct buf *bp)
{
register struct disk *du;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
int ctrlr;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
du = wddrives[dkunit(bp->b_dev)];
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
ctrlr = du->dk_ctrlr_cmd640;
#else
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
ctrlr = du->dk_ctrlr;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
Reverting wd driver back to version before Bruces new driver until the many bugs can be worked out of it...
1993-09-07 02:08:51 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
switch (du->dk_state) {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
case WANTOPEN:
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
tryagainrecal:
wdtab[ctrlr].b_active = 1;
if (wdcommand(du, 0, 0, 0, 0, WDCC_RESTORE | WD_STEP) != 0) {
wderror(bp, du, "wdcontrol: wdcommand failed");
goto maybe_retry;
}
du->dk_state = RECAL;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
case RECAL:
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (du->dk_status & WDCS_ERR || wdsetctlr(du) != 0) {
wderror(bp, du, "wdcontrol: recal failed");
maybe_retry:
if (du->dk_status & WDCS_ERR)
wdunwedge(du);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_state = WANTOPEN;
if (++wdtab[ctrlr].b_errcnt < RETRIES)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
goto tryagainrecal;
bp->b_error = ENXIO; /* XXX needs translation */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
bp->b_flags |= B_ERROR;
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
return (2);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdtab[ctrlr].b_errcnt = 0;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
du->dk_state = OPEN;
/*
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
* The rest of the initialization can be done by normal
* means.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (1);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
panic("wdcontrol");
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
return (2);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/*
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
* Wait uninterruptibly until controller is not busy, then send it a command.
* The wait usually terminates immediately because we waited for the previous
* command to terminate.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
static int
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdcommand(struct disk *du, u_int cylinder, u_int head, u_int sector,
u_int count, u_int command)
Stop gap measure until we can get Bruce's driver debugged. --- From: sos@login.dkuug.dk (S|ren Schmidt) Subject: IDE-disk hangs - solution/patches NetBSD/FreeBSD Summary: fixes for lost interrupts with IDE disks Keywords: hanging-disk, IDE-disk, lost-interrupt Due to "popular" demand I'm posting these patches to NetBSD/FreeBSD instead of mailing them around the world :-) As many have found out there is a problem when using IDE disks on FreeBSD. Following is a patch that fixes the problem with lost intterrupts. Both fixes is based on a patch posted here some month ago by Stefan Behrens?? (sorry I've lost the original article). But anyway it works (for me :-). Basically it does a timeout on lost interrupts, starting the operation again and logging and error message on the console. It additionally makes the allready present while loop timeouts independent of CPU speed, and adds minor numbers for easy access to dos partitions.
1993-11-22 23:25:46 +00:00
{
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
u_int wdc;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
wdc = du->dk_port;
If flag 0x4000 (Sleephack) is defined, assume that the disk is sleeping if all registers are 0xff. This allows me to run with flags 0xc0ff on my IBM-DMCA-21440 disk, which gives 5MB/sec sequential read :-) If you have a laptop, try adding flag 0x4000 to your disk, and tell me if it makes any difference for you.
1997-04-03 09:43:50 +00:00
if (du->cfg_flags & WDOPT_SLEEPHACK) {
/* OK, so the APM bios has put the disk into SLEEP mode,
* how can we tell ? Uhm, we can't. There is no
* standardized way of finding out, and the only way to
* wake it up is to reset it. Bummer.
*
* All the many and varied versions of the IDE/ATA standard
* explicitly tells us not to look at these registers if
* the disk is in SLEEP mode. Well, too bad really, we
* have to find out if it's in sleep mode before we can
* avoid reading the registers.
*
* I have reason to belive that most disks will return
* either 0xff or 0x00 in all but the status register
* when in SLEEP mode, but I have yet to see one return
* 0x00, so we don't check for that yet.
*
* The check for WDCS_BUSY is for the case where the
* bios spins up the disk for us, but doesn't initialize
* it correctly /phk
*/
if(inb(wdc + wd_precomp) + inb(wdc + wd_cyl_lo) +
inb(wdc + wd_cyl_hi) + inb(wdc + wd_sdh) +
inb(wdc + wd_sector) + inb(wdc + wd_seccnt) == 6 * 0xff) {
if (bootverbose)
printf("wd(%d,%d): disk aSLEEP\n",
du->dk_ctrlr, du->dk_unit);
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
wdunwedge(du);
If flag 0x4000 (Sleephack) is defined, assume that the disk is sleeping if all registers are 0xff. This allows me to run with flags 0xc0ff on my IBM-DMCA-21440 disk, which gives 5MB/sec sequential read :-) If you have a laptop, try adding flag 0x4000 to your disk, and tell me if it makes any difference for you.
1997-04-03 09:43:50 +00:00
} else if(inb(wdc + wd_status) == WDCS_BUSY) {
if (bootverbose)
printf("wd(%d,%d): disk is BUSY\n",
du->dk_ctrlr, du->dk_unit);
wdunwedge(du);
}
}
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (wdwait(du, 0, TIMEOUT) < 0)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (1);
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
if( command == WDCC_FEATURES) {
outb(wdc + wd_features, count);
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
if ( count == WDFEA_SETXFER )
outb(wdc + wd_seccnt, sector);
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
} else {
outb(wdc + wd_precomp, du->dk_dd.d_precompcyl / 4);
outb(wdc + wd_cyl_lo, cylinder);
outb(wdc + wd_cyl_hi, cylinder >> 8);
outb(wdc + wd_sdh, WDSD_IBM | (du->dk_unit << 4) | head);
outb(wdc + wd_sector, sector + 1);
outb(wdc + wd_seccnt, count);
}
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
if (wdwait(du, (command == WDCC_DIAGNOSE || command == WDCC_IDC)
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
? 0 : WDCS_READY, TIMEOUT) < 0)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (1);
outb(wdc + wd_command, command);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Another attempt at making multi-sector mode work.
1996-08-23 02:52:44 +00:00
static void
wdsetmulti(struct disk *du)
{
/*
* The config option flags low 8 bits define the maximum multi-block
* transfer size. If the user wants the maximum that the drive
* is capable of, just set the low bits of the config option to
* 0x00ff.
*/
if ((du->cfg_flags & WDOPT_MULTIMASK) != 0 && (du->dk_multi > 1)) {
int configval = du->cfg_flags & WDOPT_MULTIMASK;
du->dk_multi = min(du->dk_multi, configval);
if (wdcommand(du, 0, 0, 0, du->dk_multi, WDCC_SET_MULTI)) {
du->dk_multi = 1;
} else {
if (wdwait(du, WDCS_READY, TIMEOUT) < 0) {
du->dk_multi = 1;
}
}
} else {
du->dk_multi = 1;
}
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* issue IDC to drive to tell it just what geometry it is to be.
*/
static int
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdsetctlr(struct disk *du)
{
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
int error = 0;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#ifdef WDDEBUG
printf("wd(%d,%d): wdsetctlr: C %lu H %lu S %lu\n",
du->dk_ctrlr, du->dk_unit,
du->dk_dd.d_ncylinders, du->dk_dd.d_ntracks,
du->dk_dd.d_nsectors);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#endif
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
if (du->dk_dd.d_ntracks == 0 || du->dk_dd.d_ntracks > 16) {
struct wdparams *wp;
printf("wd%d: can't handle %lu heads from partition table ",
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_lunit, du->dk_dd.d_ntracks);
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
/* obtain parameters */
wp = &du->dk_params;
if (wp->wdp_heads > 0 && wp->wdp_heads <= 16) {
Fix -Wformat warnings, still need to do something about %b and pointer type args.
1995-05-09 12:26:00 +00:00
printf("(controller value %u restored)\n",
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
wp->wdp_heads);
du->dk_dd.d_ntracks = wp->wdp_heads;
}
else {
printf("(truncating to 16)\n");
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
du->dk_dd.d_ntracks = 16;
}
}
if (du->dk_dd.d_nsectors == 0 || du->dk_dd.d_nsectors > 255) {
printf("wd%d: cannot handle %lu sectors (max 255)\n",
du->dk_lunit, du->dk_dd.d_nsectors);
error = 1;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
}
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
if (error) {
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
wdtab[du->dk_ctrlr_cmd640].b_errcnt += RETRIES;
#else
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
wdtab[du->dk_ctrlr].b_errcnt += RETRIES;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
Bug fixes and performance improvements from John Dyson and myself: 1) check va before clearing the page clean flag. Not doing so was causing the vnode pager error 5 messages when paging from NFS. (pmap.c) 2) put back interrupt protection in idle_loop. Bruce didn't think it was necessary, John insists that it is (and I agree). (swtch.s) 3) various improvements to the clustering code (vm_machdep.c). It's now enabled/used by default. 4) bad disk blocks are now handled properly when doing clustered IOs. (wd.c, vm_machdep.c) 5) bogus bad block handling fixed in wd.c. 6) algorithm improvements to the pageout/pagescan daemons. It's amazing how well 4MB machines work now.
1994-04-20 07:06:57 +00:00
return (1);
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (wdcommand(du, du->dk_dd.d_ncylinders, du->dk_dd.d_ntracks - 1, 0,
du->dk_dd.d_nsectors, WDCC_IDC) != 0
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
|| wdwait(du, WDCS_READY, TIMEOUT) < 0) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wderror((struct buf *)NULL, du, "wdsetctlr failed");
return (1);
}
Move a couple of the initialization commands to the right place. Multi sector mode was not getting re-initialized when needed.
1996-07-27 19:01:10 +00:00
Another attempt at making multi-sector mode work.
1996-08-23 02:52:44 +00:00
wdsetmulti(du);
Move a couple of the initialization commands to the right place. Multi sector mode was not getting re-initialized when needed.
1996-07-27 19:01:10 +00:00
#ifdef NOTYET
/* set read caching and write caching */
wdcommand(du, 0, 0, 0, WDFEA_RCACHE, WDCC_FEATURES);
Another attempt at making multi-sector mode work.
1996-08-23 02:52:44 +00:00
wdwait(du, WDCS_READY, TIMEOUT);
Move a couple of the initialization commands to the right place. Multi sector mode was not getting re-initialized when needed.
1996-07-27 19:01:10 +00:00
wdcommand(du, 0, 0, 0, WDFEA_WCACHE, WDCC_FEATURES);
Another attempt at making multi-sector mode work.
1996-08-23 02:52:44 +00:00
wdwait(du, WDCS_READY, TIMEOUT);
Move a couple of the initialization commands to the right place. Multi sector mode was not getting re-initialized when needed.
1996-07-27 19:01:10 +00:00
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (0);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Clean up -Wunused warnings. Reviewed by: bde
1996-06-12 05:11:41 +00:00
#if 0
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/*
* Wait until driver is inactive, then set up controller.
*/
static int
wdwsetctlr(struct disk *du)
{
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
int stat;
int x;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
wdsleep(du->dk_ctrlr, "wdwset");
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
x = splbio();
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
stat = wdsetctlr(du);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wdflushirq(du, x);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
splx(x);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (stat);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
Clean up -Wunused warnings. Reviewed by: bde
1996-06-12 05:11:41 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
/*
* gross little callback function for wdddma interface. returns 1 for
* success, 0 for failure.
*/
static int
wdsetmode(int mode, void *wdinfo)
{
int i;
struct disk *du;
du = wdinfo;
if (bootverbose)
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
printf("wd%d: wdsetmode() setting transfer mode to %02x\n",
du->dk_lunit, mode);
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
i = wdcommand(du, 0, 0, mode, WDFEA_SETXFER,
WDCC_FEATURES) == 0 &&
wdwait(du, WDCS_READY, TIMEOUT) == 0;
return i;
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/*
* issue READP to drive to ask it what it is.
*/
static int
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
wdgetctlr(struct disk *du)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
{
int i;
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
char tb[DEV_BSIZE], tb2[DEV_BSIZE];
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
struct wdparams *wp = NULL;
Changes John's handling of the flags a little bit, added a new one, (0x4000 == sleephack) to handle disks which are confused after a sleep. Fixed a spelling error.
1995-04-24 05:12:29 +00:00
u_long flags = du->cfg_flags;
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
again:
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (wdcommand(du, 0, 0, 0, 0, WDCC_READP) != 0
1) My previous fix does nothing, now Rod's fix rewrited to implement my previous fix too (using wdp_heads controller value) and check 0 heads case too. Other fixes from Bruce: 2) Fix dk_timeout from 2 to about 4 seconds. 3) wdcontrol not retried on internal error. 4) wdwait return check changed ( "!=0" to "<0" into wdsetctlr, "<0" to "!=0" into wdgetctlr).
1994-03-04 16:43:07 +00:00
|| wdwait(du, WDCS_READY | WDCS_SEEKCMPLT | WDCS_DRQ, TIMEOUT) != 0) {
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
/*
* if we failed on the second try, assume non-32bit
*/
if( du->dk_flags & DKFL_32BIT)
goto failed;
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* XXX need to check error status after final transfer. */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/*
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
* Old drives don't support WDCC_READP. Try a seek to 0.
* Some IDE controllers return trash if there is no drive
* attached, so first test that the drive can be selected.
* This also avoids long waits for nonexistent drives.
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
*/
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (wdwait(du, 0, TIMEOUT) < 0)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (1);
outb(du->dk_port + wd_sdh, WDSD_IBM | (du->dk_unit << 4));
DELAY(5000); /* usually unnecessary; drive select is fast */
Apply my small patch to make detection of ATAPI CD-ROMs happen a little more reliably. So far I've received a couple of positive responses and no objections to these changes. There are two one-line changes: - In wdprobe(), when testing the error status of drives, don't unconditionally decide that there is no controller present if we read back a value of 0x81 (drive 0 okay, drive 1 failed) twice in a row. This may be caused by having an ATAPI CD-ROM jumpered as a master on the controller with no slave. - In wdgetctlr(), when checking for a status of WDCS_READY, check the value twice. The first time may be bogus. This stops a phantom wd2 device from being detected when an ATAPI CD-ROM is attached to the secondary controller alone as a slave. (This can cause installation to fail when sysinstall attempts to open the phantom device and wedges the system as a result. This has bitten me a couple of times on some Gateway 2000 machines.)
1996-08-12 00:53:02 +00:00
/*
* Do this twice: may get a false WDCS_READY the first time.
*/
inb(du->dk_port + wd_status);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
if ((inb(du->dk_port + wd_status) & (WDCS_BUSY | WDCS_READY))
!= WDCS_READY
|| wdcommand(du, 0, 0, 0, 0, WDCC_RESTORE | WD_STEP) != 0
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
|| wdwait(du, WDCS_READY | WDCS_SEEKCMPLT, TIMEOUT) != 0)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (1);
Load the kernel symbol table in the boot loader and not at compile time. (Boot with the -D flag if you want symbols.) Make it easier to extend `struct bootinfo' without losing either forwards or backwards compatibility. ddb_aout.c: Get the symbol table from wherever the loader put it. Nuke db_symtab[SYMTAB_SPACE]. boot.c: Enable loading of symbols. Align them on a page boundary. Add printfs about the symbol table sizes. Pass the memory sizes to the kernel. Fix initialization of `unit' (it got moved out of the loop). Fix adding the bss size (it got moved inside an ifdef). Initialize serial port when RB_SERIAL is toggled on. Fix comments. Clean up formatting of recently added code. io.c: Clean up formatting of recently added code. netboot/main.c, machdep.c, wd.c: Change names of bootinfo fields. LINT: Nuke SYMTAB_SPACE. Fix comment about DODUMP. Makefile.i386: Nuke use of dbsym. Exclude gcc symbols from kernel unless compiling with -g. Remove unused macro. Fix comments and formatting. genassym.c: Generate defines for some new bootinfo fields. Change names of old ones. locore.s: Copy only the valid part of the `struct bootinfo' passed by the loader. Reserve space for symbol table, if any. machdep.c: Check the memory sizes passed by the loader, if any. Don't use them yet. bootinfo.h: Add a size field so that we can resolve some mismatches between the loader bootinfo and the kernel boot info. The version number is not so good for this because of historical botches and because it's harder to maintain. Add memory size and symbol table fields. Change the names of everything. Hacks to save a few bytes: asm.S, boot.c, boot2.S: Replace `ouraddr' by `(BOOTSEG << 4)'. boot.c: Don't statically initialize `loadflags' to 0. Disable the "REDUNDANT" code that skips the BIOS variables. Eliminate `total'. Combine some more printfs. boot.h, disk.c, io.c, table.c: Move all statically initialzed data to table.c. io.c: Don't put the A20 gate bits in a variable.
1995-01-25 21:40:47 +00:00
if (du->dk_unit == bootinfo.bi_n_bios_used) {
Improve the case for ST506 disks a bit: Pick up Bios-geometry. That was the good news. The bad news is that bad144 is a proper mess, and I don't have time to fix it now, so you will probably not be able to use it anyway. Sorry guys, go out and buy a 100Mb IDE drive and a paddleboard :-( If somebody wants to pick up on this: bad144 needs to learn how to stay inside our slice of the disk. That's the trick.
1994-11-18 11:27:41 +00:00
du->dk_dd.d_secsize = DEV_BSIZE;
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
du->dk_dd.d_nsectors =
Load the kernel symbol table in the boot loader and not at compile time. (Boot with the -D flag if you want symbols.) Make it easier to extend `struct bootinfo' without losing either forwards or backwards compatibility. ddb_aout.c: Get the symbol table from wherever the loader put it. Nuke db_symtab[SYMTAB_SPACE]. boot.c: Enable loading of symbols. Align them on a page boundary. Add printfs about the symbol table sizes. Pass the memory sizes to the kernel. Fix initialization of `unit' (it got moved out of the loop). Fix adding the bss size (it got moved inside an ifdef). Initialize serial port when RB_SERIAL is toggled on. Fix comments. Clean up formatting of recently added code. io.c: Clean up formatting of recently added code. netboot/main.c, machdep.c, wd.c: Change names of bootinfo fields. LINT: Nuke SYMTAB_SPACE. Fix comment about DODUMP. Makefile.i386: Nuke use of dbsym. Exclude gcc symbols from kernel unless compiling with -g. Remove unused macro. Fix comments and formatting. genassym.c: Generate defines for some new bootinfo fields. Change names of old ones. locore.s: Copy only the valid part of the `struct bootinfo' passed by the loader. Reserve space for symbol table, if any. machdep.c: Check the memory sizes passed by the loader, if any. Don't use them yet. bootinfo.h: Add a size field so that we can resolve some mismatches between the loader bootinfo and the kernel boot info. The version number is not so good for this because of historical botches and because it's harder to maintain. Add memory size and symbol table fields. Change the names of everything. Hacks to save a few bytes: asm.S, boot.c, boot2.S: Replace `ouraddr' by `(BOOTSEG << 4)'. boot.c: Don't statically initialize `loadflags' to 0. Disable the "REDUNDANT" code that skips the BIOS variables. Eliminate `total'. Combine some more printfs. boot.h, disk.c, io.c, table.c: Move all statically initialzed data to table.c. io.c: Don't put the A20 gate bits in a variable.
1995-01-25 21:40:47 +00:00
bootinfo.bi_bios_geom[du->dk_unit] & 0xff;
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
du->dk_dd.d_ntracks =
Load the kernel symbol table in the boot loader and not at compile time. (Boot with the -D flag if you want symbols.) Make it easier to extend `struct bootinfo' without losing either forwards or backwards compatibility. ddb_aout.c: Get the symbol table from wherever the loader put it. Nuke db_symtab[SYMTAB_SPACE]. boot.c: Enable loading of symbols. Align them on a page boundary. Add printfs about the symbol table sizes. Pass the memory sizes to the kernel. Fix initialization of `unit' (it got moved out of the loop). Fix adding the bss size (it got moved inside an ifdef). Initialize serial port when RB_SERIAL is toggled on. Fix comments. Clean up formatting of recently added code. io.c: Clean up formatting of recently added code. netboot/main.c, machdep.c, wd.c: Change names of bootinfo fields. LINT: Nuke SYMTAB_SPACE. Fix comment about DODUMP. Makefile.i386: Nuke use of dbsym. Exclude gcc symbols from kernel unless compiling with -g. Remove unused macro. Fix comments and formatting. genassym.c: Generate defines for some new bootinfo fields. Change names of old ones. locore.s: Copy only the valid part of the `struct bootinfo' passed by the loader. Reserve space for symbol table, if any. machdep.c: Check the memory sizes passed by the loader, if any. Don't use them yet. bootinfo.h: Add a size field so that we can resolve some mismatches between the loader bootinfo and the kernel boot info. The version number is not so good for this because of historical botches and because it's harder to maintain. Add memory size and symbol table fields. Change the names of everything. Hacks to save a few bytes: asm.S, boot.c, boot2.S: Replace `ouraddr' by `(BOOTSEG << 4)'. boot.c: Don't statically initialize `loadflags' to 0. Disable the "REDUNDANT" code that skips the BIOS variables. Eliminate `total'. Combine some more printfs. boot.h, disk.c, io.c, table.c: Move all statically initialzed data to table.c. io.c: Don't put the A20 gate bits in a variable.
1995-01-25 21:40:47 +00:00
((bootinfo.bi_bios_geom[du->dk_unit] >> 8) & 0xff)
+ 1;
Improve the case for ST506 disks a bit: Pick up Bios-geometry. That was the good news. The bad news is that bad144 is a proper mess, and I don't have time to fix it now, so you will probably not be able to use it anyway. Sorry guys, go out and buy a 100Mb IDE drive and a paddleboard :-( If somebody wants to pick up on this: bad144 needs to learn how to stay inside our slice of the disk. That's the trick.
1994-11-18 11:27:41 +00:00
/* XXX Why 2 ? */
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
du->dk_dd.d_ncylinders =
Load the kernel symbol table in the boot loader and not at compile time. (Boot with the -D flag if you want symbols.) Make it easier to extend `struct bootinfo' without losing either forwards or backwards compatibility. ddb_aout.c: Get the symbol table from wherever the loader put it. Nuke db_symtab[SYMTAB_SPACE]. boot.c: Enable loading of symbols. Align them on a page boundary. Add printfs about the symbol table sizes. Pass the memory sizes to the kernel. Fix initialization of `unit' (it got moved out of the loop). Fix adding the bss size (it got moved inside an ifdef). Initialize serial port when RB_SERIAL is toggled on. Fix comments. Clean up formatting of recently added code. io.c: Clean up formatting of recently added code. netboot/main.c, machdep.c, wd.c: Change names of bootinfo fields. LINT: Nuke SYMTAB_SPACE. Fix comment about DODUMP. Makefile.i386: Nuke use of dbsym. Exclude gcc symbols from kernel unless compiling with -g. Remove unused macro. Fix comments and formatting. genassym.c: Generate defines for some new bootinfo fields. Change names of old ones. locore.s: Copy only the valid part of the `struct bootinfo' passed by the loader. Reserve space for symbol table, if any. machdep.c: Check the memory sizes passed by the loader, if any. Don't use them yet. bootinfo.h: Add a size field so that we can resolve some mismatches between the loader bootinfo and the kernel boot info. The version number is not so good for this because of historical botches and because it's harder to maintain. Add memory size and symbol table fields. Change the names of everything. Hacks to save a few bytes: asm.S, boot.c, boot2.S: Replace `ouraddr' by `(BOOTSEG << 4)'. boot.c: Don't statically initialize `loadflags' to 0. Disable the "REDUNDANT" code that skips the BIOS variables. Eliminate `total'. Combine some more printfs. boot.h, disk.c, io.c, table.c: Move all statically initialzed data to table.c. io.c: Don't put the A20 gate bits in a variable.
1995-01-25 21:40:47 +00:00
(bootinfo.bi_bios_geom[du->dk_unit] >> 16) + 2;
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
du->dk_dd.d_secpercyl =
du->dk_dd.d_ntracks * du->dk_dd.d_nsectors;
du->dk_dd.d_secperunit =
du->dk_dd.d_secpercyl * du->dk_dd.d_ncylinders;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#if 0
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
du->dk_dd.d_partitions[WDRAW].p_size =
Improve the case for ST506 disks a bit: Pick up Bios-geometry. That was the good news. The bad news is that bad144 is a proper mess, and I don't have time to fix it now, so you will probably not be able to use it anyway. Sorry guys, go out and buy a 100Mb IDE drive and a paddleboard :-( If somebody wants to pick up on this: bad144 needs to learn how to stay inside our slice of the disk. That's the trick.
1994-11-18 11:27:41 +00:00
du->dk_dd.d_secperunit;
du->dk_dd.d_type = DTYPE_ST506;
du->dk_dd.d_subtype |= DSTYPE_GEOMETRY;
strncpy(du->dk_dd.d_typename, "Bios geometry",
sizeof du->dk_dd.d_typename);
strncpy(du->dk_params.wdp_model, "ST506",
sizeof du->dk_params.wdp_model);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#endif
Load the kernel symbol table in the boot loader and not at compile time. (Boot with the -D flag if you want symbols.) Make it easier to extend `struct bootinfo' without losing either forwards or backwards compatibility. ddb_aout.c: Get the symbol table from wherever the loader put it. Nuke db_symtab[SYMTAB_SPACE]. boot.c: Enable loading of symbols. Align them on a page boundary. Add printfs about the symbol table sizes. Pass the memory sizes to the kernel. Fix initialization of `unit' (it got moved out of the loop). Fix adding the bss size (it got moved inside an ifdef). Initialize serial port when RB_SERIAL is toggled on. Fix comments. Clean up formatting of recently added code. io.c: Clean up formatting of recently added code. netboot/main.c, machdep.c, wd.c: Change names of bootinfo fields. LINT: Nuke SYMTAB_SPACE. Fix comment about DODUMP. Makefile.i386: Nuke use of dbsym. Exclude gcc symbols from kernel unless compiling with -g. Remove unused macro. Fix comments and formatting. genassym.c: Generate defines for some new bootinfo fields. Change names of old ones. locore.s: Copy only the valid part of the `struct bootinfo' passed by the loader. Reserve space for symbol table, if any. machdep.c: Check the memory sizes passed by the loader, if any. Don't use them yet. bootinfo.h: Add a size field so that we can resolve some mismatches between the loader bootinfo and the kernel boot info. The version number is not so good for this because of historical botches and because it's harder to maintain. Add memory size and symbol table fields. Change the names of everything. Hacks to save a few bytes: asm.S, boot.c, boot2.S: Replace `ouraddr' by `(BOOTSEG << 4)'. boot.c: Don't statically initialize `loadflags' to 0. Disable the "REDUNDANT" code that skips the BIOS variables. Eliminate `total'. Combine some more printfs. boot.h, disk.c, io.c, table.c: Move all statically initialzed data to table.c. io.c: Don't put the A20 gate bits in a variable.
1995-01-25 21:40:47 +00:00
bootinfo.bi_n_bios_used ++;
Improve the case for ST506 disks a bit: Pick up Bios-geometry. That was the good news. The bad news is that bad144 is a proper mess, and I don't have time to fix it now, so you will probably not be able to use it anyway. Sorry guys, go out and buy a 100Mb IDE drive and a paddleboard :-( If somebody wants to pick up on this: bad144 needs to learn how to stay inside our slice of the disk. That's the trick.
1994-11-18 11:27:41 +00:00
return 0;
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/*
* Fake minimal drive geometry for reading the MBR.
* readdisklabel() may enlarge it to read the label and the
* bad sector table.
*/
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du->dk_dd.d_secsize = DEV_BSIZE;
du->dk_dd.d_nsectors = 17;
du->dk_dd.d_ntracks = 1;
du->dk_dd.d_ncylinders = 1;
du->dk_dd.d_secpercyl = 17;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_dd.d_secperunit = 17;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#if 0
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/*
* Fake maximal drive size for writing the label.
*/
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
du->dk_dd.d_partitions[RAW_PART].p_size = 64 * 16 * 1024;
Fixed for DQFLQUIET flag from: Terry Lee <terry@uivlsisd.csl.uiuc.edu> I submitted the part of patch00155 that fixes wd.c not reporting any bad blocks. Unfortunately, the patch created a new problem that makes new installations using that kernel impossible. I have included a patch below that should fix both problems. The fix comes from Tom Ivar Helbekkmo's wd driver (the Norway wd driver). I've also eliminated the use of the DKFL_QUIET flag completely, since it is not used anymore. Fixed wdgetctrl routines for so that it returns a meaniful string for non IDE disk drives. It also nolonger returns success when a drive is not really there.
1993-08-08 07:03:21 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/*
* Fake some more of the label for printing by disklabel(1)
* in case there is no real label.
*/
du->dk_dd.d_type = DTYPE_ST506;
du->dk_dd.d_subtype |= DSTYPE_GEOMETRY;
strncpy(du->dk_dd.d_typename, "Fake geometry",
Fixed for DQFLQUIET flag from: Terry Lee <terry@uivlsisd.csl.uiuc.edu> I submitted the part of patch00155 that fixes wd.c not reporting any bad blocks. Unfortunately, the patch created a new problem that makes new installations using that kernel impossible. I have included a patch below that should fix both problems. The fix comes from Tom Ivar Helbekkmo's wd driver (the Norway wd driver). I've also eliminated the use of the DKFL_QUIET flag completely, since it is not used anymore. Fixed wdgetctrl routines for so that it returns a meaniful string for non IDE disk drives. It also nolonger returns success when a drive is not really there.
1993-08-08 07:03:21 +00:00
sizeof du->dk_dd.d_typename);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/* Fake the model name for printing by wdattach(). */
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
strncpy(du->dk_params.wdp_model, "unknown",
Fixed for DQFLQUIET flag from: Terry Lee <terry@uivlsisd.csl.uiuc.edu> I submitted the part of patch00155 that fixes wd.c not reporting any bad blocks. Unfortunately, the patch created a new problem that makes new installations using that kernel impossible. I have included a patch below that should fix both problems. The fix comes from Tom Ivar Helbekkmo's wd driver (the Norway wd driver). I've also eliminated the use of the DKFL_QUIET flag completely, since it is not used anymore. Fixed wdgetctrl routines for so that it returns a meaniful string for non IDE disk drives. It also nolonger returns success when a drive is not really there.
1993-08-08 07:03:21 +00:00
sizeof du->dk_params.wdp_model);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
/* obtain parameters */
wp = &du->dk_params;
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
if (du->dk_flags & DKFL_32BIT)
insl(du->dk_port + wd_data, tb, sizeof(tb) / sizeof(long));
else
insw(du->dk_port + wd_data, tb, sizeof(tb) / sizeof(short));
/* try 32-bit data path (VLB IDE controller) */
Implemented the multi-sector I/O enable and the 32 bit I/O probe through the use of the config file flags as opposed to the option "NSECS_MULTI". "NSECS_MULTI" has been removed from the driver. The new capability allows boot-time modification of the config.
1995-04-24 04:32:31 +00:00
if (flags & WDOPT_32BIT) {
if (! (du->dk_flags & DKFL_32BIT)) {
bcopy(tb, tb2, sizeof(struct wdparams));
du->dk_flags |= DKFL_32BIT;
goto again;
}
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
Implemented the multi-sector I/O enable and the 32 bit I/O probe through the use of the config file flags as opposed to the option "NSECS_MULTI". "NSECS_MULTI" has been removed from the driver. The new capability allows boot-time modification of the config.
1995-04-24 04:32:31 +00:00
/* check that we really have 32-bit controller */
if (bcmp (tb, tb2, sizeof(struct wdparams)) != 0) {
Fixes and improvements from John Dyson: Fixed the I/O statistics Allow WD1007 type controllers to work Support MULTI-BLOCK I/O Correct delay to use port 0x84, reading the status register might not be a long enough delay. Changed probe message to match SCSI type devices.
1995-03-22 05:23:01 +00:00
failed:
Implemented the multi-sector I/O enable and the 32 bit I/O probe through the use of the config file flags as opposed to the option "NSECS_MULTI". "NSECS_MULTI" has been removed from the driver. The new capability allows boot-time modification of the config.
1995-04-24 04:32:31 +00:00
/* test failed, use 16-bit i/o mode */
bcopy(tb2, tb, sizeof(struct wdparams));
du->dk_flags &= ~DKFL_32BIT;
}
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
bcopy(tb, wp, sizeof(struct wdparams));
/* shuffle string byte order */
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
for (i = 0; i < sizeof(wp->wdp_model); i += 2) {
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
u_short *p;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
p = (u_short *) (wp->wdp_model + i);
*p = ntohs(*p);
}
More clean up of the drive Id string, trancate trailing spaces.
1994-02-22 22:13:37 +00:00
/*
* Clean up the wdp_model by converting nulls to spaces, and
* then removing the trailing spaces.
*/
for (i=0; i < sizeof(wp->wdp_model); i++) {
The model string can be 40 bytes long with no terminating null, so convert all nulls to spaces and then print it with a %40s.
1994-02-22 18:51:27 +00:00
if (wp->wdp_model[i] == '\0') {
wp->wdp_model[i] = ' ';
}
}
More clean up of the drive Id string, trancate trailing spaces.
1994-02-22 22:13:37 +00:00
for (i=sizeof(wp->wdp_model)-1; i>=0 && wp->wdp_model[i]==' '; i--) {
wp->wdp_model[i] = '\0';
}
The model string can be 40 bytes long with no terminating null, so convert all nulls to spaces and then print it with a %40s.
1994-02-22 18:51:27 +00:00
Another attempt at making multi-sector mode work.
1996-08-23 02:52:44 +00:00
/*
* find out the drives maximum multi-block transfer capability
*/
du->dk_multi = wp->wdp_nsecperint & 0xff;
wdsetmulti(du);
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
/*
* check drive's DMA capability
*/
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
if (wddma[du->dk_interface].wdd_candma) {
du->dk_dmacookie = wddma[du->dk_interface].wdd_candma(du->dk_port, du->dk_ctrlr);
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
/* does user want this? */
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
if ((du->cfg_flags & WDOPT_DMA) &&
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
/* have we got a DMA controller? */
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
du->dk_dmacookie &&
/* can said drive do DMA? */
wddma[du->dk_interface].wdd_dmainit(du->dk_dmacookie, wp, wdsetmode, du)) {
du->dk_flags |= DKFL_USEDMA;
}
} else {
du->dk_dmacookie = NULL;
}
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#ifdef WDDEBUG
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
printf(
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
"\nwd(%d,%d): wdgetctlr: gc %x cyl %d trk %d sec %d type %d sz %d model %s\n",
Fixed calculation of the number of cylinders. wdp_cylinders (was wdp_fixedcyl) gives it directly. wdp_removedcyl is "reserved" except in ancient ATA-1 drafts and shouldn't be added. This fixes PR 1288. Changed some fields and comments in struct wdparams to match a less- ancient ATA draft. Fixed bit number for `rdy' in status string.
1996-06-08 10:03:38 +00:00
du->dk_ctrlr, du->dk_unit, wp->wdp_config, wp->wdp_cylinders,
wp->wdp_heads, wp->wdp_sectors, wp->wdp_buffertype,
wp->wdp_buffersize, wp->wdp_model);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* update disklabel given drive information */
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du->dk_dd.d_secsize = DEV_BSIZE;
Fixed calculation of the number of cylinders. wdp_cylinders (was wdp_fixedcyl) gives it directly. wdp_removedcyl is "reserved" except in ancient ATA-1 drafts and shouldn't be added. This fixes PR 1288. Changed some fields and comments in struct wdparams to match a less- ancient ATA draft. Fixed bit number for `rdy' in status string.
1996-06-08 10:03:38 +00:00
du->dk_dd.d_ncylinders = wp->wdp_cylinders; /* +- 1 */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
du->dk_dd.d_ntracks = wp->wdp_heads;
du->dk_dd.d_nsectors = wp->wdp_sectors;
du->dk_dd.d_secpercyl = du->dk_dd.d_ntracks * du->dk_dd.d_nsectors;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
du->dk_dd.d_secperunit = du->dk_dd.d_secpercyl * du->dk_dd.d_ncylinders;
Add yet another kludge to this driver. Man page update to follow.
1996-07-21 09:28:50 +00:00
if (WDOPT_FORCEHD(du->cfg_flags)) {
du->dk_dd.d_ntracks = WDOPT_FORCEHD(du->cfg_flags);
du->dk_dd.d_secpercyl =
du->dk_dd.d_ntracks * du->dk_dd.d_nsectors;
du->dk_dd.d_ncylinders =
du->dk_dd.d_secperunit / du->dk_dd.d_secpercyl;
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#if 0
du->dk_dd.d_partitions[RAW_PART].p_size = du->dk_dd.d_secperunit;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* dubious ... */
bcopy("ESDI/IDE", du->dk_dd.d_typename, 9);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
bcopy(wp->wdp_model + 20, du->dk_dd.d_packname, 14 - 1);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
/* better ... */
du->dk_dd.d_type = DTYPE_ESDI;
du->dk_dd.d_subtype |= DSTYPE_GEOMETRY;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (0);
}
int
Declare all the args of wdclose() and wdioctl(). Cosmetic.
1995-02-26 01:15:30 +00:00
wdclose(dev_t dev, int flags, int fmt, struct proc *p)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
{
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
dsclose(dev, fmt, wddrives[dkunit(dev)]->dk_slices);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
int
Declare all the args of wdclose() and wdioctl(). Cosmetic.
1995-02-26 01:15:30 +00:00
wdioctl(dev_t dev, int cmd, caddr_t addr, int flags, struct proc *p)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
{
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
int lunit = dkunit(dev);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
register struct disk *du;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
int error;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#ifdef notyet
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
struct uio auio;
struct iovec aiov;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
struct format_op *fop;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du = wddrives[lunit];
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
wdsleep(du->dk_ctrlr, "wdioct");
Change dsioctl() interface to allow DIOCSYNCSLICEINFO to update the caller's slice pointer and to print the device name in error messages.
1995-04-30 15:14:34 +00:00
error = dsioctl("wd", dev, cmd, addr, flags, &du->dk_slices,
wdstrategy1, (ds_setgeom_t *)NULL);
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (error != -1)
return (error);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
switch (cmd) {
Changes to support the "new" bad144. Adds a "BADSCAN" ioctl and support. Reviewed by: Bruce Evans Submitted by: Poul-Henning Kamp
1995-05-16 07:52:17 +00:00
case DIOCSBADSCAN:
if (*(int *)addr)
du->dk_flags |= DKFL_BADSCAN;
else
du->dk_flags &= ~DKFL_BADSCAN;
return (0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#ifdef notyet
case DIOCWFORMAT:
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (!(flag & FWRITE))
return (EBADF);
fop = (struct format_op *)addr;
aiov.iov_base = fop->df_buf;
aiov.iov_len = fop->df_count;
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_resid = fop->df_count;
auio.uio_segflg = 0;
auio.uio_offset = fop->df_startblk * du->dk_dd.d_secsize;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#error /* XXX the 386BSD interface is different */
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
error = physio(wdformat, &rwdbuf[lunit], 0, dev, B_WRITE,
minphys, &auio);
fop->df_count -= auio.uio_resid;
fop->df_reg[0] = du->dk_status;
fop->df_reg[1] = du->dk_error;
return (error);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
#endif
default:
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
return (ENOTTY);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#ifdef B_FORMAT
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
int
wdformat(struct buf *bp)
{
bp->b_flags |= B_FORMAT;
Julian Elischer's disklabel fixes.
1994-10-27 20:45:13 +00:00
wdstrategy(bp);
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
/*
Julian Elischer's disklabel fixes.
1994-10-27 20:45:13 +00:00
* phk put this here, better that return(wdstrategy(bp));
* XXX
*/
return -1;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
#endif
int
wdsize(dev_t dev)
{
struct disk *du;
Fix numerous bugs in wdsize(). It closed the wrong device (usually the B partition instead of the C partition). It closed a device without having opened it. It didn't open a device often enough. This caused swap partitions on slices other than the first slice looked at to be unavailable for swapping. It didn't check the device number sufficiently.
1995-05-08 16:48:23 +00:00
int lunit;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Fix numerous bugs in wdsize(). It closed the wrong device (usually the B partition instead of the C partition). It closed a device without having opened it. It didn't open a device often enough. This caused swap partitions on slices other than the first slice looked at to be unavailable for swapping. It didn't check the device number sufficiently.
1995-05-08 16:48:23 +00:00
lunit = dkunit(dev);
if (lunit >= NWD || dktype(dev) != 0)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (-1);
Fix numerous bugs in wdsize(). It closed the wrong device (usually the B partition instead of the C partition). It closed a device without having opened it. It didn't open a device often enough. This caused swap partitions on slices other than the first slice looked at to be unavailable for swapping. It didn't check the device number sufficiently.
1995-05-08 16:48:23 +00:00
du = wddrives[lunit];
if (du == NULL)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
return (-1);
Fix numerous bugs in wdsize(). It closed the wrong device (usually the B partition instead of the C partition). It closed a device without having opened it. It didn't open a device often enough. This caused swap partitions on slices other than the first slice looked at to be unavailable for swapping. It didn't check the device number sufficiently.
1995-05-08 16:48:23 +00:00
return (dssize(dev, &du->dk_slices, wdopen, wdclose));
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/*
* Dump core after a system crash.
*/
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
int
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wddump(dev_t dev)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
{
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
register struct disk *du;
struct disklabel *lp;
long num; /* number of sectors to write */
int lunit, part;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
long blkoff, blknum;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
long blkchk, blkcnt, blknext;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
long cylin, head, sector;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
long secpertrk, secpercyl, nblocks;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
u_long ds_offset;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
char *addr;
static int wddoingadump = 0;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Toss any characters present prior to dump. */
Changed cncheckc() interface so that it is 8-bit clean - return -1 instead of 0 if there is no input. syscons.c: Added missing spl locking in sccncheckc(). Return the same value as sccngetc() would. It is wrong for sccngetc() to return non-ASCII, but stripping the non-ASCII bits doesn't help.
1996-09-14 04:27:46 +00:00
while (cncheckc() != -1)
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Check for acceptable device. */
/* XXX should reset to maybe allow du->dk_state < OPEN. */
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
lunit = dkunit(dev); /* eventually support floppies? */
part = dkpart(dev);
if (lunit >= NWD || (du = wddrives[lunit]) == NULL
|| du->dk_state < OPEN
|| (lp = dsgetlabel(dev, du->dk_slices)) == NULL)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (ENXIO);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Size of memory to dump, in disk sectors. */
First pass at cleaning up macros relating to pages, clusters and all that.
1996-05-02 10:43:17 +00:00
num = (u_long)Maxmem * PAGE_SIZE / du->dk_dd.d_secsize;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
secpertrk = du->dk_dd.d_nsectors;
secpercyl = du->dk_dd.d_secpercyl;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
nblocks = lp->d_partitions[part].p_size;
blkoff = lp->d_partitions[part].p_offset;
/* XXX */
ds_offset = du->dk_slices->dss_slices[dkslice(dev)].ds_offset;
blkoff += ds_offset;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#if 0
pg("part %x, nblocks %d, dumplo %d num %d\n",
part, nblocks, dumplo, num);
#endif
/* Check transfer bounds against partition size. */
if (dumplo < 0 || dumplo + num > nblocks)
return (EINVAL);
/* Check if we are being called recursively. */
if (wddoingadump)
return (EFAULT);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#if 0
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Mark controller active for if we panic during the dump. */
wdtab[du->dk_ctrlr].b_active = 1;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#endif
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wddoingadump = 1;
/* Recalibrate the drive. */
DELAY(5); /* ATA spec XXX NOT */
if (wdcommand(du, 0, 0, 0, 0, WDCC_RESTORE | WD_STEP) != 0
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
|| wdwait(du, WDCS_READY | WDCS_SEEKCMPLT, TIMEOUT) != 0
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
|| wdsetctlr(du) != 0) {
wderror((struct buf *)NULL, du, "wddump: recalibrate failed");
return (EIO);
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_flags |= DKFL_SINGLE;
addr = (char *) 0;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
blknum = dumplo + blkoff;
while (num > 0) {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
blkcnt = num;
if (blkcnt > MAXTRANSFER)
blkcnt = MAXTRANSFER;
/* Keep transfer within current cylinder. */
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
if ((blknum + blkcnt - 1) / secpercyl != blknum / secpercyl)
blkcnt = secpercyl - (blknum % secpercyl);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
blknext = blknum + blkcnt;
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
Did a s/[ \t]*$//
1995-04-24 05:09:53 +00:00
/*
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
* See if one of the sectors is in the bad sector list
* (if we have one). If the first sector is bad, then
* reduce the transfer to this one bad sector; if another
* sector is bad, then reduce reduce the transfer to
* avoid any bad sectors.
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
*/
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
if (du->dk_flags & DKFL_SINGLE
&& dsgetbad(dev, du->dk_slices) != NULL) {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
for (blkchk = blknum; blkchk < blknum + blkcnt; blkchk++) {
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
daddr_t blknew;
blknew = transbad144(dsgetbad(dev, du->dk_slices),
blkchk - ds_offset) + ds_offset;
if (blknew != blkchk) {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Found bad block. */
blkcnt = blkchk - blknum;
if (blkcnt > 0) {
blknext = blknum + blkcnt;
goto out;
}
blkcnt = 1;
blknext = blknum + blkcnt;
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
#if 1 || defined(WDDEBUG)
printf("bad block %lu -> %lu\n",
blknum, blknew);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
break;
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
}
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
out:
/* Compute disk address. */
cylin = blknum / secpercyl;
head = (blknum % secpercyl) / secpertrk;
sector = blknum % secpertrk;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#if 0
/* Let's just talk about this first... */
pg("cylin l%d head %ld sector %ld addr 0x%x count %ld",
cylin, head, sector, addr, blkcnt);
#endif
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Do the write. */
if (wdcommand(du, cylin, head, sector, blkcnt, WDCC_WRITE)
!= 0) {
wderror((struct buf *)NULL, du,
"wddump: timeout waiting to to give command");
return (EIO);
}
while (blkcnt != 0) {
Revert the logic behind my last change, and use a function called `is_physical_memory()' now for the decision whether to dump some region of memory or not. Suggested by: davidg
1997-09-13 16:12:15 +00:00
if (is_physical_memory((vm_offset_t)addr))
Do not ever try to coredump adapter memory regions. PR: 4486 Submitted by: tegge@idi.ntnu.no (Tor Egge) Implement a function is_adapter_memory() in order to determine what should nto be dumped at all. Currently, only populated with the ``ISA memory hole''. Adapter regions of other busses should be added.
1997-09-10 12:31:40 +00:00
pmap_enter(kernel_pmap, (vm_offset_t)CADDR1,
Revert the logic behind my last change, and use a function called `is_physical_memory()' now for the decision whether to dump some region of memory or not. Suggested by: davidg
1997-09-13 16:12:15 +00:00
trunc_page(addr), VM_PROT_READ, TRUE);
Do not ever try to coredump adapter memory regions. PR: 4486 Submitted by: tegge@idi.ntnu.no (Tor Egge) Implement a function is_adapter_memory() in order to determine what should nto be dumped at all. Currently, only populated with the ``ISA memory hole''. Adapter regions of other busses should be added.
1997-09-10 12:31:40 +00:00
else
pmap_enter(kernel_pmap, (vm_offset_t)CADDR1,
Revert the logic behind my last change, and use a function called `is_physical_memory()' now for the decision whether to dump some region of memory or not. Suggested by: davidg
1997-09-13 16:12:15 +00:00
trunc_page(0), VM_PROT_READ, TRUE);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Ready to send data? */
DELAY(5); /* ATA spec */
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (wdwait(du, WDCS_READY | WDCS_SEEKCMPLT | WDCS_DRQ, TIMEOUT)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
< 0) {
wderror((struct buf *)NULL, du,
"wddump: timeout waiting for DRQ");
return (EIO);
}
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
if (du->dk_flags & DKFL_32BIT)
outsl(du->dk_port + wd_data,
First pass at cleaning up macros relating to pages, clusters and all that.
1996-05-02 10:43:17 +00:00
CADDR1 + ((int)addr & PAGE_MASK),
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
DEV_BSIZE / sizeof(long));
else
outsw(du->dk_port + wd_data,
First pass at cleaning up macros relating to pages, clusters and all that.
1996-05-02 10:43:17 +00:00
CADDR1 + ((int)addr & PAGE_MASK),
Support 32bit access to IDE disks, if it seems to work for the drive. You will normally have to have a VLB or other 32bit IDE "controller" for this to work. Depending on your setup, this may gain you 20-100 % speed from your disk. Reviewed by: phk Submitted by: vak@cronyx.ru
1995-02-04 19:39:36 +00:00
DEV_BSIZE / sizeof(short));
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
addr += DEV_BSIZE;
Teach both disk drivers how to cope with a hardware watchdog while dumping core.. I'm tired of getting 1/2 of a core-dump conditional on -DHW_WDOG for now this will migrate to 2.2 as that's where I need it.
1997-08-09 01:44:25 +00:00
/*
* If we are dumping core, it may take a while.
* So reassure the user and hold off any watchdogs.
*/
if ((unsigned)addr % (1024 * 1024) == 0) {
#ifdef HW_WDOG
if (wdog_tickler)
(*wdog_tickler)();
#endif /* HW_WDOG */
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
printf("%ld ", num / (1024 * 1024 / DEV_BSIZE));
Teach both disk drivers how to cope with a hardware watchdog while dumping core.. I'm tired of getting 1/2 of a core-dump conditional on -DHW_WDOG for now this will migrate to 2.2 as that's where I need it.
1997-08-09 01:44:25 +00:00
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
num--;
blkcnt--;
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Wait for completion. */
DELAY(5); /* ATA spec XXX NOT */
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (wdwait(du, WDCS_READY | WDCS_SEEKCMPLT, TIMEOUT) < 0) {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wderror((struct buf *)NULL, du,
"wddump: timeout waiting for status");
return (EIO);
}
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Check final status. */
if (du->dk_status
& (WDCS_READY | WDCS_SEEKCMPLT | WDCS_DRQ | WDCS_ERR)
!= (WDCS_READY | WDCS_SEEKCMPLT)) {
wderror((struct buf *)NULL, du,
"wddump: extra DRQ, or error");
return (EIO);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
/* Update block count. */
blknum = blknext;
/* Operator aborting dump? */
Changed cncheckc() interface so that it is 8-bit clean - return -1 instead of 0 if there is no input. syscons.c: Added missing spl locking in sccncheckc(). Return the same value as sccngetc() would. It is wrong for sccngetc() to return non-ASCII, but stripping the non-ASCII bits doesn't help.
1996-09-14 04:27:46 +00:00
if (cncheckc() != -1)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (EINTR);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
return (0);
Initial import, 0.1 + pk 0.2.4-B1
1993-06-12 14:58:17 +00:00
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
static void
wderror(struct buf *bp, struct disk *du, char *mesg)
{
if (bp == NULL)
Clean up dmesg output so that it matches the scsi stuff. Small hack, if heads > 16, output a warning and then set the head count to 16. This stops the infinite loop on this error and allows people to later fix the DOS parameter table later with fdisk.
1994-02-21 12:32:33 +00:00
printf("wd%d: %s:\n", du->dk_lunit, mesg);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
else
I commit this for Bruce, who has serious connectivity problmes pt. This should NOT go into 2.0.5 /phk Support disk slices. This involves mainly replacing inline code with function calls. Support for ST506 drives is temporarily broken since the `setgeom' arg to dsopen() is not implemented completely enough to use. The `setgeom' arg will go away and ST506 drives will be supported in another way. A large amount of dead code is left in wdopen() as a reminder of the problems here. Close the device in wdsize(). Open tracking was broken on all drives with a swap device. Remove support for soft write protection. There are no ioctls to set it. It was used to disable writing to unlabelled disks, but we want to support writing to foreign partitions on unlabeled disks. Use generic dkbad routines to do about 2/3 of the work for supporting bad144. Improve disk statistics: estimate 4MB/sec instead of 8MB/sec for the transfer rate (ISA max is 4MB/sec, old IDE max is 3.3MB/sec); fix dk_xfer[] (it counted sectors, not transfers); keep the estimate dk_seek[] = dk_xfer[] (was sectors, is now transfers); only count words actually transferred (the count is still too high after a failed write and after retries). Remove wdxfer[]. Fix indentation in wdattach(). Fix resulting botched printing of the disk size for ST506 drives. Print the disk geometry less cryptically.
1995-04-14 22:31:58 +00:00
diskerr(bp, "wd", mesg, LOG_PRINTF, du->dk_skip,
dsgetlabel(bp->b_dev, du->dk_slices));
Clean up dmesg output so that it matches the scsi stuff. Small hack, if heads > 16, output a warning and then set the head count to 16. This stops the infinite loop on this error and allows people to later fix the DOS parameter table later with fdisk.
1994-02-21 12:32:33 +00:00
printf("wd%d: status %b error %b\n", du->dk_lunit,
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du->dk_status, WDCS_BITS, du->dk_error, WDERR_BITS);
}
/*
* Discard any interrupts that were latched by the interrupt system while
* we were doing polled i/o.
*/
static void
wdflushirq(struct disk *du, int old_ipl)
{
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
wdtab[du->dk_ctrlr_cmd640].b_active = 2;
splx(old_ipl);
(void)splbio();
wdtab[du->dk_ctrlr_cmd640].b_active = 0;
#else
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wdtab[du->dk_ctrlr].b_active = 2;
splx(old_ipl);
(void)splbio();
wdtab[du->dk_ctrlr].b_active = 0;
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
}
/*
* Reset the controller.
*/
static int
wdreset(struct disk *du)
{
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
int err = 0;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
if ((du->dk_flags & (DKFL_DMA|DKFL_USEDMA)) && du->dk_dmacookie)
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
wddma[du->dk_interface].wdd_dmadone(du->dk_dmacookie);
Add support for busmaster DMA on some PCI IDE chipsets. I changed a few bits here and there, mainly renaming wd82371.c to ide_pci.c now that it's supposed to handle different chipsets. It runs on my P6 natoma board with two Maxtor drives, and also on a Fujitsu machine I have at work with an Opti chipset and a Quantum drive. Submitted by:cgull@smoke.marlboro.vt.us <John Hood> Original readme: *** WARNING *** This code has so far been tested on exactly one motherboard with two identical drives known for their good DMA support. This code, in the right circumstances, could corrupt data subtly, silently, and invisibly, in much the same way that older PCI IDE controllers do. It's ALPHA-quality code; there's one or two major gaps in my understanding of PCI IDE still. Don't use this code on any system with data that you care about; it's only good for hack boxes. Expect that any data may be silently and randomly corrupted at any moment. It's a disk driver. It has bugs. Disk drivers with bugs munch data. It's a fact of life. I also *STRONGLY* recommend getting a copy of your chipset's manual and the ATA-2 or ATA-3 spec and making sure that timing modes on your disk drives and IDE controller are being setup correctly by the BIOS-- because the driver makes only the lamest of attempts to do this just now. *** END WARNING *** that said, i happen to think the code is working pretty well... WHAT IT DOES: this code adds support to the wd driver for bus mastering PCI IDE controllers that follow the SFF-8038 standard. (all the bus mastering PCI IDE controllers i've seen so far do follow this standard.) it should provide busmastering on nearly any current P5 or P6 chipset, specifically including any Intel chipset using one of the PIIX south bridges-- this includes the '430FX, '430VX, '430HX, '430TX, '440LX, and (i think) the Orion '450GX chipsets. specific support is also included for the VIA Apollo VP-1 chipset, as it appears in the relabeled "HXPro" incarnation seen on cheap US$70 taiwanese motherboards (that's what's in my development machine). it works out of the box on controllers that do DMA mode2; if my understanding is correct, it'll probably work on Ultra-DMA33 controllers as well. it'll probably work on busmastering IDE controllers in PCI slots, too, but this is an area i am less sure about. it cuts CPU usage considerably and improves drive performance slightly. usable numbers are difficult to come by with existing benchmark tools, but experimentation on my K5-P90 system, with VIA VP-1 chipset and Quantum Fireball 1080 drives, shows that disk i/o on raw partitions imposes perhaps 5% cpu load. cpu load during filesystem i/o drops a lot, from near 100% to anywhere between 30% and 70%. (the improvement may not be as large on an Intel chipset; from what i can tell, the VIA VP-1 may not be very efficient with PCI I/O.) disk performance improves by 5% or 10% with these drives. real, visible, end-user performance improvement on a single user machine is about nil. :) a kernel compile was sped up by a whole three seconds. it *does* feel a bit better-behaved when the system is swapping heavily, but a better disk driver is not the fix for *that* problem. THE CODE: this code is a patch to wd.c and wd82371.c, and associated header files. it should be considered alpha code; more work needs to be done. wd.c has fairly clean patches to add calls to busmaster code, as implemented in wd82371.c and potentially elsewhere (one could imagine, say, a Mac having a different DMA controller). wd82371.c has been considerably reworked: the wddma interface that it presents has been changed (expect more changes), many bugs have been fixed, a new internal interface has been added for supporting different chipsets, and the PCI probe has been considerably extended. the interface between wd82371.c and wd.c is still fairly clean, but i'm not sure it's in the right place. there's a mess of issues around ATA/ATAPI that need to be sorted out, including ATAPI support, CD-ROM support, tape support, LS-120/Zip support, SFF-8038i DMA, UltraDMA, PCI IDE controllers, bus probes, buggy controllers, controller timing setup, drive timing setup, world peace and kitchen sinks. whatever happens with all this and however it gets partitioned, it is fairly clear that wd.c needs some significant rework-- probably a complete rewrite. timing setup on disk controllers is something i've entirely punted on. on my development machine, it appears that the BIOS does at least some of the necessary timing setup. i chose to restrict operation to drives that are already configured for Mode4 PIO and Mode2 multiword DMA, since the timing is essentially the same and many if not most chipsets use the same control registers for DMA and PIO timing. does anybody *know* whether BIOSes are required to do timing setup for DMA modes on drives under their care? error recovery is probably weak. early on in development, i was getting drive errors induced by bugs in the driver; i used these to flush out the worst of the bugs in the driver's error handling, but problems may remain. i haven't got a drive with bad sectors i can watch the driver flail on. complaints about how wd82371.c has been reindented will be ignored until the FreeBSD project has a real style policy, there is a mechanism for individual authors to match it (indent flags or an emacs c-mode or whatever), and it is enforced. if i'm going to use a source style i don't like, it would help if i could figure out what it *is* (style(9) is about half of a policy), and a way to reasonably duplicate it. i ended up wasting a while trying to figure out what the right thing to do was before deciding reformatting the whole thing was the worst possible thing to do, except for all the other possibilities. i have maintained wd.c's indentation; that was not too hard, fortunately. TO INSTALL: my dev box is freebsd 2.2.2 release. fortunately, wd.c is a living fossil, and has diverged very little recently. included in this tarball is a patch file, 'otherdiffs', for all files except wd82371.c, my edited wd82371.c, a patch file, 'wd82371.c-diff-exact', against the 2.2.2 dist of 82371.c, and another patch file, 'wd82371.c-diff-whitespace', generated with diff -b (ignore whitespace). most of you not using 2.2.2 will probably have to use this last patchfile with 'patch --ignore-whitespace'. apply from the kernel source tree root. as far as i can tell, this should apply cleanly on anything from -current back to 2.2.2 and probably back to 2.2.0. you, the kernel hacker, can figure out what to do from here. if you need more specific directions, you probably should not be experimenting with this code yet. to enable DMA support, set flag 0x2000 for that drive in your config file or in userconfig, as you would the 32-bit-PIO flag. the driver will then turn on DMA support if your drive and controller pass its tests. it's a bit picky, probably. on discovering DMA mode failures or disk errors or transfers that the DMA controller can't deal with, the driver will fall back to PIO, so it is wise to setup the flags as if PIO were still important. 'controller wdc0 at isa? port "IO_WD1" bio irq 14 flags 0xa0ffa0ff vector wdintr' should work with nearly any PCI IDE controller. i would *strongly* suggest booting single-user at first, and thrashing the drive a bit while it's still mounted read-only. this should be fairly safe, even if the driver goes completely out to lunch. it might save you a reinstall. one way to tell whether the driver is really using DMA is to check the interrupt count during disk i/o with vmstat; DMA mode will add an extremely low number of interrupts, as compared to even multi-sector PIO. boot -v will give you a copious register dump of timing-related info on Intel and VIAtech chipsets, as well as PIO/DMA mode information on all hard drives. refer to your ATA and chipset documentation to interpret these. WHAT I'D LIKE FROM YOU and THINGS TO TEST: reports. success reports, failure reports, any kind of reports. :) send them to cgull+ide@smoke.marlboro.vt.us. i'd also like to see the kernel messages from various BIOSes (boot -v; dmesg), along with info on the motherboard and BIOS on that machine. i'm especially interested in reports on how this code works on the various Intel chipsets, and whether the register dump works correctly. i'm also interested in hearing about other chipsets. i'm especially interested in hearing success/failure reports for PCI IDE controllers on cards, such as CMD's or Promise's new busmastering IDE controllers. UltraDMA-33 reports. interoperation with ATAPI peripherals-- FreeBSD doesn't work with my old Hitachi IDE CDROM, so i can't tell if I've broken anything. :) i'd especially like to hear how the drive copes in DMA operation on drives with bad sectors. i haven't been able to find any such yet. success/failure reports on older IDE drives with early support for DMA modes-- those introduced between 1.5 and 3 years ago, typically ranging from perhaps 400MB to 1.6GB. failure reports on operation with more than one drive would be appreciated. the driver was developed with two drives on one controller, the worst-case situation, and has been tested with one drive on each controller, but you never know... any reports of messages from the driver during normal operation, especially "reverting to PIO mode", or "dmaverify odd vaddr or length" (the DMA controller is strongly halfword oriented, and i'm curious to know if any FreeBSD usage actually needs misaligned transfers). performance reports. beware that bonnie's CPU usage reporting is useless for IDE drives; the best test i've found has been to run a program that runs a spin loop at an idle priority and reports how many iterations it manages, and even that sometimes produces numbers i don't believe. performance reports of multi-drive operation are especially interesting; my system cannot sustain full throughput on two drives on separate controllers, but that may just be a lame motherboard. THINGS I'M STILL MISSING CLUE ON: * who's responsible for configuring DMA timing modes on IDE drives? the BIOS or the driver? * is there a spec for dealing with Ultra-DMA extensions? * are there any chipsets or with bugs relating to DMA transfer that should be blacklisted? * are there any ATA interfaces that use some other kind of DMA controller in conjunction with standard ATA protocol? FINAL NOTE: after having looked at the ATA-3 spec, all i can say is, "it's ugly". *especially* electrically. the IDE bus is best modeled as an unterminated transmission line, these days. for maximum reliability, keep your IDE cables as short as possible and as few as possible. from what i can tell, most current chipsets have both IDE ports wired into a single buss, to a greater or lesser degree. using two cables means you double the length of this bus. SCSI may have its warts, but at least the basic analog design of the bus is still somewhat reasonable. IDE passed beyond the veil two years ago. --John Hood, cgull@smoke.marlboro.vt.us
1997-07-29 12:57:25 +00:00
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
(void)wdwait(du, 0, TIMEOUT);
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
outb(du->dk_altport, WDCTL_IDS | WDCTL_RST);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
DELAY(10 * 1000);
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
outb(du->dk_altport, WDCTL_IDS);
Make ATAPI support completely optional - if ATAPI isn't defined, the original code/algorithm is used completely (with one bugfix). This makes it "safe" to bring these changes into -stable.
1995-09-30 15:19:44 +00:00
#ifdef ATAPI
Bring IDE CDROM support up to latest version (1.8a?) from Serge. Submitted by: vak@cronyx.ru
1995-09-30 00:11:19 +00:00
if (wdwait(du, WDCS_READY | WDCS_SEEKCMPLT, TIMEOUT) != 0)
err = 1; /* no IDE drive found */
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
du->dk_error = inb(du->dk_port + wd_error);
Bring IDE CDROM support up to latest version (1.8a?) from Serge. Submitted by: vak@cronyx.ru
1995-09-30 00:11:19 +00:00
if (du->dk_error != 0x01)
err = 1; /* the drive is incompatible */
Make ATAPI support completely optional - if ATAPI isn't defined, the original code/algorithm is used completely (with one bugfix). This makes it "safe" to bring these changes into -stable.
1995-09-30 15:19:44 +00:00
#else
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
if (wdwait(du, WDCS_READY | WDCS_SEEKCMPLT, TIMEOUT) != 0) {
printf("wdreset: error1: 0x%x\n", du->dk_error);
Make ATAPI support completely optional - if ATAPI isn't defined, the original code/algorithm is used completely (with one bugfix). This makes it "safe" to bring these changes into -stable.
1995-09-30 15:19:44 +00:00
return (1);
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
}
Bring IDE CDROM support up to latest version (1.8a?) from Serge. Submitted by: vak@cronyx.ru
1995-09-30 00:11:19 +00:00
#endif
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
outb(du->dk_altport, WDCTL_4BIT);
Bring IDE CDROM support up to latest version (1.8a?) from Serge. Submitted by: vak@cronyx.ru
1995-09-30 00:11:19 +00:00
return (err);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
}
/*
* Sleep until driver is inactive.
* This is used only for avoiding rare race conditions, so it is unimportant
* that the sleep may be far too short or too long.
*/
static void
wdsleep(int ctrlr, char *wmesg)
{
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
int s = splbio();
Add support for the buggy CMD640B PCI EIDE controller chip, which can't perform overlapping commands on both of its channels. To enable the CMD640B work-around, the kernel must be compiled with "options CMD640". Without that option there should be no difference in the code produced compared to the previous revision of wd.c. Submitted by: Wolfgang Helbig <helbig@ba-stuttgart.de>
1997-03-11 23:17:28 +00:00
#ifdef CMD640
if (eide_quirks & Q_CMD640B)
ctrlr = PRIMARY;
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
while (wdtab[ctrlr].b_active)
tsleep((caddr_t)&wdtab[ctrlr].b_active, PZERO - 1, wmesg, 1);
Update the wd.c driver to use the new TAILQ scheme for device buffer queue. Also, create a new subroutine 'tqdisksort' that is an improved version of the original disksort that also uses TAILQs.
1995-11-23 07:24:41 +00:00
splx(s);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
}
Fixed up the compiler warnings.
1994-02-06 17:03:17 +00:00
static void
Ran ft.c through ident. Added a missing #ifdef INET wrapper in lpt.c Main change: Removed the timeout_func_t casts from timeout calls and correctly defined the timeout routines to conform to the new format. lpt.c doesn't have this change. Reviewed by: Submitted by:
1994-08-23 07:52:29 +00:00
wdtimeout(void *cdu)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
{
struct disk *du;
int x;
Extra intr's have nothing to do with LAPTOP. Made five-strikes-we-shut-up for timeouts. Inspired by: bde in 1.1.5.1-patch011
1994-10-27 05:39:12 +00:00
static int timeouts;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
du = (struct disk *)cdu;
x = splbio();
if (du->dk_timeout != 0 && --du->dk_timeout == 0) {
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
if(timeouts++ <= 5) {
char *msg;
msg = (timeouts > 5) ?
"Last time I say: interrupt timeout. Probably a portable PC." :
"interrupt timeout";
wderror((struct buf *)NULL, du, msg);
if (du->dk_dmacookie)
printf("wd%d: wdtimeout() DMA status %b\n",
du->dk_lunit,
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
wddma[du->dk_interface].wdd_dmastatus(du->dk_dmacookie),
Upgrade of EIDE DMA support, Johns comments: * lots of fixes to error handling-- mostly works now * improve DMA timing config for Triton chipsets-- PIIX4 and UDMA drive still untested * generally improve DMA config in many ways-- mostly cleanup * clean up boot-time messages * rewrite PRD generation algorithm * first wd timeout is now longer, to handle drive spinup Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-04 18:49:53 +00:00
WDDS_BITS);
}
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
wdunwedge(du);
wdflushirq(du, x);
du->dk_skip = 0;
du->dk_flags |= DKFL_SINGLE;
wdstart(du->dk_ctrlr);
}
Ran ft.c through ident. Added a missing #ifdef INET wrapper in lpt.c Main change: Removed the timeout_func_t casts from timeout calls and correctly defined the timeout routines to conform to the new format. lpt.c doesn't have this change. Reviewed by: Submitted by:
1994-08-23 07:52:29 +00:00
timeout(wdtimeout, cdu, hz);
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
splx(x);
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/*
* Reset the controller after it has become wedged. This is different from
* wdreset() so that wdreset() can be used in the probe and so that this
* can restore the geometry .
*/
static int
wdunwedge(struct disk *du)
{
struct disk *du1;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
int lunit;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/* Schedule other drives for recalibration. */
for (lunit = 0; lunit < NWD; lunit++)
if ((du1 = wddrives[lunit]) != NULL && du1 != du
&& du1->dk_ctrlr == du->dk_ctrlr
&& du1->dk_state > WANTOPEN)
du1->dk_state = WANTOPEN;
DELAY(RECOVERYTIME);
if (wdreset(du) == 0) {
/*
* XXX - recalibrate current drive now because some callers
* aren't prepared to have its state change.
*/
if (wdcommand(du, 0, 0, 0, 0, WDCC_RESTORE | WD_STEP) == 0
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
&& wdwait(du, WDCS_READY | WDCS_SEEKCMPLT, TIMEOUT) == 0
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
&& wdsetctlr(du) == 0)
return (0);
}
wderror((struct buf *)NULL, du, "wdunwedge failed");
return (1);
}
/*
* Wait uninterruptibly until controller is not busy and either certain
* status bits are set or an error has occurred.
* The wait is usually short unless it is for the controller to process
* an entire critical command.
* Return 1 for (possibly stale) controller errors, -1 for timeout errors,
* or 0 for no errors.
* Return controller status in du->dk_status and, if there was a controller
* error, return the error code in du->dk_error.
*/
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#ifdef WD_COUNT_RETRIES
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
static int min_retries[NWDC];
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
static int
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
wdwait(struct disk *du, u_char bits_wanted, int timeout)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
{
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
int wdc;
u_char status;
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
#define POLLING 1000
wdc = du->dk_port;
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
timeout += POLLING;
Slight change to wdwait() to improve performance while not violating the ATA spec. Submitted by: John Dyson
1995-02-27 06:42:35 +00:00
Change the initial delay in wdwait to a DELAY macro call. It is really too long, but does not impact performance very much when using the NSECS_MULTI option. This will help solve some mysterious IDE I/O problems.
1995-04-22 22:44:30 +00:00
/*
* This delay is really too long, but does not impact the performance
Implemented the multi-sector I/O enable and the 32 bit I/O probe through the use of the config file flags as opposed to the option "NSECS_MULTI". "NSECS_MULTI" has been removed from the driver. The new capability allows boot-time modification of the config.
1995-04-24 04:32:31 +00:00
* as much when using the multi-sector option. Shorter delays have
Change the initial delay in wdwait to a DELAY macro call. It is really too long, but does not impact performance very much when using the NSECS_MULTI option. This will help solve some mysterious IDE I/O problems.
1995-04-22 22:44:30 +00:00
* caused I/O errors on some drives and system configs. This should
* probably be fixed if we develop a better short term delay mechanism.
*/
DELAY(1);
Slight change to wdwait() to improve performance while not violating the ATA spec. Submitted by: John Dyson
1995-02-27 06:42:35 +00:00
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
do {
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#ifdef WD_COUNT_RETRIES
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (min_retries[du->dk_ctrlr] > timeout
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
|| min_retries[du->dk_ctrlr] == 0)
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
min_retries[du->dk_ctrlr] = timeout;
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
du->dk_status = status = inb(wdc + wd_status);
Latest fixes from Serge: I tried to solve the problem of IDE probing compatibility in this version. When compiled without an ATAPI option, the wd driver is fully backward compatible with 2.0.5. With ATAPI option, the wdprobe becomes strictly weaker. That is, if wdprobe works without ATAPI option, it will always work with it too. Another problem was with the CD-ROM drive attached as a slave in the IDE bus, where there is no master. All IDE CD-ROM drives are shipped in slave configuration, and most users just plug them in, never thinking about jumpers. It works fine with ms-dos and ms-windows, and this version of the driver supports it as well. The eject op can now load disks. Just repeat it twice, and the disk will be ejected and then loaded back. The disc cannot be ejected if it is mounted. Submitted by: Serge Vakulenko, <vak@cronyx.ru>
1995-10-14 15:41:10 +00:00
#ifdef ATAPI
/*
* Atapi drives have a very interesting feature, when attached
* as a slave on the IDE bus, and there is no master.
* They release the bus after getting the command.
* We should reselect the drive here to get the status.
*/
if (status == 0xff) {
outb(wdc + wd_sdh, WDSD_IBM | du->dk_unit << 4);
du->dk_status = status = inb(wdc + wd_status);
}
#endif
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
if (!(status & WDCS_BUSY)) {
if (status & WDCS_ERR) {
du->dk_error = inb(wdc + wd_error);
/*
* We once returned here. This is wrong
* because the error bit is apparently only
* valid after the controller has interrupted
* (e.g., the error bit is stale when we wait
* for DRQ for writes). So we can't depend
* on the error bit at all when polling for
* command completion.
*/
}
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
if ((status & bits_wanted) == bits_wanted) {
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (status & WDCS_ERR);
Addition of support of the slightly rogue Promise IDE interface(Dyson), support of multiple PCI IDE controllers(Dyson), and some updates and cleanups from John Hood, who originally made our IDE DMA stuff work :-). I have run tests with 7 IDE drives connected to my system, all in DMA mode, with no errors. Modulo any bugs, this stuff makes IDE look really good (within it's limitations.) Submitted by: John Hood <cgull@smoke.marlboro.vt.us>
1997-09-20 07:41:58 +00:00
}
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
}
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
if (timeout < TIMEOUT)
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
/*
* Switch to a polling rate of about 1 KHz so that
* the timeout is almost machine-independent. The
* controller is taking a long time to respond, so
* an extra msec won't matter.
*/
DELAY(1000);
Slight change to wdwait() to improve performance while not violating the ATA spec. Submitted by: John Dyson
1995-02-27 06:42:35 +00:00
else
DELAY(1);
Added third parameter to wdwait which contains the amount of time to 'wait' based on suggestion by Rod in -hackers. No functional change as all the TIMEOUTS are identical currently, but this will probably change.
1994-02-11 12:02:35 +00:00
} while (--timeout != 0);
New wd driver, based on Bruce Evans 'wx/altwd' driver which has passed enough tests to be considered more stable than current driver. Lots of work by Bruce, David G., and Guido have gone into this version, and more is to come in the future. Support for multiple controllers is in, but doesn't work correctly with different controllers (IDE AND MFM), but multiple alike controllers appears to work. Most of the stray interrupts problems should be fixed, although you will get a couple 'extra interrupts' when disklabeling and on startup.
1994-01-04 20:05:26 +00:00
return (-1);
}
the second set of changes in a move towards getting devices to be totally dynamic. this is only the devices in i386/isa I'll do more tomorrow. they're completely masked by #ifdef JREMOD at this stage... the eventual aim is that every driver will do a SYSINIT at startup BEFORE the probes, which will effectively link it into the devsw tables etc. If I'd thought about it more I'd have put that in in this set (damn) The ioconf lines generated by config will also end up in the device's own scope as well, so ioconf.c will eventually be gutted the SYSINIT call to the driver will include a phase where the driver links it's ioconf line into a chain of such. when this phase is done then the user can modify them with the boot: -c config menu if he wants, just like now.. config will put the config lines out in the .h file (e.g. in aha.h will be the addresses for the aha driver to look.) as I said this is a very small first step.. the aim of THIS set of edits is to not have to edit conf.c at all when adding a new device.. the tabe will be a simple skeleton.. when this is done, it will allow other changes to be made, all teh time still having a fully working kernel tree, but the logical outcome is the complete REMOVAL of the devsw tables. By the end of this, linked in drivers will be exactly the same as run-time loaded drivers, except they JUST HAPPEN to already be linked and present at startup.. the SYSINIT calls will be the equivalent of the "init" call made to a newly loaded driver in every respect. For this edit, each of the files has the following code inserted into it: obviously, tailored to suit.. ----------------------somewhere at the top: #ifdef JREMOD #include <sys/conf.h> #define CDEV_MAJOR 13 #define BDEV_MAJOR 4 static void sd_devsw_install(); #endif /*JREMOD */ ---------------------somewhere that's run during bootup: EVENTUALLY a SYSINIT #ifdef JREMOD sd_devsw_install(); #endif /*JREMOD*/ -----------------------at the bottom: #ifdef JREMOD struct bdevsw sd_bdevsw = { sdopen, sdclose, sdstrategy, sdioctl, /*4*/ sddump, sdsize, 0 }; struct cdevsw sd_cdevsw = { sdopen, sdclose, rawread, rawwrite, /*13*/ sdioctl, nostop, nullreset, nodevtotty,/* sd */ seltrue, nommap, sdstrategy }; static sd_devsw_installed = 0; static void sd_devsw_install() { dev_t descript; if( ! sd_devsw_installed ) { descript = makedev(CDEV_MAJOR,0); cdevsw_add(&descript,&sd_cdevsw,NULL); #if defined(BDEV_MAJOR) descript = makedev(BDEV_MAJOR,0); bdevsw_add(&descript,&sd_bdevsw,NULL); #endif /*BDEV_MAJOR*/ sd_devsw_installed = 1; } } #endif /* JREMOD */
1995-11-28 09:42:06 +00:00
static wd_devsw_installed = 0;
OK, that's it.. That's EVERY SINGLE driver that has an entry in conf.c.. my next trick will be to define cdevsw[] and bdevsw[] as empty arrays and remove all those DAMNED defines as well.. Each of these drivers has a SYSINIT linker set entry that comes in very early.. and asks teh driver to add it's own entry to the two devsw[] tables. some slight reworking of the commits from yesterday (added the SYSINIT stuff and some usually wrong but token DEVFS entries to all these devices. BTW does anyone know where the 'ata' entries in conf.c actually reside? seems we don't actually have a 'ataopen() etc... If you want to add a new device in conf.c please make sure I know so I can keep it up to date too.. as before, this is all dependent on #if defined(JREMOD) (and #ifdef DEVFS in parts)
1995-11-29 10:49:16 +00:00
static void wd_drvinit(void *unused)
the second set of changes in a move towards getting devices to be totally dynamic. this is only the devices in i386/isa I'll do more tomorrow. they're completely masked by #ifdef JREMOD at this stage... the eventual aim is that every driver will do a SYSINIT at startup BEFORE the probes, which will effectively link it into the devsw tables etc. If I'd thought about it more I'd have put that in in this set (damn) The ioconf lines generated by config will also end up in the device's own scope as well, so ioconf.c will eventually be gutted the SYSINIT call to the driver will include a phase where the driver links it's ioconf line into a chain of such. when this phase is done then the user can modify them with the boot: -c config menu if he wants, just like now.. config will put the config lines out in the .h file (e.g. in aha.h will be the addresses for the aha driver to look.) as I said this is a very small first step.. the aim of THIS set of edits is to not have to edit conf.c at all when adding a new device.. the tabe will be a simple skeleton.. when this is done, it will allow other changes to be made, all teh time still having a fully working kernel tree, but the logical outcome is the complete REMOVAL of the devsw tables. By the end of this, linked in drivers will be exactly the same as run-time loaded drivers, except they JUST HAPPEN to already be linked and present at startup.. the SYSINIT calls will be the equivalent of the "init" call made to a newly loaded driver in every respect. For this edit, each of the files has the following code inserted into it: obviously, tailored to suit.. ----------------------somewhere at the top: #ifdef JREMOD #include <sys/conf.h> #define CDEV_MAJOR 13 #define BDEV_MAJOR 4 static void sd_devsw_install(); #endif /*JREMOD */ ---------------------somewhere that's run during bootup: EVENTUALLY a SYSINIT #ifdef JREMOD sd_devsw_install(); #endif /*JREMOD*/ -----------------------at the bottom: #ifdef JREMOD struct bdevsw sd_bdevsw = { sdopen, sdclose, sdstrategy, sdioctl, /*4*/ sddump, sdsize, 0 }; struct cdevsw sd_cdevsw = { sdopen, sdclose, rawread, rawwrite, /*13*/ sdioctl, nostop, nullreset, nodevtotty,/* sd */ seltrue, nommap, sdstrategy }; static sd_devsw_installed = 0; static void sd_devsw_install() { dev_t descript; if( ! sd_devsw_installed ) { descript = makedev(CDEV_MAJOR,0); cdevsw_add(&descript,&sd_cdevsw,NULL); #if defined(BDEV_MAJOR) descript = makedev(BDEV_MAJOR,0); bdevsw_add(&descript,&sd_bdevsw,NULL); #endif /*BDEV_MAJOR*/ sd_devsw_installed = 1; } } #endif /* JREMOD */
1995-11-28 09:42:06 +00:00
{
OK, that's it.. That's EVERY SINGLE driver that has an entry in conf.c.. my next trick will be to define cdevsw[] and bdevsw[] as empty arrays and remove all those DAMNED defines as well.. Each of these drivers has a SYSINIT linker set entry that comes in very early.. and asks teh driver to add it's own entry to the two devsw[] tables. some slight reworking of the commits from yesterday (added the SYSINIT stuff and some usually wrong but token DEVFS entries to all these devices. BTW does anyone know where the 'ata' entries in conf.c actually reside? seems we don't actually have a 'ataopen() etc... If you want to add a new device in conf.c please make sure I know so I can keep it up to date too.. as before, this is all dependent on #if defined(JREMOD) (and #ifdef DEVFS in parts)
1995-11-29 10:49:16 +00:00
the second set of changes in a move towards getting devices to be totally dynamic. this is only the devices in i386/isa I'll do more tomorrow. they're completely masked by #ifdef JREMOD at this stage... the eventual aim is that every driver will do a SYSINIT at startup BEFORE the probes, which will effectively link it into the devsw tables etc. If I'd thought about it more I'd have put that in in this set (damn) The ioconf lines generated by config will also end up in the device's own scope as well, so ioconf.c will eventually be gutted the SYSINIT call to the driver will include a phase where the driver links it's ioconf line into a chain of such. when this phase is done then the user can modify them with the boot: -c config menu if he wants, just like now.. config will put the config lines out in the .h file (e.g. in aha.h will be the addresses for the aha driver to look.) as I said this is a very small first step.. the aim of THIS set of edits is to not have to edit conf.c at all when adding a new device.. the tabe will be a simple skeleton.. when this is done, it will allow other changes to be made, all teh time still having a fully working kernel tree, but the logical outcome is the complete REMOVAL of the devsw tables. By the end of this, linked in drivers will be exactly the same as run-time loaded drivers, except they JUST HAPPEN to already be linked and present at startup.. the SYSINIT calls will be the equivalent of the "init" call made to a newly loaded driver in every respect. For this edit, each of the files has the following code inserted into it: obviously, tailored to suit.. ----------------------somewhere at the top: #ifdef JREMOD #include <sys/conf.h> #define CDEV_MAJOR 13 #define BDEV_MAJOR 4 static void sd_devsw_install(); #endif /*JREMOD */ ---------------------somewhere that's run during bootup: EVENTUALLY a SYSINIT #ifdef JREMOD sd_devsw_install(); #endif /*JREMOD*/ -----------------------at the bottom: #ifdef JREMOD struct bdevsw sd_bdevsw = { sdopen, sdclose, sdstrategy, sdioctl, /*4*/ sddump, sdsize, 0 }; struct cdevsw sd_cdevsw = { sdopen, sdclose, rawread, rawwrite, /*13*/ sdioctl, nostop, nullreset, nodevtotty,/* sd */ seltrue, nommap, sdstrategy }; static sd_devsw_installed = 0; static void sd_devsw_install() { dev_t descript; if( ! sd_devsw_installed ) { descript = makedev(CDEV_MAJOR,0); cdevsw_add(&descript,&sd_cdevsw,NULL); #if defined(BDEV_MAJOR) descript = makedev(BDEV_MAJOR,0); bdevsw_add(&descript,&sd_bdevsw,NULL); #endif /*BDEV_MAJOR*/ sd_devsw_installed = 1; } } #endif /* JREMOD */
1995-11-28 09:42:06 +00:00
if( ! wd_devsw_installed ) {
Make a "DWIM" function for adding [bc]devsw entries for bdev drivers. Saves about 280 butes of source per driver, 56 bytes in object size and another 56 bytes moves from data to bss. No functional change intended nor expected. GENERIC should be about one k smaller now :-)
1996-07-23 21:52:43 +00:00
bdevsw_add_generic(BDEV_MAJOR,CDEV_MAJOR, &wd_bdevsw);
the second set of changes in a move towards getting devices to be totally dynamic. this is only the devices in i386/isa I'll do more tomorrow. they're completely masked by #ifdef JREMOD at this stage... the eventual aim is that every driver will do a SYSINIT at startup BEFORE the probes, which will effectively link it into the devsw tables etc. If I'd thought about it more I'd have put that in in this set (damn) The ioconf lines generated by config will also end up in the device's own scope as well, so ioconf.c will eventually be gutted the SYSINIT call to the driver will include a phase where the driver links it's ioconf line into a chain of such. when this phase is done then the user can modify them with the boot: -c config menu if he wants, just like now.. config will put the config lines out in the .h file (e.g. in aha.h will be the addresses for the aha driver to look.) as I said this is a very small first step.. the aim of THIS set of edits is to not have to edit conf.c at all when adding a new device.. the tabe will be a simple skeleton.. when this is done, it will allow other changes to be made, all teh time still having a fully working kernel tree, but the logical outcome is the complete REMOVAL of the devsw tables. By the end of this, linked in drivers will be exactly the same as run-time loaded drivers, except they JUST HAPPEN to already be linked and present at startup.. the SYSINIT calls will be the equivalent of the "init" call made to a newly loaded driver in every respect. For this edit, each of the files has the following code inserted into it: obviously, tailored to suit.. ----------------------somewhere at the top: #ifdef JREMOD #include <sys/conf.h> #define CDEV_MAJOR 13 #define BDEV_MAJOR 4 static void sd_devsw_install(); #endif /*JREMOD */ ---------------------somewhere that's run during bootup: EVENTUALLY a SYSINIT #ifdef JREMOD sd_devsw_install(); #endif /*JREMOD*/ -----------------------at the bottom: #ifdef JREMOD struct bdevsw sd_bdevsw = { sdopen, sdclose, sdstrategy, sdioctl, /*4*/ sddump, sdsize, 0 }; struct cdevsw sd_cdevsw = { sdopen, sdclose, rawread, rawwrite, /*13*/ sdioctl, nostop, nullreset, nodevtotty,/* sd */ seltrue, nommap, sdstrategy }; static sd_devsw_installed = 0; static void sd_devsw_install() { dev_t descript; if( ! sd_devsw_installed ) { descript = makedev(CDEV_MAJOR,0); cdevsw_add(&descript,&sd_cdevsw,NULL); #if defined(BDEV_MAJOR) descript = makedev(BDEV_MAJOR,0); bdevsw_add(&descript,&sd_bdevsw,NULL); #endif /*BDEV_MAJOR*/ sd_devsw_installed = 1; } } #endif /* JREMOD */
1995-11-28 09:42:06 +00:00
wd_devsw_installed = 1;
If you're going to mechanically replicate something in 50 files it's best to not have a (compiles cleanly) typo in it! (sigh)
1995-11-29 14:41:20 +00:00
}
the second set of changes in a move towards getting devices to be totally dynamic. this is only the devices in i386/isa I'll do more tomorrow. they're completely masked by #ifdef JREMOD at this stage... the eventual aim is that every driver will do a SYSINIT at startup BEFORE the probes, which will effectively link it into the devsw tables etc. If I'd thought about it more I'd have put that in in this set (damn) The ioconf lines generated by config will also end up in the device's own scope as well, so ioconf.c will eventually be gutted the SYSINIT call to the driver will include a phase where the driver links it's ioconf line into a chain of such. when this phase is done then the user can modify them with the boot: -c config menu if he wants, just like now.. config will put the config lines out in the .h file (e.g. in aha.h will be the addresses for the aha driver to look.) as I said this is a very small first step.. the aim of THIS set of edits is to not have to edit conf.c at all when adding a new device.. the tabe will be a simple skeleton.. when this is done, it will allow other changes to be made, all teh time still having a fully working kernel tree, but the logical outcome is the complete REMOVAL of the devsw tables. By the end of this, linked in drivers will be exactly the same as run-time loaded drivers, except they JUST HAPPEN to already be linked and present at startup.. the SYSINIT calls will be the equivalent of the "init" call made to a newly loaded driver in every respect. For this edit, each of the files has the following code inserted into it: obviously, tailored to suit.. ----------------------somewhere at the top: #ifdef JREMOD #include <sys/conf.h> #define CDEV_MAJOR 13 #define BDEV_MAJOR 4 static void sd_devsw_install(); #endif /*JREMOD */ ---------------------somewhere that's run during bootup: EVENTUALLY a SYSINIT #ifdef JREMOD sd_devsw_install(); #endif /*JREMOD*/ -----------------------at the bottom: #ifdef JREMOD struct bdevsw sd_bdevsw = { sdopen, sdclose, sdstrategy, sdioctl, /*4*/ sddump, sdsize, 0 }; struct cdevsw sd_cdevsw = { sdopen, sdclose, rawread, rawwrite, /*13*/ sdioctl, nostop, nullreset, nodevtotty,/* sd */ seltrue, nommap, sdstrategy }; static sd_devsw_installed = 0; static void sd_devsw_install() { dev_t descript; if( ! sd_devsw_installed ) { descript = makedev(CDEV_MAJOR,0); cdevsw_add(&descript,&sd_cdevsw,NULL); #if defined(BDEV_MAJOR) descript = makedev(BDEV_MAJOR,0); bdevsw_add(&descript,&sd_bdevsw,NULL); #endif /*BDEV_MAJOR*/ sd_devsw_installed = 1; } } #endif /* JREMOD */
1995-11-28 09:42:06 +00:00
}
OK, that's it.. That's EVERY SINGLE driver that has an entry in conf.c.. my next trick will be to define cdevsw[] and bdevsw[] as empty arrays and remove all those DAMNED defines as well.. Each of these drivers has a SYSINIT linker set entry that comes in very early.. and asks teh driver to add it's own entry to the two devsw[] tables. some slight reworking of the commits from yesterday (added the SYSINIT stuff and some usually wrong but token DEVFS entries to all these devices. BTW does anyone know where the 'ata' entries in conf.c actually reside? seems we don't actually have a 'ataopen() etc... If you want to add a new device in conf.c please make sure I know so I can keep it up to date too.. as before, this is all dependent on #if defined(JREMOD) (and #ifdef DEVFS in parts)
1995-11-29 10:49:16 +00:00
SYSINIT(wddev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE+CDEV_MAJOR,wd_drvinit,NULL)
From: bde@kralizec.zeta.org.au (Bruce Evans) o merged and fixed timeout code from sos. o merged DOS partition support from sos. o fixed "extra" interrupts. o check if malloc retuns NULL :-(. o print drive size as in NetBSD o after an error, return the residual count, not 0. o give up early for > 16 heads. o cleaned up the old-drive detection in wdgetctlr(). o rewrote wddump(). I'd been putting this off because I'd had the dump disabled since 0.0 and had no idea if even the standard version worked. It didn't: (0) syscons' sgetc() interface broke some keyboard stuff. (1) CADDR1 is mapped using pmap() and pmap got a NULL pointer trap (at least when I got to wddump() using call diediedie from ddb) because the pte for CADDR1 is only supposed to be hacked on directly (e.g. in physcopyseg()). (2) bad144 handling was not done. (3) it was slooow (3600 sectors/minute) because my controller doesn't cache writes. o miscellaneous other cleanups, e.g., removed scattered patchkit/terry dates. o lots of reformatting. To be done: o Merge/fix TIH/NetBSD bad144 code (doesn't belong in any particular driver. Why aren't we using i386/dkbad.c?). o Merge/fix Dyson/NetBSD clustering code (large parts should be shared). o Fix/extend the partition in use bits. Support extended partitions. This should be shared by all disk drivers. Swap to a DOS partition so that the swap space can be shared with linux. o Don't use polling except for initialization. Need to reorganise the state machine. Then "extra" interrupts shouldn't happen (except maybe one for initialization). o Fix disklabel, boot and driver inconsistencies with bad144 in standard versions. o Look at latest linux clustering methods. Our disksort() gets in the way of clustering. o cleaned up the ATA changes (needs more work. I think the ATA specs say that the only thing really wrong with the original version of wx is that there was no test for drive-ready before some commands. 400 nsec delay suffices for almost everything and 400 nsec delay is usually "free").
1994-02-01 05:55:21 +00:00
#endif /* NWDC > 0 */
Reference in New Issue Copy Permalink