mirror/freebsd
1
0
Fork 0
mirror of https://git.FreeBSD.org/src.git synced 2024-12-23 11:18:54 +00:00
Code Issues Releases Activity
70af957f57
freebsd/sys/i386/scsi/aic7xxx.c

2448 lines
63 KiB
C
Raw Normal View History

My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
* Generic driver for the aic7xxx based adaptec SCSI controllers
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
* Copyright (c) 1994, 1995, 1996 Justin T. Gibbs.
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
* All rights reserved.
*
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
* Product specific probe and attach routines can be found in:
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
* i386/eisa/aic7770.c 27/284X and aic7770 motherboard controllers
* pci/aic7870.c 3940, 2940, aic7870 and aic7850 controllers
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
*
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* Portions of this driver are based on the FreeBSD 1742 Driver:
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
*
* Written by Julian Elischer (julian@tfs.com)
* for TRW Financial Systems for use under the MACH(2.5) operating system.
*
* TRW Financial Systems, in accordance with their agreement with Carnegie
* Mellon University, makes this software available to CMU to distribute
* or use in any manner that they see fit as long as this message is kept with
* the software. For this reason TFS also grants any other persons or
* organisations permission to use or modify this software.
*
* TFS supplies this software to be publicly redistributed
* on the understanding that TFS is not responsible for the correct
* functioning of this software in any circumstances.
*
* commenced: Sun Sep 27 18:14:01 PDT 1992
*
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
* $Id: aic7xxx.c,v 1.59 1996/03/23 11:29:20 phk Exp $
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
*/
/*
* TODO:
* Implement Target Mode
*
*/
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/buf.h>
#include <sys/proc.h>
Replaced #includes of <sys/user.h> by less gross headers, usually <sys/vm.h>. Many device drivers need only the definition of vtophys() from vm. Added nearby #includes of <sys/conf.h> where appropriate.
1995-12-06 23:52:35 +00:00
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
#include <scsi/scsi_all.h>
#include <scsi/scsiconf.h>
Replaced #includes of <sys/user.h> by less gross headers, usually <sys/vm.h>. Many device drivers need only the definition of vtophys() from vm. Added nearby #includes of <sys/conf.h> where appropriate.
1995-12-06 23:52:35 +00:00
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
#include <machine/clock.h>
Replaced #includes of <sys/user.h> by less gross headers, usually <sys/vm.h>. Many device drivers need only the definition of vtophys() from vm. Added nearby #includes of <sys/conf.h> where appropriate.
1995-12-06 23:52:35 +00:00
#include <vm/vm.h>
Untangled the vm.h include file spaghetti.
1995-12-07 12:48:31 +00:00
#include <vm/vm_param.h>
#include <vm/pmap.h>
Replaced #includes of <sys/user.h> by less gross headers, usually <sys/vm.h>. Many device drivers need only the definition of vtophys() from vm. Added nearby #includes of <sys/conf.h> where appropriate.
1995-12-06 23:52:35 +00:00
Point include to i386/scsi
1995-01-13 02:27:08 +00:00
#include <i386/scsi/aic7xxx.h>
Replaced #includes of <sys/user.h> by less gross headers, usually <sys/vm.h>. Many device drivers need only the definition of vtophys() from vm. Added nearby #includes of <sys/conf.h> where appropriate.
1995-12-06 23:52:35 +00:00
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
#include <dev/aic7xxx/aic7xxx_reg.h>
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
#define PAGESIZ 4096
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
#include <sys/kernel.h>
#define KVTOPHYS(x) vtophys(x)
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
#define MIN(a,b) ((a < b) ? a : b)
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
#define ALL_TARGETS -1
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
-Wall fixes.
1995-11-06 05:21:13 +00:00
u_long ahc_unit = 0;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
static int ahc_debug = AHC_SHOWABORTS|AHC_SHOWMISC;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/**** bit definitions for SCSIDEF ****/
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
#define HSCSIID 0x07 /* our SCSI ID */
#define HWSCSIID 0x0f /* our SCSI ID if Wide Bus */
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
static u_int32_t ahc_adapter_info __P((int unit));
static void ahcminphys __P((struct buf *bp));
static int32_t ahc_scsi_cmd __P((struct scsi_xfer *xs));
Completed function declarations and added prototypes. Sorted prototypes.
1995-12-15 23:49:42 +00:00
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static struct scsi_adapter ahc_switch =
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
{
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahc_scsi_cmd,
ahcminphys,
0,
0,
ahc_adapter_info,
"ahc",
{ 0, 0 }
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
};
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/* the below structure is so we have a default dev struct for our link struct */
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static struct scsi_device ahc_dev =
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
NULL, /* Use default error handler */
NULL, /* have a queue, served by this */
NULL, /* have no async handler */
NULL, /* Use default 'done' routine */
"ahc",
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
0,
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{ 0, 0 }
};
/*
* Since the sequencer can disable pausing in a critical section, we
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* must loop until it actually stops.
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
* XXX Should add a timeout in here??
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
#define PAUSE_SEQUENCER(ahc) \
outb(HCNTRL + ahc->baseport, ahc->pause); \
\
while ((inb(HCNTRL + ahc->baseport) & PAUSE) == 0) \
;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
#define UNPAUSE_SEQUENCER(ahc) \
outb( HCNTRL + ahc->baseport, ahc->unpause )
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* Restart the sequencer program from address zero
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
#define RESTART_SEQUENCER(ahc) \
do { \
outb( SEQCTL + ahc->baseport, SEQRESET|FASTMODE ); \
} while (inb(SEQADDR0 + ahc->baseport) != 0 && \
inb(SEQADDR1 + ahc->baseport != 0)); \
\
UNPAUSE_SEQUENCER(ahc);
static u_char ahc_abort_wscb __P((struct ahc_data *ahc, struct scb *scbp,
u_char prev, u_long iobase,
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
u_char timedout_scb, u_int32_t xs_error));
static void ahc_add_waiting_scb __P((u_long iobase, struct scb *scb));
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
static void ahc_done __P((struct ahc_data *ahc, struct scb *scbp));
static void ahc_free_scb __P((struct ahc_data *ahc, struct scb *scb,
int flags));
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
static inline void ahc_fetch_scb __P((struct ahc_data *ahc, struct scb *scb,
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
int iosize));
Completed function declarations and added prototypes. Sorted prototypes.
1995-12-15 23:49:42 +00:00
static struct scb *
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_get_scb __P((struct ahc_data *ahc, int flags));
Completed function declarations and added prototypes. Sorted prototypes.
1995-12-15 23:49:42 +00:00
static void ahc_loadseq __P((u_long iobase));
static int ahc_match_scb __P((struct scb *scb, int target, char channel));
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
static int ahc_poll __P((struct ahc_data *ahc, int wait));
Completed function declarations and added prototypes. Sorted prototypes.
1995-12-15 23:49:42 +00:00
#ifdef AHC_DEBUG
static void ahc_print_active_scb __P((struct ahc_data *ahc));
static void ahc_print_scb __P((struct scb *scb));
#endif
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
static int ahc_reset_channel __P((struct ahc_data *ahc, char channel,
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
u_char timedout_scb, u_int32_t xs_error,
u_char initiate_reset));
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
static int ahc_reset_device __P((struct ahc_data *ahc, int target,
char channel, u_char timedout_scb,
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
u_int32_t xs_error));
Completed function declarations and added prototypes. Sorted prototypes.
1995-12-15 23:49:42 +00:00
static void ahc_reset_current_bus __P((u_long iobase));
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
static void ahc_run_done_queue __P((struct ahc_data *ahc));
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
static void ahc_scsirate __P((struct ahc_data* ahc, u_char *scsirate,
int period, int offset, int target));
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
static inline void ahc_send_scb __P((struct ahc_data *ahc, struct scb *scb));
Completed function declarations and added prototypes. Sorted prototypes.
1995-12-15 23:49:42 +00:00
static timeout_t
ahc_timeout;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
static void ahc_busy_target __P((int target, char channel,
u_long iobase));
Completed function declarations and added prototypes. Sorted prototypes.
1995-12-15 23:49:42 +00:00
static void ahc_unbusy_target __P((int target, char channel,
u_long iobase));
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
#ifdef AHC_DEBUG
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahc_print_scb(scb)
struct scb *scb;
{
-Wall fixes.
1995-11-06 05:21:13 +00:00
printf("scb:%p control:0x%x tcl:0x%x cmdlen:%d cmdpointer:0x%lx\n"
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
,scb
,scb->control
,scb->target_channel_lun
,scb->cmdlen
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
,scb->cmdpointer );
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
printf(" datlen:%d data:0x%lx segs:0x%x segp:0x%lx\n"
,scb->datalen
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
,scb->data
,scb->SG_segment_count
,scb->SG_list_pointer);
-Wall fixes.
1995-11-06 05:21:13 +00:00
printf(" sg_addr:%lx sg_len:%ld\n"
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
,scb->ahc_dma[0].addr
,scb->ahc_dma[0].len);
}
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
ahc_print_active_scb(ahc)
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
struct ahc_data *ahc;
{
int cur_scb_offset;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
u_long iobase = ahc->baseport;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
PAUSE_SEQUENCER(ahc);
cur_scb_offset = inb(SCBPTR + iobase);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
UNPAUSE_SEQUENCER(ahc);
ahc_print_scb(ahc->scbarray[cur_scb_offset]);
}
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
#endif
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
#define PARERR 0x08
#define ILLOPCODE 0x04
#define ILLSADDR 0x02
#define ILLHADDR 0x01
static struct {
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
u_char errno;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
char *errmesg;
} hard_error[] = {
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
{ ILLHADDR, "Illegal Host Access" },
{ ILLSADDR, "Illegal Sequencer Address referrenced" },
{ ILLOPCODE, "Illegal Opcode in sequencer program" },
{ PARERR, "Sequencer Ram Parity Error" }
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
};
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* Valid SCSIRATE values. (p. 3-17)
* Provides a mapping of tranfer periods in ns to the proper value to
* stick in the scsiscfr reg to use that transfer rate.
*/
static struct {
short sxfr;
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
/* Rates in Ultra mode have bit 8 of sxfr set */
#define ULTRA_SXFR 0x100
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
short period; /* in ns */
char *rate;
} ahc_syncrates[] = {
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
{ 0x100, 50, "20.0" },
{ 0x110, 62, "16.0" },
{ 0x120, 75, "13.4" },
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
{ 0x130, 87, "11.4" },
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
{ 0x140, 100, "10.0" },
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
{ 0x150, 112, "8.8" },
{ 0x160, 125, "8.0" },
{ 0x170, 137, "7.2" },
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
{ 0x000, 100, "10.0" },
{ 0x010, 125, "8.0" },
{ 0x020, 150, "6.67" },
{ 0x030, 175, "5.7" },
{ 0x040, 200, "5.0" },
{ 0x050, 225, "4.4" },
{ 0x060, 250, "4.0" },
{ 0x070, 275, "3.6" }
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
};
static int ahc_num_syncrates =
sizeof(ahc_syncrates) / sizeof(ahc_syncrates[0]);
/*
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
* Allocate a controller structures for a new device and initialize it.
* ahc_reset should be called before now since we assume that the card
* is paused.
*
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
*/
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
struct ahc_data *
ahc_alloc(unit, iobase, type, flags)
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
int unit;
u_long iobase;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahc_type type;
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
ahc_flag flags;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
/*
* find unit and check we have that many defined
*/
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
struct ahc_data *ahc;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
/*
* Allocate a storage area for us
*/
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc = malloc(sizeof(struct ahc_data), M_TEMP, M_NOWAIT);
if (!ahc) {
printf("ahc%d: cannot malloc!\n", unit);
return NULL;
}
bzero(ahc, sizeof(struct ahc_data));
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
SLIST_INIT(&ahc->free_scb);
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
ahc->unit = unit;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
ahc->baseport = iobase;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahc->type = type;
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
ahc->flags = flags;
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
ahc->unpause = (inb(HCNTRL + iobase) & IRQMS) | INTEN;
ahc->pause = ahc->unpause | PAUSE;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
return (ahc);
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
void
ahc_free(ahc)
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
struct ahc_data *ahc;
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
{
free(ahc, M_DEVBUF);
return;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
void
ahc_reset(iobase)
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
u_long iobase;
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
{
u_char hcntrl;
int wait;
/* Retain the IRQ type accross the chip reset */
hcntrl = (inb(HCNTRL + iobase) & IRQMS) | INTEN;
outb(HCNTRL + iobase, CHIPRST | PAUSE);
/*
* Ensure that the reset has finished
*/
wait = 1000;
while (wait--) {
DELAY(1000);
if(!(inb(HCNTRL + iobase) & CHIPRST))
break;
}
if(wait == 0) {
-Wall fixes.
1995-11-06 05:21:13 +00:00
printf("ahc at 0x%lx: WARNING - Failed chip reset! "
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
"Trying to initialize anyway.\n", iobase);
}
outb(HCNTRL + iobase, hcntrl | PAUSE);
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* Look up the valid period to SCSIRATE conversion in our table.
*/
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_scsirate(ahc, scsirate, period, offset, target )
struct ahc_data *ahc;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
u_char *scsirate;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
short period;
u_char offset;
int target;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
int i;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
for (i = 0; i < ahc_num_syncrates; i++) {
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if ((ahc_syncrates[i].period - period) >= 0) {
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
/*
* Watch out for Ultra speeds when ultra is not
* enabled and vice-versa.
*/
if (ahc->type & AHC_ULTRA) {
if (!(ahc_syncrates[i].sxfr & ULTRA_SXFR)) {
printf("ahc%d: target %d requests "
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
"%sMHz transfers, but adapter "
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
"in Ultra mode can only sync at "
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
"10MHz or above\n", ahc->unit,
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
target, ahc_syncrates[i].rate);
break; /* Use Async */
}
}
else {
if (ahc_syncrates[i].sxfr & ULTRA_SXFR) {
/*
* This should only happen if the
* drive is the first to negotiate
* and chooses a high rate. We'll
* just move down the table util
* we hit a non ultra speed.
*/
continue;
}
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
*scsirate = (ahc_syncrates[i].sxfr) | (offset & 0x0f);
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(bootverbose) {
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
printf("ahc%d: target %d synchronous at %sMHz,"
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
" offset = 0x%x\n", ahc->unit, target,
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
ahc_syncrates[i].rate, offset );
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
return;
}
}
Be more careful with how we do a chip reset. Clean up some comments.
1995-04-15 21:37:32 +00:00
/* Default to asyncronous transfers. Also reject this SDTR request. */
*scsirate = 0;
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(bootverbose) {
printf("ahc%d: target %d using asyncronous transfers\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, target );
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
/*
* Attach all the sub-devices we can find
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
*/
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
int
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_attach(ahc)
struct ahc_data *ahc;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
struct scsibus_data *scbus;
/*
* fill in the prototype scsi_link.
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->sc_link.adapter_unit = ahc->unit;
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
ahc->sc_link.adapter_targ = ahc->our_id;
Use the new adapter_softc field in the scsi_link structure so that these drivers don't need to maintain an array of configured units. The bt driver still needs to because ISA interrupt handlers take a unit number. :(
1996-01-07 19:24:36 +00:00
ahc->sc_link.adapter_softc = ahc;
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
ahc->sc_link.adapter = &ahc_switch;
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
ahc->sc_link.opennings = 2;
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
ahc->sc_link.device = &ahc_dev;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahc->sc_link.flags = DEBUGLEVEL;
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
ahc->sc_link.fordriver = 0;
Register with devconfg so we show up in lsdev.
1994-11-18 20:34:30 +00:00
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
/*
* Prepare the scsibus_data area for the upperlevel
* scsi code.
*/
scbus = scsi_alloc_bus();
if(!scbus)
return 0;
scbus->adapter_link = &ahc->sc_link;
if(ahc->type & AHC_WIDE)
scbus->maxtarg = 15;
/*
* ask the adapter what subunits are present
*/
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(bootverbose)
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
printf("ahc%d: Probing channel A\n", ahc->unit);
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
scsi_attachdevs(scbus);
scbus = NULL; /* Upper-level SCSI code owns this now */
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
if(ahc->type & AHC_TWIN) {
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
/* Configure the second scsi bus */
ahc->sc_link_b = ahc->sc_link;
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
ahc->sc_link_b.adapter_targ = ahc->our_id_b;
ahc->sc_link_b.adapter_bus = 1;
Specify the controller bus in the scsi_link structure to allow hardwired buses on multi-bus controllers. Currently only affects the 274xT controllers. Reviewed by: Peter Dufault(dufault@hda.com), Rod Grimes(rgrimes@FreeBSD.org)
1995-07-17 23:35:16 +00:00
ahc->sc_link_b.fordriver = (void *)SELBUSB;
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
scbus = scsi_alloc_bus();
if(!scbus)
return 0;
scbus->adapter_link = &ahc->sc_link_b;
if(ahc->type & AHC_WIDE)
scbus->maxtarg = 15;
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(bootverbose)
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
printf("ahc%d: Probing Channel B\n", ahc->unit);
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
scsi_attachdevs(scbus);
scbus = NULL; /* Upper-level SCSI code owns this now */
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
}
Remove hard coded assumption that SCSI busses have 7 targets. Change some leading spaces to tabs. This change forces the controller drivers to allocate a scsibus_data struct via a call to scsi_alloc_bus(), fill in the adapter_link field, and optionally modify any other fields of the struct. Scsi_alloc_bus() initializes all fields to the default, so the changes in most drivers are very minimal. For drivers that support Wide controllers, the maxtarg field will have to be updated to allow probing of all targets (for an example, look at the aic7xxx driver). Scsi_attachdevs() now takes a scsibus_data* as its argument instead of an sc_link*. This allows us to expand the role of the scsibus_data struct for other bus level configuration setings (max number of transactions, current transaction opennings, etc for better tagged queuing support). Reviewed by: Rodney Grimes <rgrimes>, Peter Dufault <dufault>, Julian Elischer <julian>
1995-08-23 23:03:17 +00:00
return 1;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
/* Send an SCB down to the card via PIO maintaining the SCB pointer */
static inline void
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_send_scb(ahc, scb)
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
struct ahc_data *ahc;
struct scb *scb;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
{
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
u_long iobase = ahc->baseport;
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
u_char cur_scb;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
cur_scb = inb(SCBPTR + iobase);
outb(SCBPTR + iobase, scb->position);
outb(SCBCNT + iobase, SCBAUTO);
if( ahc->type == AHC_284 )
/* Can only do 8bit PIO */
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
outsb(SCBARRAY+iobase, scb, SCB_PIO_TRANSFER_SIZE);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
else
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
outsl(SCBARRAY+iobase, scb,
(SCB_PIO_TRANSFER_SIZE + 3) / 4);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
outb(SCBCNT + iobase, 0);
outb(SCBPTR + iobase, cur_scb);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
static inline void
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_fetch_scb(ahc, scb, iosize)
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
struct ahc_data *ahc;
struct scb *scb;
int iosize;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
{
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
u_long iobase = ahc->baseport;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
outb(SCBCNT + iobase, 0x80); /* SCBAUTO */
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
/* Can only do 8bit PIO for reads */
insb(SCBARRAY+iobase, scb, iosize);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
outb(SCBCNT + iobase, 0);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
/*
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
* Add this SCB to the head of the "waiting for selection" list.
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
*/
static
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
void ahc_add_waiting_scb (iobase, scb)
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
u_long iobase;
struct scb *scb;
{
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
u_char next;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
u_char curscb;
curscb = inb(SCBPTR + iobase);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
next = inb(WAITING_SCBH + iobase);
outb(SCBPTR+iobase, scb->position);
outb(SCB_NEXT_WAITING+iobase, next);
outb(WAITING_SCBH + iobase, scb->position);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
outb(SCBPTR + iobase, curscb);
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
* Catch an interrupt from the adapter
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
*/
Make PCI interrupt handlers return void like everybody else does. Reviewed by: davidg
1996-01-23 21:48:28 +00:00
void
ahc_intr(arg)
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
void *arg;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
int intstat;
u_char status;
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
u_long iobase;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
struct scb *scb = NULL;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
struct scsi_xfer *xs = NULL;
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
struct ahc_data *ahc = (struct ahc_data *)arg;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
iobase = ahc->baseport;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
intstat = inb(INTSTAT + iobase);
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
/*
* Is this interrupt for me? or for
* someone who is sharing my interrupt
*/
if (!(intstat & INT_PEND))
Make PCI interrupt handlers return void like everybody else does. Reviewed by: davidg
1996-01-23 21:48:28 +00:00
return;
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
if (intstat & BRKADRINT) {
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/* We upset the sequencer :-( */
/* Lookup the error message */
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
int i, error = inb(ERROR + iobase);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
int num_errors = sizeof(hard_error)/sizeof(hard_error[0]);
for(i = 0; error != 1 && i < num_errors; i++)
error >>= 1;
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
panic("ahc%d: brkadrint, %s at seqaddr = 0x%x\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, hard_error[i].errmesg,
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
(inb(SEQADDR1 + iobase) << 8) |
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
inb(SEQADDR0 + iobase));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
if (intstat & SEQINT) {
u_short targ_mask;
u_char target = (inb(SCSIID + iobase) >> 4) & 0x0f;
u_char scratch_offset = target;
char channel =
inb(SBLKCTL + iobase) & SELBUSB ? 'B': 'A';
if (channel == 'B')
scratch_offset += 8;
targ_mask = (0x01 << scratch_offset);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
switch (intstat & SEQINT_MASK) {
case BAD_PHASE:
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
panic("ahc%d:%c:%d: unknown scsi bus phase. "
"Attempting to continue\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, channel, target);
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
break;
case SEND_REJECT:
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
{
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
u_char rejbyte = inb(REJBYTE + iobase);
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
if(( rejbyte & 0xf0) == 0x20) {
/* Tagged Message */
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
printf("\nahc%d:%c:%d: Tagged message "
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
"received without identify. "
"Disabling tagged commands "
"for this target.\n",
ahc->unit, channel, target);
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
ahc->tagenable &= ~targ_mask;
}
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
else
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
printf("ahc%d:%c:%d: Warning - "
"unknown message recieved from "
"target (0x%x). Rejecting\n",
ahc->unit, channel, target,
rejbyte);
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
break;
}
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
case NO_IDENT:
panic("ahc%d:%c:%d: Target did not send an IDENTIFY "
"message. SAVED_TCL == 0x%x\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, channel, target,
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
inb(SAVED_TCL + iobase));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
break;
case NO_MATCH:
{
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
printf("ahc%d:%c:%d: no active SCB for "
"reconnecting target - "
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
"issuing ABORT\n", ahc->unit, channel,
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
target);
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
printf("SAVED_TCL == 0x%x\n",
inb(SAVED_TCL + iobase));
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
ahc_unbusy_target(target, channel, iobase);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
outb(SCBARRAY + iobase, 0);
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
outb(CLRSINT1 + iobase, CLRSELTIMEO);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
RESTART_SEQUENCER(ahc);
break;
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
case SDTR_MSG:
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
{
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
short period;
u_char offset, rate;
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
u_char targ_scratch;
u_char maxoffset;
/*
* Help the sequencer to translate the
* negotiated transfer rate. Transfer is
* 1/4 the period in ns as is returned by
* the sync negotiation message. So, we must
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
* multiply by four
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
period = inb(ARG_1 + iobase) << 2;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
offset = inb(ACCUM + iobase);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
targ_scratch = inb(TARG_SCRATCH + iobase
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
+ scratch_offset);
if(targ_scratch & WIDEXFER)
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
maxoffset = 0x08;
else
maxoffset = 0x0f;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_scsirate(ahc, &rate, period,
MIN(offset,maxoffset),
target);
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
/* Preserve the WideXfer flag */
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
targ_scratch = rate | (targ_scratch & WIDEXFER);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(TARG_SCRATCH + iobase + scratch_offset,
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
targ_scratch);
outb(SCSIRATE + iobase, targ_scratch);
if( (targ_scratch & 0x0f) == 0 )
Be more careful with how we do a chip reset. Clean up some comments.
1995-04-15 21:37:32 +00:00
{
/*
* The requested rate was so low
* that asyncronous transfers are
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* faster (not to mention the
Be more careful with how we do a chip reset. Clean up some comments.
1995-04-15 21:37:32 +00:00
* controller won't support them),
* so we issue a message reject to
* ensure we go to asyncronous
* transfers.
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(RETURN_1 + iobase, SEND_REJ);
Be more careful with how we do a chip reset. Clean up some comments.
1995-04-15 21:37:32 +00:00
}
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
/* See if we initiated Sync Negotiation */
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
else if(ahc->sdtrpending & targ_mask)
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
{
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
/*
* Don't send an SDTR back to
* the target
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(RETURN_1 + iobase, 0);
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
}
else{
/*
* Send our own SDTR in reply
*/
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
#ifdef AHC_DEBUG
if(ahc_debug & AHC_SHOWMISC)
printf("Sending SDTR!!\n");
#endif
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(RETURN_1 + iobase, SEND_SDTR);
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
* Negate the flags
*/
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
ahc->needsdtr &= ~targ_mask;
ahc->sdtrpending &= ~targ_mask;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
break;
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
case WDTR_MSG:
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
{
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
u_char scratch, bus_width;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
bus_width = inb(ARG_1 + iobase);
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scratch = inb(TARG_SCRATCH + iobase
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
+ scratch_offset);
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
if(ahc->wdtrpending & targ_mask)
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
{
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
* Don't send a WDTR back to the
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
* target, since we asked first.
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(RETURN_1 + iobase, 0);
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
switch(bus_width)
{
case BUS_8_BIT:
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
scratch &= 0x7f;
break;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
case BUS_16_BIT:
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(bootverbose)
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
printf("ahc%d: target "
"%d using 16Bit "
"transfers\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit,
target);
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
scratch |= 0x80;
break;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
case BUS_32_BIT:
/*
* How can we do 32bit
* transfers on a 16bit
* bus?
*/
outb(RETURN_1 + iobase,
SEND_REJ);
printf("ahc%d: target "
"%d requested 32Bit "
"transfers. "
"Rejecting...\n",
ahc->unit, target);
break;
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
default:
break;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
}
}
else {
/*
* Send our own WDTR in reply
*/
switch(bus_width)
{
case BUS_8_BIT:
scratch &= 0x7f;
break;
case BUS_32_BIT:
/* Negotiate 16_BITS */
bus_width = BUS_16_BIT;
case BUS_16_BIT:
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(bootverbose)
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
printf("ahc%d: target "
"%d using 16Bit "
"transfers\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit,
target);
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
scratch |= 0x80;
break;
default:
break;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(RETURN_1 + iobase,
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
bus_width | SEND_WDTR);
}
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
ahc->needwdtr &= ~targ_mask;
ahc->wdtrpending &= ~targ_mask;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(TARG_SCRATCH + iobase + scratch_offset,
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
scratch);
outb(SCSIRATE + iobase, scratch);
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
break;
}
The sequencer sets Intstat to REJECT_MSG, not MSG_REJECT. Submitted by: "Daniel M. Eischen" <deischen@iworks.InterWorks.org>
1996-01-05 16:13:44 +00:00
case REJECT_MSG:
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
{
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
* What we care about here is if we had an
* outstanding SDTR or WDTR message for this
* target. If we did, this is a signal that
* the target is refusing negotiation.
*/
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
u_char targ_scratch;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
targ_scratch = inb(TARG_SCRATCH + iobase
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
+ scratch_offset);
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
if(ahc->wdtrpending & targ_mask){
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
/* note 8bit xfers and clear flag */
targ_scratch &= 0x7f;
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
ahc->needwdtr &= ~targ_mask;
ahc->wdtrpending &= ~targ_mask;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
printf("ahc%d:%c:%d: refuses "
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
"WIDE negotiation. Using "
"8bit transfers\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, channel, target);
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
}
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
else if(ahc->sdtrpending & targ_mask){
Set the current transfer parameters directly after a mesg reject for either Wide or Sync negotiation attempts as we do in all other phases of negotiation.
1995-01-27 17:37:05 +00:00
/* note asynch xfers and clear flag */
targ_scratch &= 0xf0;
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
ahc->needsdtr &= ~targ_mask;
ahc->sdtrpending &= ~targ_mask;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
printf("ahc%d:%c:%d: refuses "
Set the current transfer parameters directly after a mesg reject for either Wide or Sync negotiation attempts as we do in all other phases of negotiation.
1995-01-27 17:37:05 +00:00
"syncronous negotiation. Using "
"asyncronous transfers\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, channel, target);
Set the current transfer parameters directly after a mesg reject for either Wide or Sync negotiation attempts as we do in all other phases of negotiation.
1995-01-27 17:37:05 +00:00
}
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
else {
Set the current transfer parameters directly after a mesg reject for either Wide or Sync negotiation attempts as we do in all other phases of negotiation.
1995-01-27 17:37:05 +00:00
/*
* Otherwise, we ignore it.
*/
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
#ifdef AHC_DEBUG
if(ahc_debug & AHC_SHOWMISC)
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
printf("ahc%d:%c:%d: Message
reject -- ignored\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, channel,
target);
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
#endif
Set the current transfer parameters directly after a mesg reject for either Wide or Sync negotiation attempts as we do in all other phases of negotiation.
1995-01-27 17:37:05 +00:00
break;
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(TARG_SCRATCH + iobase + scratch_offset,
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
targ_scratch);
Set the current transfer parameters directly after a mesg reject for either Wide or Sync negotiation attempts as we do in all other phases of negotiation.
1995-01-27 17:37:05 +00:00
outb(SCSIRATE + iobase, targ_scratch);
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
break;
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
case BAD_STATUS:
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
int scb_index;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/* The sequencer will notify us when a command
* has an error that would be of interest to
* the kernel. This allows us to leave the sequencer
* running in the common case of command completes
* without error.
*/
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
scb_index = inb(SCBPTR + iobase);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
scb = ahc->scbarray[scb_index];
Set the return value (outb'd to the sequencer) for non zero status interrupts to zero to start with. This is the value we return in most cases. The sense code then outbs the apropriate value to cause sense retrieval. The return value was uninitialized before this change (something that was an okay thing with the old sequencer code, but not with the semantics of the new). This problem caused us to always retrieve sense upon recieving a non zero status byte. This is exactly what was happening for the "target busy" status returned by exabyte tape drives when they rewind or power on with a tape installed. The request sense proved fatal. This should fix the tape problems.
1995-05-01 09:49:45 +00:00
/*
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* Set the default return value to 0 (don't
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
* send sense). The sense code will change
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* this if needed and this reduces code
Set the return value (outb'd to the sequencer) for non zero status interrupts to zero to start with. This is the value we return in most cases. The sense code then outbs the apropriate value to cause sense retrieval. The return value was uninitialized before this change (something that was an okay thing with the old sequencer code, but not with the semantics of the new). This problem caused us to always retrieve sense upon recieving a non zero status byte. This is exactly what was happening for the "target busy" status returned by exabyte tape drives when they rewind or power on with a tape installed. The request sense proved fatal. This should fix the tape problems.
1995-05-01 09:49:45 +00:00
* duplication.
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(RETURN_1 + iobase, 0);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
if (!scb || !(scb->flags & SCB_ACTIVE)) {
Make PCI interrupt handlers return void like everybody else does. Reviewed by: davidg
1996-01-23 21:48:28 +00:00
printf("ahc%d:%c:%d: ahc_intr - referenced scb "
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
"not valid during seqint 0x%x scb(%d)\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, channel, target, intstat,
scb_index);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
goto clear;
Set the return value (outb'd to the sequencer) for non zero status interrupts to zero to start with. This is the value we return in most cases. The sense code then outbs the apropriate value to cause sense retrieval. The return value was uninitialized before this change (something that was an okay thing with the old sequencer code, but not with the semantics of the new). This problem caused us to always retrieve sense upon recieving a non zero status byte. This is exactly what was happening for the "target busy" status returned by exabyte tape drives when they rewind or power on with a tape installed. The request sense proved fatal. This should fix the tape problems.
1995-05-01 09:49:45 +00:00
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Set the return value (outb'd to the sequencer) for non zero status interrupts to zero to start with. This is the value we return in most cases. The sense code then outbs the apropriate value to cause sense retrieval. The return value was uninitialized before this change (something that was an okay thing with the old sequencer code, but not with the semantics of the new). This problem caused us to always retrieve sense upon recieving a non zero status byte. This is exactly what was happening for the "target busy" status returned by exabyte tape drives when they rewind or power on with a tape installed. The request sense proved fatal. This should fix the tape problems.
1995-05-01 09:49:45 +00:00
xs = scb->xs;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_fetch_scb(ahc, scb, SCB_PIO_TRANSFER_SIZE);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
#ifdef AHC_DEBUG
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
if((ahc_debug & AHC_SHOWSCBS)
&& xs->sc_link->target == DEBUGTARG)
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
ahc_print_scb(scb);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
#endif
Set the return value (outb'd to the sequencer) for non zero status interrupts to zero to start with. This is the value we return in most cases. The sense code then outbs the apropriate value to cause sense retrieval. The return value was uninitialized before this change (something that was an okay thing with the old sequencer code, but not with the semantics of the new). This problem caused us to always retrieve sense upon recieving a non zero status byte. This is exactly what was happening for the "target busy" status returned by exabyte tape drives when they rewind or power on with a tape installed. The request sense proved fatal. This should fix the tape problems.
1995-05-01 09:49:45 +00:00
xs->status = scb->target_status;
switch(scb->target_status){
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
case SCSI_OK:
printf("ahc%d: Interrupted for staus of"
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
" 0???\n", ahc->unit);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
break;
case SCSI_CHECK:
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
#ifdef AHC_DEBUG
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
if(ahc_debug & AHC_SHOWSENSE)
{
sc_print_addr(xs->sc_link);
printf("requests Check Status\n");
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
#endif
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
if((xs->error == XS_NOERROR) &&
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
!(scb->flags & SCB_SENSE)) {
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
u_char control = scb->control;
u_short active;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
struct ahc_dma_seg *sg = scb->ahc_dma;
struct scsi_sense *sc = &(scb->sense_cmd);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
u_char tcl = scb->target_channel_lun;
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
#ifdef AHC_DEBUG
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
if(ahc_debug & AHC_SHOWSENSE)
{
sc_print_addr(xs->sc_link);
printf("Sending Sense\n");
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
#endif
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
bzero(scb, SCB_PIO_TRANSFER_SIZE);
scb->control |= control & DISCENB;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
scb->flags |= SCB_SENSE;
sc->op_code = REQUEST_SENSE;
sc->byte2 = xs->sc_link->lun << 5;
sc->length = sizeof(struct scsi_sense_data);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
sc->control = 0;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
sg->addr = KVTOPHYS(&xs->sense);
sg->len = sizeof(struct scsi_sense_data);
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
scb->target_channel_lun = tcl;
scb->SG_segment_count = 1;
scb->SG_list_pointer = KVTOPHYS(sg);
scb->cmdpointer = KVTOPHYS(sc);
scb->cmdlen = sizeof(*sc);
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
scb->data = sg->addr;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scb->datalen = sg->len;
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_send_scb(ahc, scb);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
outb(SCB_NEXT_WAITING+iobase,
SCB_LIST_NULL);
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
/*
* Ensure that the target is "BUSY"
* so we don't get overlapping
* commands if we happen to be doing
* tagged I/O.
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_busy_target(target,channel,iobase);
/*
* Make us the next command to run
*/
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_add_waiting_scb(iobase, scb);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(RETURN_1 + iobase, SEND_SENSE);
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
break;
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
}
/*
* Clear the SCB_SENSE Flag and have
* the sequencer do a normal command
* complete with either a "DRIVER_STUFFUP"
* error or whatever other error condition
* we already had.
*/
scb->flags &= ~SCB_SENSE;
if(xs->error == XS_NOERROR)
xs->error = XS_DRIVER_STUFFUP;
break;
case SCSI_BUSY:
xs->error = XS_BUSY;
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
sc_print_addr(xs->sc_link);
printf("Target Busy\n");
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
break;
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
case SCSI_QUEUE_FULL:
/*
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
* The upper level SCSI code will eventually
* handle this properly.
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
*/
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
sc_print_addr(xs->sc_link);
printf("Queue Full\n");
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
xs->error = XS_BUSY;
break;
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
default:
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
sc_print_addr(xs->sc_link);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
printf("unexpected targ_status: %x\n",
scb->target_status);
xs->error = XS_DRIVER_STUFFUP;
break;
}
break;
}
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
case RESIDUAL:
{
int scb_index;
scb_index = inb(SCBPTR + iobase);
scb = ahc->scbarray[scb_index];
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
xs = scb->xs;
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
/*
* Don't clobber valid resid info with
* a resid coming from a check sense
* operation.
*/
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
if(!(scb->flags & SCB_SENSE)) {
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
int resid_sgs;
/*
* Remainder of the SG where the transfer
* stopped.
*/
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
xs->resid =
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
(inb(iobase+SCB_RESID_DCNT2)<<16) |
(inb(iobase+SCB_RESID_DCNT1)<<8) |
inb(iobase+SCB_RESID_DCNT0);
/*
* Add up the contents of all residual
* SG segments that are after the SG where
* the transfer stopped.
*/
resid_sgs = inb(SCB_RESID_SGCNT + iobase) - 1;
while(resid_sgs > 0) {
int sg;
sg = scb->SG_segment_count - resid_sgs;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
xs->resid += scb->ahc_dma[sg].len;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
resid_sgs--;
}
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
xs->flags |= SCSI_RESID_VALID;
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
#ifdef AHC_DEBUG
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
if(ahc_debug & AHC_SHOWMISC) {
sc_print_addr(xs->sc_link);
-Wall fixes.
1995-11-06 05:21:13 +00:00
printf("Handled Residual of %ld bytes\n"
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
,xs->resid);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
}
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
#endif
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
}
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
break;
}
Log ABORT_TAG messages to the console. When attempting to abort a command, don't assume that just because the sequecer happens to have SCBPTR pointing at the scb we want that it is an active command.
1995-04-01 19:53:04 +00:00
case ABORT_TAG:
{
int scb_index;
scb_index = inb(SCBPTR + iobase);
scb = ahc->scbarray[scb_index];
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
xs = scb->xs;
Log ABORT_TAG messages to the console. When attempting to abort a command, don't assume that just because the sequecer happens to have SCBPTR pointing at the scb we want that it is an active command.
1995-04-01 19:53:04 +00:00
/*
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* We didn't recieve a valid tag back from
Log ABORT_TAG messages to the console. When attempting to abort a command, don't assume that just because the sequecer happens to have SCBPTR pointing at the scb we want that it is an active command.
1995-04-01 19:53:04 +00:00
* the target on a reconnect.
*/
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
sc_print_addr(xs->sc_link);
printf("invalid tag recieved -- sending ABORT_TAG\n");
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
xs->error = XS_DRIVER_STUFFUP;
Log ABORT_TAG messages to the console. When attempting to abort a command, don't assume that just because the sequecer happens to have SCBPTR pointing at the scb we want that it is an active command.
1995-04-01 19:53:04 +00:00
untimeout(ahc_timeout, (caddr_t)scb);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_done(ahc, scb);
Log ABORT_TAG messages to the console. When attempting to abort a command, don't assume that just because the sequecer happens to have SCBPTR pointing at the scb we want that it is an active command.
1995-04-01 19:53:04 +00:00
break;
}
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
case AWAITING_MSG:
{
int scb_index;
scb_index = inb(SCBPTR + iobase);
scb = ahc->scbarray[scb_index];
/*
* This SCB had a zero length command, informing
* the sequencer that we wanted to send a special
* message to this target. We only do this for
* BUS_DEVICE_RESET messages currently.
*/
if(scb->flags & SCB_DEVICE_RESET)
{
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(MSG0 + iobase,
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
MSG_BUS_DEVICE_RESET);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(MSG_LEN + iobase, 1);
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
}
else
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
panic("ahc_intr: AWAITING_MSG for an SCB that "
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
"does not have a waiting message");
break;
}
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
case IMMEDDONE:
{
/*
* Take care of device reset messages
*/
u_char scbindex = inb(SCBPTR + iobase);
scb = ahc->scbarray[scbindex];
if(scb->flags & SCB_DEVICE_RESET) {
u_char targ_scratch;
int found;
/*
* Go back to async/narrow transfers and
* renegotiate.
*/
ahc_unbusy_target(target, channel, iobase);
ahc->needsdtr |= ahc->needsdtr_orig & targ_mask;
ahc->needwdtr |= ahc->needwdtr_orig & targ_mask;
ahc->sdtrpending &= ~targ_mask;
ahc->wdtrpending &= ~targ_mask;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
targ_scratch = inb(TARG_SCRATCH + iobase
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
+ scratch_offset);
targ_scratch &= SXFR;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(TARG_SCRATCH + iobase + scratch_offset,
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
targ_scratch);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
found = ahc_reset_device(ahc, target,
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
channel, SCB_LIST_NULL,
XS_NOERROR);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
sc_print_addr(scb->xs->sc_link);
printf("Bus Device Reset delivered. "
"%d SCBs aborted\n", found);
ahc->in_timeout = FALSE;
ahc_run_done_queue(ahc);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
}
else
Make PCI interrupt handlers return void like everybody else does. Reviewed by: davidg
1996-01-23 21:48:28 +00:00
panic("ahc_intr: Immediate complete for "
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
"unknown operation.");
break;
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
#if NOT_YET
/* XXX Fill these in later */
case MESG_BUFFER_BUSY:
break;
case MSGIN_PHASEMIS:
break;
#endif
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
default:
printf("ahc: seqint, "
"intstat == 0x%x, scsisigi = 0x%x\n",
intstat, inb(SCSISIGI + iobase));
break;
}
clear:
/*
* Clear the upper byte that holds SEQINT status
* codes and clear the SEQINT bit.
*/
outb(CLRINT + iobase, CLRSEQINT);
/*
* The sequencer is paused immediately on
* a SEQINT, so we should restart it when
* we leave this section.
*/
UNPAUSE_SEQUENCER(ahc);
}
if (intstat & SCSIINT) {
int scb_index = inb(SCBPTR + iobase);
status = inb(SSTAT1 + iobase);
scb = ahc->scbarray[scb_index];
xs = scb->xs;
if (!scb || !(scb->flags & SCB_ACTIVE)) {
printf("ahc%d: ahc_intr - referenced scb not "
"valid during scsiint 0x%x scb(%d)\n",
ahc->unit, status, scb_index);
outb(CLRSINT1 + iobase, status);
UNPAUSE_SEQUENCER(ahc);
outb(CLRINT + iobase, CLRSCSIINT);
scb = NULL;
goto cmdcomplete;
}
if (status & SCSIRSTI) {
char channel;
PAUSE_SEQUENCER(ahc);
channel = inb(SBLKCTL + iobase);
channel = channel & SELBUSB ? 'B' : 'A';
printf("ahc%d: Someone reset channel %c\n",
channel);
ahc_reset_channel(ahc,
channel,
SCB_LIST_NULL,
XS_BUSY,
/* Initiate Reset */FALSE);
}
else if (status & SCSIPERR) {
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
/*
* Determine the bus phase and
* queue an appropriate message
*/
char *phase;
u_char mesg_out = MSG_NOP;
u_char lastphase = inb(LASTPHASE + iobase);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
sc_print_addr(xs->sc_link);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
switch(lastphase) {
case P_DATAOUT:
phase = "Data-Out";
break;
case P_DATAIN:
phase = "Data-In";
mesg_out = MSG_INITIATOR_DET_ERROR;
break;
case P_COMMAND:
phase = "Command";
break;
case P_MESGOUT:
phase = "Message-Out";
break;
case P_STATUS:
phase = "Status";
mesg_out = MSG_INITIATOR_DET_ERROR;
break;
case P_MESGIN:
phase = "Message-In";
mesg_out = MSG_MSG_PARITY_ERROR;
break;
default:
phase = "unknown";
break;
}
printf("parity error during %s phase.\n", phase);
/*
* We've set the hardware to assert ATN if we
* get a parity error on "in" phases, so all we
* need to do is stuff the message buffer with
* the appropriate message. In phases have set
* mesg_out to something other than MSG_NOP.
*/
if(mesg_out != MSG_NOP) {
outb(MSG0 + iobase, mesg_out);
outb(MSG_LEN + iobase, 1);
}
else
/*
* Should we allow the target to make
* this decision for us?
*/
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
xs->error = XS_DRIVER_STUFFUP;
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
}
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
else if (status & SELTO) {
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
u_char waiting;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
u_char flags;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
xs->error = XS_SELTIMEOUT;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
* Clear any pending messages for the timed out
* target, and mark the target as free
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
flags = inb(FLAGS + iobase);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(MSG_LEN + iobase, 0);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
ahc_unbusy_target(xs->sc_link->target,
((long)xs->sc_link->fordriver & SELBUSB)
? 'B' : 'A',
iobase);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
outb(SCBARRAY + iobase, 0);
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(CLRSINT1 + iobase, CLRSELTIMEO);
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(CLRINT + iobase, CLRSCSIINT);
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
/* Shift the waiting for selection queue forward */
waiting = inb(WAITING_SCBH + iobase);
outb(SCBPTR + iobase, waiting);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
waiting = inb(SCB_NEXT_WAITING + iobase);
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
outb(WAITING_SCBH + iobase, waiting);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
RESTART_SEQUENCER(ahc);
}
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
else if (!(status & BUSFREE)) {
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
sc_print_addr(xs->sc_link);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
printf("Unknown SCSIINT. Status = 0x%x\n", status);
outb(CLRSINT1 + iobase, status);
UNPAUSE_SEQUENCER(ahc);
outb(CLRINT + iobase, CLRSCSIINT);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
scb = NULL;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
if(scb != NULL) {
/* We want to process the command */
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
untimeout(ahc_timeout, (caddr_t)scb);
ahc_done(ahc, scb);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
cmdcomplete:
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
if (intstat & CMDCMPLT) {
int scb_index;
do {
scb_index = inb(QOUTFIFO + iobase);
scb = ahc->scbarray[scb_index];
if (!scb || !(scb->flags & SCB_ACTIVE)) {
printf("ahc%d: WARNING "
"no command for scb %d (cmdcmplt)\n"
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
"QOUTCNT == %d\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, scb_index,
inb(QOUTCNT + iobase));
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(CLRINT + iobase, CLRCMDINT);
continue;
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(CLRINT + iobase, CLRCMDINT);
untimeout(ahc_timeout, (caddr_t)scb);
ahc_done(ahc, scb);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
} while (inb(QOUTCNT + iobase) & ahc->qcntmask);
}
Fixed the type of ahcintr(). The type of an ISA interrupt handler is incompatible with the type of a PCI interrupt handler. A new entry point `ahc_pci_intr()' is used for PCI. ISA and PCI interrupts are penalized equally (:-) by calling a common handler `ahc_intr()'. This should be reorganized. Some strings now name the wrong function...
1995-11-04 14:43:30 +00:00
}
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* We have a scb which has been processed by the
* adaptor, now we look to see how the operation
* went.
*/
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_done(ahc, scb)
struct ahc_data *ahc;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
struct scb *scb;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
struct scsi_xfer *xs = scb->xs;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
SC_DEBUG(xs->sc_link, SDEV_DB2, ("ahc_done\n"));
/*
* Put the results of the operation
* into the xfer and call whoever started it
*/
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
if(scb->flags & SCB_SENSE)
xs->error = XS_SENSE;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
if ((xs->flags & SCSI_ERR_OK) && !(xs->error == XS_SENSE)) {
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/* All went correctly OR errors expected */
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
xs->error = XS_NOERROR;
}
xs->flags |= ITSDONE;
Disable tagged queuing by default. option AHC_TAGENABLE will turn it on. This is temporary until I can get a device flags added.
1995-04-09 06:39:01 +00:00
#ifdef AHC_TAGENABLE
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
if(xs->cmd->opcode == 0x12 && xs->error == XS_NOERROR)
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
{
struct scsi_inquiry_data *inq_data;
u_short mask = 0x01 << (xs->sc_link->target |
(scb->target_channel_lun & 0x08));
/*
* Sneak a look at the results of the SCSI Inquiry
* command and see if we can do Tagged queing. This
* should really be done by the higher level drivers.
*/
inq_data = (struct scsi_inquiry_data *)xs->data;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if((inq_data->flags & SID_CmdQue) && !(ahc->tagenable & mask))
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
{
printf("ahc%d: target %d Tagged Queuing Device\n",
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->unit, xs->sc_link->target);
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
ahc->tagenable |= mask;
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
#ifdef QUEUE_FULL_SUPPORTED
Fix typo in QUEUE_FULL_SUPPORTED section. Submitted by: Satoshi Asami <asami>
1995-11-07 07:01:05 +00:00
xs->sc_link->opennings += 2;
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
#endif
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
}
}
Disable tagged queuing by default. option AHC_TAGENABLE will turn it on. This is temporary until I can get a device flags added.
1995-04-09 06:39:01 +00:00
#endif
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
ahc_free_scb(ahc, scb, xs->flags);
scsi_done(xs);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
/*
* Start the board, ready for normal operation
*/
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
int
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_init(ahc)
struct ahc_data *ahc;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
u_long iobase = ahc->baseport;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
u_char scsi_conf, sblkctl, i;
int max_targ = 15;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
Move aic7xxx register definitions to sys/dev/aic7xxx/aic7xxx_reg.h. Start the revamp of the initialiation process. New routines include ahc_alloc, ahc_free, and ahc_reset. These help divide the work of staring up a board more logically between probe and attach. ahcintr now takes a (void *) and returns int. The pci code uses it directly. Until the PCI code for shared edged triggered interrupts is removed, the eisa code uses a stub (ahc_eisa_intr) that throws away the int returned by ahcintr. Use MHz instead of MB/s for printing out sync rates. Print out "aic7880" instead of "aic7870" for the new aic7880 chips.
1995-11-05 04:50:55 +00:00
* Assume we have a board at this stage and it has been reset.
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
*/
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
/* Determine channel configuration and who we are on the scsi bus. */
switch ( (sblkctl = inb(SBLKCTL + iobase) & 0x0a) ) {
case 0:
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->our_id = (inb(SCSICONF + iobase) & HSCSIID);
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(ahc->type == AHC_394)
printf("Channel %c, SCSI Id=%d, ",
ahc->flags & AHC_CHNLB ? 'B' : 'A',
ahc->our_id);
else
printf("Single Channel, SCSI Id=%d, ", ahc->our_id);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(FLAGS + iobase, SINGLE_BUS);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
break;
case 2:
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->our_id = (inb(SCSICONF + 1 + iobase) & HWSCSIID);
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(ahc->type == AHC_394)
printf("Wide Channel %c, SCSI Id=%d, ",
ahc->flags & AHC_CHNLB ? 'B' : 'A',
ahc->our_id);
else
printf("Wide Channel, SCSI Id=%d, ", ahc->our_id);
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
ahc->type |= AHC_WIDE;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(FLAGS + iobase, WIDE_BUS);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
break;
case 8:
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->our_id = (inb(SCSICONF + iobase) & HSCSIID);
ahc->our_id_b = (inb(SCSICONF + 1 + iobase) & HSCSIID);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
printf("Twin Channel, A SCSI Id=%d, B SCSI Id=%d, ",
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
ahc->our_id, ahc->our_id_b);
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
ahc->type |= AHC_TWIN;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(FLAGS + iobase, TWIN_BUS);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
break;
default:
printf(" Unsupported adapter type. Ignoring\n");
return(-1);
}
/* Determine the number of SCBs */
{
int i;
int j;
outb(SCBPTR + iobase, 0);
outb(SCBARRAY + iobase, 0);
for(i = 1; i < AHC_SCB_MAX; i<<=1) {
outb(SCBPTR + iobase, i);
outb(SCBARRAY + iobase, i);
if(inb(SCBARRAY + iobase) == i){
outb(SCBPTR + iobase, 0);
if(inb(SCBARRAY + iobase) != i)
continue;
}
break;
}
/*
* Now go backward from i so we can detect odd
* SCB configurations (like the aic7850 which
* has 3).
*/
j = i >> 1;
outb(SCBPTR + iobase, 0);
outb(SCBARRAY + iobase, 0);
for(i--;i > j; i--) {
outb(SCBPTR + iobase, i);
outb(SCBARRAY + iobase, i);
if(inb(SCBARRAY + iobase) == i){
outb(SCBPTR + iobase, 0);
if(inb(SCBARRAY + iobase) == i)
continue;
}
break;
}
ahc->maxscbs = i+1;
for(ahc->qcntmask = ahc->maxscbs; i > 0; i >>= 1 )
ahc->qcntmask |= i;
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
printf("%d SCBs\n", ahc->maxscbs);
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
#ifdef AHC_DEBUG
if(ahc_debug & AHC_SHOWMISC) {
struct scb test;
printf("ahc%d: hardware scb %ld bytes; kernel scb; "
"ahc_dma %d bytes\n",
ahc->unit, (u_long)&(test.next) - (u_long)(&test),
sizeof(test),
sizeof(struct ahc_dma_seg));
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
#endif /* AHC_DEBUG */
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
/* Set the SCSI Id, SXFRCTL0, SXFRCTL1, and SIMODE1, for both channels*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if(ahc->type & AHC_TWIN)
Initialize SIMODE1, SCSIID, and SXFRCTL1 in ahc_init instead of at each reconnection/selection with a target.
1995-01-22 00:48:39 +00:00
{
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
Initialize SIMODE1, SCSIID, and SXFRCTL1 in ahc_init instead of at each reconnection/selection with a target.
1995-01-22 00:48:39 +00:00
* The device is gated to channel B after a chip reset,
* so set those values first
*/
outb(SCSIID + iobase, ahc->our_id_b);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
scsi_conf = inb(SCSICONF + 1 + iobase) & (ENSPCHK|STIMESEL);
Initialize SIMODE1, SCSIID, and SXFRCTL1 in ahc_init instead of at each reconnection/selection with a target.
1995-01-22 00:48:39 +00:00
outb(SXFRCTL1 + iobase, scsi_conf|ENSTIMER|ACTNEGEN|STPWEN);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(SIMODE1 + iobase, ENSELTIMO|ENSCSIRST|ENSCSIPERR);
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
if(ahc->type & AHC_ULTRA)
Initialize SXFRCTL0 with DFON|SPIOEN|ULTRAEN.
1995-10-29 05:57:48 +00:00
outb(SXFRCTL0 + iobase, DFON|SPIOEN|ULTRAEN);
else
outb(SXFRCTL0 + iobase, DFON|SPIOEN);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
/* Reset the bus */
outb(SCSISEQ + iobase, SCSIRSTO);
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
DELAY(1000);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
outb(SCSISEQ + iobase, 0);
Initialize SIMODE1, SCSIID, and SXFRCTL1 in ahc_init instead of at each reconnection/selection with a target.
1995-01-22 00:48:39 +00:00
/* Select Channel A */
outb(SBLKCTL + iobase, 0);
}
outb(SCSIID + iobase, ahc->our_id);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
scsi_conf = inb(SCSICONF + iobase) & (ENSPCHK|STIMESEL);
Initialize SIMODE1, SCSIID, and SXFRCTL1 in ahc_init instead of at each reconnection/selection with a target.
1995-01-22 00:48:39 +00:00
outb(SXFRCTL1 + iobase, scsi_conf|ENSTIMER|ACTNEGEN|STPWEN);
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(SIMODE1 + iobase, ENSELTIMO|ENSCSIRST|ENSCSIPERR);
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
if(ahc->type & AHC_ULTRA)
Initialize SXFRCTL0 with DFON|SPIOEN|ULTRAEN.
1995-10-29 05:57:48 +00:00
outb(SXFRCTL0 + iobase, DFON|SPIOEN|ULTRAEN);
else
outb(SXFRCTL0 + iobase, DFON|SPIOEN);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
/* Reset the bus */
outb(SCSISEQ + iobase, SCSIRSTO);
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
DELAY(1000);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
outb(SCSISEQ + iobase, 0);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
* Look at the information that board initialization or
* the board bios has left us. In the lower four bits of each
* target's scratch space any value other than 0 indicates
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* that we should initiate syncronous transfers. If it's zero,
* the user or the BIOS has decided to disable syncronous
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
* negotiation to that target so we don't activate the needsdtr
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
* flag.
*/
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
ahc->needsdtr_orig = 0;
ahc->needwdtr_orig = 0;
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
/* Grab the disconnection disable table and invert it for our needs */
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if(ahc->flags & AHC_USEDEFAULTS) {
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
printf("ahc%d: Host Adapter Bios disabled. Using default SCSI "
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
"device parameters\n", ahc->unit);
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
ahc->discenable = 0xff;
}
else
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc->discenable = ~((inb(DISC_DSB + iobase + 1) << 8)
| inb(DISC_DSB + iobase));
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if(!(ahc->type & (AHC_WIDE|AHC_TWIN)))
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
max_targ = 7;
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
for(i = 0; i <= max_targ; i++){
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
u_char target_settings;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if (ahc->flags & AHC_USEDEFAULTS) {
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
target_settings = 0; /* 10MHz */
ahc->needsdtr_orig |= (0x01 << i);
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
ahc->needwdtr_orig |= (0x01 << i);
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
}
else {
/* Take the settings leftover in scratch RAM. */
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
target_settings = inb(TARG_SCRATCH + i + iobase);
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
if(target_settings & 0x0f){
ahc->needsdtr_orig |= (0x01 << i);
/*Default to a asyncronous transfers(0 offset)*/
target_settings &= 0xf0;
}
if(target_settings & 0x80){
ahc->needwdtr_orig |= (0x01 << i);
/*
* We'll set the Wide flag when we
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
* are successful with Wide negotiation.
* Turn it off for now so we aren't
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
* confused.
*/
target_settings &= 0x7f;
}
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(TARG_SCRATCH+i+iobase,target_settings);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
* If we are not a WIDE device, forget WDTR. This
* makes the driver work on some cards that don't
* leave these fields cleared when the BIOS is not
* installed.
*/
if(!(ahc->type & AHC_WIDE))
ahc->needwdtr_orig = 0;
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
ahc->needsdtr = ahc->needsdtr_orig;
Add $Ids. Use DMA for downloading SCBs to the sequencer. Implement SDTR, WDTR, and message reject handlers so they don't need to exist in the sequencer. All three of these cases are not on the critical path, so it makes little sense to use up precious sequencer ram for them.
1995-01-16 16:33:47 +00:00
ahc->needwdtr = ahc->needwdtr_orig;
Keep track of when we actually are awaiting a reply to an SDTR or WDTR message instead of relying on the fact that we are scheduled to send them. The old method worked 99.9% of the time, but someone reported some periferals that did MSG_REJECT at odd times (sometimes before we could send an SDTR or WDTR) that we would construe as the response to an SDTR or WDTR message. This also removes a possible race condition where after a bus reset (the result of a command time out not during intial probe time), we might queue two commands both requesting SDTR, WDTR or both.
1995-02-03 17:15:12 +00:00
ahc->sdtrpending = 0;
ahc->wdtrpending = 0;
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
ahc->tagenable = 0;
Increase the DELAY used in ahc_poll. This is only used during intial probe of the device and shouldn't affect normal operation. It seems that some version of the aic7870 are extreamly slow in clearing their interrupt register so the added delay is required to ensure we don't poll a stale interrupt value. Clean up debugging printfs and stuff superflous output into #ifdef AHC_DEBUG. Ensure that sdtrpending and wdtrpending are cleared in all cases after we recieve an SDTR or WDTR message. With this commit, the 294x should work or at least be much closer to working properly.
1995-03-17 23:58:27 +00:00
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
#ifdef AHC_DEBUG
/* How did we do? */
if(ahc_debug & AHC_SHOWMISC)
printf("NEEDSDTR == 0x%x\nNEEDWDTR == 0x%x\n"
"DISCENABLE == 0x%x\n", ahc->needsdtr,
ahc->needwdtr, ahc->discenable);
#endif
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
* Clear the control byte for every SCB so that the sequencer
* doesn't get confused and think that one of them is valid
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
*/
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
for(i = 0; i < ahc->maxscbs; i++) {
outb(SCBPTR + iobase, i);
outb(SCBARRAY + iobase, 0);
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
/*
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
* Set the number of availible SCBs
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(SCBCOUNT + iobase, ahc->maxscbs);
/*
* 2s compliment of SCBCOUNT
*/
i = ahc->maxscbs;
outb(COMP_SCBCOUNT + iobase, -i & 0xff);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
/*
* QCount mask to deal with broken aic7850s that
* sporatically get garbage in the upper bits of
* their QCount registers.
*/
outb(QCNTMASK + iobase, ahc->qcntmask);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
/* We don't have any busy targets right now */
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb( ACTIVE_A + iobase, 0 );
outb( ACTIVE_B + iobase, 0 );
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
/* We don't have any waiting selections */
outb( WAITING_SCBH + iobase, SCB_LIST_NULL );
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
/* Message out buffer starts empty */
outb(MSG_LEN + iobase, 0x00);
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
/*
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
* Load the Sequencer program and Enable the adapter
* in "fast" mode.
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
*/
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(bootverbose)
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
printf("ahc%d: Downloading Sequencer Program...", ahc->unit);
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
ahc_loadseq(iobase);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
Clean up a few nits in the aic7xxx driver: 1) Make the driver "quiet" by sticking most boot messages behind bootverbose conditionals. This means that you won't see the sync and wide negotiation, but you will find out if they fail. 2) Add support to the 93cx6 serial eeprom code to read at an abitrary offset. This is needed so that we can access the second half of the eeprom on 3940 cards where the second channel's config is stored. 3) Add flags argument to ahcprobe(). This is used by the pci probe code to tell the generic driver that an adapter should be treated as a channel B device as well as notify it of the presence of external SCB SRAM. These are needed for some motherboard implementations of the aic7870 and for the 3940 controllers. 4) Print "Channel A"/"Channel B" instead of "Single Channel" for the two busses of the 3940. I received many reports of confusion about how the 3940 was probed since most people belived that only one ahc entry was needed. This will hopefully make it clearer. 5) Walk the SCBs to determine just how many their are if external SCB ram is detected. 6) Hard code that external SCB ram is present for the 3940 since it doesn't use the documented reporting facility for reporting the SRAM. :( 255 commands per channel are supported on the 3940. 7) Read the seeprom starting at addres 32 for the second channel of the 3940 so we get the right info for that channel. 8) Clean up printing of the "Disabling tagged queuing message". 9) Queue timeouts if they occur while we are handling a timeout. The code was totally unprotected in this scenario. Reviewed by: Timeout code reviewed by David Greenman <davidg>
1995-09-05 23:52:03 +00:00
if(bootverbose)
printf("Done\n");
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
outb(SEQCTL + iobase, FASTMODE);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
UNPAUSE_SEQUENCER(ahc);
/*
* Note that we are going and return (to probe)
*/
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
ahc->flags = AHC_INIT;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
return (0);
}
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahcminphys(bp)
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
struct buf *bp;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
* Even though the card can transfer up to 16megs per command
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
* we are limited by the number of segments in the dma segment
* list that we can hold. The worst case is that all pages are
* discontinuous physically, hense the "page per segment" limit
* enforced here.
*/
if (bp->b_bcount > ((AHC_NSEG - 1) * PAGESIZ)) {
bp->b_bcount = ((AHC_NSEG - 1) * PAGESIZ);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
/*
* start a scsi operation given the command and
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* the data address, target, and lun all of which
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
* are stored in the scsi_xfer struct
*/
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
static int32_t
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahc_scsi_cmd(xs)
struct scsi_xfer *xs;
{
Use the new adapter_softc field in the scsi_link structure so that these drivers don't need to maintain an array of configured units. The bt driver still needs to because ISA interrupt handlers take a unit number. :(
1996-01-07 19:24:36 +00:00
struct scb *scb;
struct ahc_dma_seg *sg;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
int seg; /* scatter gather seg being worked on */
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
int thiskv;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
physaddr thisphys, nextphys;
int bytes_this_seg, bytes_this_page, datalen, flags;
Use the new adapter_softc field in the scsi_link structure so that these drivers don't need to maintain an array of configured units. The bt driver still needs to because ISA interrupt handlers take a unit number. :(
1996-01-07 19:24:36 +00:00
struct ahc_data *ahc;
u_short mask;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
int s;
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
Use the new adapter_softc field in the scsi_link structure so that these drivers don't need to maintain an array of configured units. The bt driver still needs to because ISA interrupt handlers take a unit number. :(
1996-01-07 19:24:36 +00:00
ahc = (struct ahc_data *)xs->sc_link->adapter_softc;
mask = (0x01 << (xs->sc_link->target
| ((u_long)xs->sc_link->fordriver & 0x08)));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
SC_DEBUG(xs->sc_link, SDEV_DB2, ("ahc_scsi_cmd\n"));
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
* get an scb to use. If the transfer
* is from a buf (possibly from interrupt time)
* then we can't allow it to sleep
*/
flags = xs->flags;
if (flags & ITSDONE) {
Use the new adapter_softc field in the scsi_link structure so that these drivers don't need to maintain an array of configured units. The bt driver still needs to because ISA interrupt handlers take a unit number. :(
1996-01-07 19:24:36 +00:00
printf("ahc%d: Already done?", ahc->unit);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
xs->flags &= ~ITSDONE;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
if (!(flags & INUSE)) {
Use the new adapter_softc field in the scsi_link structure so that these drivers don't need to maintain an array of configured units. The bt driver still needs to because ISA interrupt handlers take a unit number. :(
1996-01-07 19:24:36 +00:00
printf("ahc%d: Not in use?", ahc->unit);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
xs->flags |= INUSE;
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if (!(scb = ahc_get_scb(ahc, flags))) {
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
xs->error = XS_DRIVER_STUFFUP;
return (TRY_AGAIN_LATER);
}
Fix -Wformat warnings from LINT kernel.
1995-05-11 19:26:53 +00:00
SC_DEBUG(xs->sc_link, SDEV_DB3, ("start scb(%p)\n", scb));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
scb->xs = xs;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
if (flags & SCSI_RESET)
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
scb->flags |= SCB_DEVICE_RESET|SCB_IMMED;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* Put all the arguments for the xfer in the scb
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
*/
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
if(ahc->tagenable & mask)
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scb->control |= TAG_ENB;
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
if(ahc->discenable & mask)
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scb->control |= DISCENB;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
if((ahc->needwdtr & mask) && !(ahc->wdtrpending & mask))
Keep track of when we actually are awaiting a reply to an SDTR or WDTR message instead of relying on the fact that we are scheduled to send them. The old method worked 99.9% of the time, but someone reported some periferals that did MSG_REJECT at odd times (sometimes before we could send an SDTR or WDTR) that we would construe as the response to an SDTR or WDTR message. This also removes a possible race condition where after a bus reset (the result of a command time out not during intial probe time), we might queue two commands both requesting SDTR, WDTR or both.
1995-02-03 17:15:12 +00:00
{
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scb->control |= NEEDWDTR;
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
ahc->wdtrpending |= mask;
Keep track of when we actually are awaiting a reply to an SDTR or WDTR message instead of relying on the fact that we are scheduled to send them. The old method worked 99.9% of the time, but someone reported some periferals that did MSG_REJECT at odd times (sometimes before we could send an SDTR or WDTR) that we would construe as the response to an SDTR or WDTR message. This also removes a possible race condition where after a bus reset (the result of a command time out not during intial probe time), we might queue two commands both requesting SDTR, WDTR or both.
1995-02-03 17:15:12 +00:00
}
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
else if((ahc->needsdtr & mask) && !(ahc->sdtrpending & mask))
Keep track of when we actually are awaiting a reply to an SDTR or WDTR message instead of relying on the fact that we are scheduled to send them. The old method worked 99.9% of the time, but someone reported some periferals that did MSG_REJECT at odd times (sometimes before we could send an SDTR or WDTR) that we would construe as the response to an SDTR or WDTR message. This also removes a possible race condition where after a bus reset (the result of a command time out not during intial probe time), we might queue two commands both requesting SDTR, WDTR or both.
1995-02-03 17:15:12 +00:00
{
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scb->control |= NEEDSDTR;
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
ahc->sdtrpending |= mask;
Keep track of when we actually are awaiting a reply to an SDTR or WDTR message instead of relying on the fact that we are scheduled to send them. The old method worked 99.9% of the time, but someone reported some periferals that did MSG_REJECT at odd times (sometimes before we could send an SDTR or WDTR) that we would construe as the response to an SDTR or WDTR message. This also removes a possible race condition where after a bus reset (the result of a command time out not during intial probe time), we might queue two commands both requesting SDTR, WDTR or both.
1995-02-03 17:15:12 +00:00
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
scb->target_channel_lun = ((xs->sc_link->target << 4) & 0xF0) |
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
((u_long)xs->sc_link->fordriver & 0x08) |
Compile cleanly -Wall. Remove some code that was an attempt to make the 294x probe more stable, but since the guess was wrong, go back to the clearer aproach I used the first time around.
1995-03-07 08:59:28 +00:00
(xs->sc_link->lun & 0x07);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
scb->cmdlen = xs->cmdlen;
scb->cmdpointer = KVTOPHYS(xs->cmd);
Major overhaul of the aic7xxx driver: - catch the interrupt type (EDGE/LEVEL) before chip reset instead of guessing the right type. - Add pause variable to the ahc struct to better handle the different interrupt types and pausing the sequencer. - CLRINTSTAT -> CLRSCSIINT: This is a documented bit in the CLRINT register in newer Adaptec documentation, so use their name for it. - Report valid residual byte counts. - Don't mess with the target scratch areas > id 8 on single, narrow, channel devices. The BIOS does a checksum of this area and can flip out if we zero it out. - Initialize the sequencer FLAGS scratch ram variable in the single channel devices to 0. This was the cause of the annoying warning where we would get a cmdcmplt the first time we did any type of transfer negotiation with no valid scb. It also fixes the problem that looked like the INTSTAT register wasn't clearing fast enough. This only showed up on 294x cards, not motherboard aic7870s. - Add the AHC_AIC7870 type and use it as the superset of aic7870 based controllers. - clear the sync offset section of the targ scratch area so that we default to asyncronous transfers. This was only a problem for wide controllers because there was a scenario where the offset wouldn't get updated before a data(out/in) phase would occur. This required some change in the sequencer code since we were depending on this field to hold the rate to negotiate. - allow sync and wide negotiated commands to be tagged (the sequencer now handles this properly).
1995-03-31 13:54:41 +00:00
xs->resid = 0;
Merge RELENG_2_0_5 into HEAD
1995-06-11 19:33:05 +00:00
xs->status = 0;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
if (xs->datalen) { /* should use S/G only if not zero length */
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
scb->SG_list_pointer = KVTOPHYS(scb->ahc_dma);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
sg = scb->ahc_dma;
seg = 0;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
{
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* Set up the scatter gather block
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
*/
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
SC_DEBUG(xs->sc_link, SDEV_DB4,
Fix -Wformat warnings from LINT kernel.
1995-05-11 19:26:53 +00:00
("%ld @%p:- ", xs->datalen, xs->data));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
datalen = xs->datalen;
thiskv = (int) xs->data;
thisphys = KVTOPHYS(thiskv);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
while ((datalen) && (seg < AHC_NSEG)) {
bytes_this_seg = 0;
/* put in the base address */
sg->addr = thisphys;
Fix -Wformat warnings from LINT kernel.
1995-05-11 19:26:53 +00:00
SC_DEBUGN(xs->sc_link, SDEV_DB4, ("0x%lx",
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
thisphys));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/* do it at least once */
nextphys = thisphys;
while ((datalen) && (thisphys == nextphys)) {
/*
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* This page is contiguous (physically)
* with the the last, just extend the
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
* length
*/
/* how far to the end of the page */
nextphys = (thisphys & (~(PAGESIZ - 1)))
+ PAGESIZ;
bytes_this_page = nextphys - thisphys;
/**** or the data ****/
bytes_this_page = min(bytes_this_page
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
,datalen);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
bytes_this_seg += bytes_this_page;
datalen -= bytes_this_page;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/* get more ready for the next page */
thiskv = (thiskv & (~(PAGESIZ - 1)))
+ PAGESIZ;
if (datalen)
thisphys = KVTOPHYS(thiskv);
}
/*
* next page isn't contiguous, finish the seg
*/
SC_DEBUGN(xs->sc_link, SDEV_DB4,
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
("(0x%x)", bytes_this_seg));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
sg->len = bytes_this_seg;
sg++;
seg++;
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
}
} /*end of iov/kv decision */
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
scb->SG_segment_count = seg;
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
/* Copy the first SG into the data pointer area */
scb->data = scb->ahc_dma->addr;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scb->datalen = scb->ahc_dma->len;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
SC_DEBUGN(xs->sc_link, SDEV_DB4, ("\n"));
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
if (datalen) {
/* there's still data, must have run out of segs! */
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
printf("ahc_scsi_cmd%d: more than %d DMA segs\n",
Use the new adapter_softc field in the scsi_link structure so that these drivers don't need to maintain an array of configured units. The bt driver still needs to because ISA interrupt handlers take a unit number. :(
1996-01-07 19:24:36 +00:00
ahc->unit, AHC_NSEG);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
xs->error = XS_DRIVER_STUFFUP;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_free_scb(ahc, scb, flags);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
return (COMPLETE);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
}
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
else {
/*
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* No data xfer, use non S/G values
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
*/
scb->SG_segment_count = 0;
scb->SG_list_pointer = 0;
First pass cleanup of this driver. This pass does not include the sequencer optimizations I have been working on yet, but does bring in some bug fixes and performance improvments that were easy to regression test: Setup the data fifo threshold and bus off timing correctly for 27/284x cards. Users of these adapters with fast periferals (greater than 5MB/s) will notice a big performance difference. (Sometimes as large as going from 3.7->8.3MB/s). Fix handling of the active target flags. Some of the outbs where missing the base offset in the abort code. The abort code still needs lots of work. Support 3940 controllers, but only with 16 SCBs for now. Eventually I'll add support for all 255, but I need to find a tester for the code first since we have to enable the cards external SRAM to do this. Add Dan Eischen's serial eeprom reading facilities. This allows the 2940 adapters to pull additional information left over from SCSI-Select right out out of the configuration seeprom. If the BIOS is disabled on 274x controllers, reset all target parameters to there defaults since you can't rely on what is stored in scratch ram. Report motherboard controllers as such. Stick the first SG address and count into the SCB data and count areas for all transfers in preparation of a later sequencer optimization. Keep track of which targets can are allowed to have the disconnection priveledge since this will be handled by the kernel driver in the future. If a target issues a message reject in response to a tagged message, disable tagged queuing for that target. Some seagates say they can do tagged queuing, but lie, and its a shame to have to disable tagged queuing on all devices just because you have one that can't cope.
1995-07-04 21:14:45 +00:00
scb->data = 0;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scb->datalen = 0;
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Add tagged queueing support. Right now, we only allow two tagged commands per target. I could have followed the route of the ncr driver and gone to great lengths to get the SCSI subsystem to support more, but I think I'll use the time saved to help Julian and Peter make tagged queuing a better handled generic feature. This also includes some comment and enum clean up and a possible fix for the hanging PCI controllers.
1995-02-22 01:43:25 +00:00
#ifdef AHC_DEBUG
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
if((ahc_debug & AHC_SHOWSCBS) && (xs->sc_link->target == DEBUGTARG))
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahc_print_scb(scb);
#endif
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
if (!(flags & SCSI_NOMASK)) {
s = splbio();
PAUSE_SEQUENCER(ahc);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_send_scb(ahc, scb);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
outb(QINFIFO + ahc->baseport, scb->position);
UNPAUSE_SEQUENCER(ahc);
timeout(ahc_timeout, (caddr_t)scb, (xs->timeout * hz) / 1000);
splx(s);
SC_DEBUG(xs->sc_link, SDEV_DB3, ("cmd_sent\n"));
return (SUCCESSFULLY_QUEUED);
}
/*
* If we can't use interrupts, poll for completion
*/
s = splbio();
PAUSE_SEQUENCER(ahc);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_send_scb(ahc, scb);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
outb(QINFIFO + ahc->baseport, scb->position);
UNPAUSE_SEQUENCER(ahc);
SC_DEBUG(xs->sc_link, SDEV_DB3, ("cmd_wait\n"));
do {
if (ahc_poll(ahc, xs->timeout)) {
if (!(xs->flags & SCSI_SILENT))
printf("cmd fail\n");
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_timeout(scb);
break;
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
}
} while (!(xs->flags & ITSDONE)); /* a non command complete intr */
splx(s);
return (COMPLETE);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
* Return some information to the caller about
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
* the adapter and it's capabilities.
*/
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
static u_int32_t
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
ahc_adapter_info(unit)
int unit;
{
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
return (2); /* 2 outstanding requests at a time per device */
}
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* A scb (and hence an scb entry on the board is put onto the
* free list.
*/
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_free_scb(ahc, scb, flags)
struct ahc_data *ahc;
int flags;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
struct scb *scb;
{
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
unsigned int opri;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
opri = splbio();
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
scb->flags = SCB_FREE;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
SLIST_INSERT_HEAD(&ahc->free_scb, scb, next);
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
#ifdef AHC_DEBUG
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
ahc->activescbs--;
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
#endif
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* If there were none, wake abybody waiting for
* one to come free, starting with queued entries
*/
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
if (!scb->next.sle_next) {
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
wakeup((caddr_t)&ahc->free_scb);
}
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
splx(opri);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* Get a free scb
* If there are none, see if we can allocate a
* new one. Otherwise either return an error or sleep
*/
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static struct scb *
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_get_scb(ahc, flags)
struct ahc_data *ahc;
int flags;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
unsigned opri;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
struct scb *scbp;
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
opri = splbio();
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* If we can and have to, sleep waiting for one to come free
* but only if we can't allocate a new one.
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
*/
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
while (1) {
if( scbp = ahc->free_scb.slh_first ) {
SLIST_REMOVE_HEAD(&ahc->free_scb, next);
}
else if (ahc->numscbs < ahc->maxscbs) {
scbp = (struct scb *) malloc(sizeof(struct scb),
M_TEMP, M_NOWAIT);
if (scbp) {
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
bzero(scbp, sizeof(struct scb));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
scbp->position = ahc->numscbs;
Additions to support the WAITING_SCB list that the sequencer maintains. It is the kernel driver's responsibility to do the list manipulation whenever a selection timeout or a request sense occurs. Print out the interrupt type that the device has been set to. It seems that one of the Asus motherboards botches this and David thought a diagnostic would be nice. Fix a bug in my diagnostic code that David found. Reviewed by: Wcarchive and David Greenman
1995-04-27 17:47:17 +00:00
ahc->numscbs++;
scbp->flags = SCB_ACTIVE;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
/*
* Place in the scbarray
* Never is removed. Position
* in ahc->scbarray is the scbarray
* position on the board we will
* load it into.
*/
ahc->scbarray[scbp->position] = scbp;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
}
else {
printf("ahc%d: Can't malloc SCB\n", ahc->unit);
}
}
else {
if (!(flags & SCSI_NOSLEEP)) {
tsleep((caddr_t)&ahc->free_scb, PRIBIO,
"ahcscb", 0);
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
continue;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
}
}
break;
}
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
if (scbp) {
Fixes to the aic7xxx sequencer code and device driver from Justin Gibbs: 1) If a target initiated a sync negotiation with us and happened to chose a value above 15, the old code inadvertantly truncated it with an "& 0x0f". If the periferal picked something really bad like 0x32, you'd end up with an offset of 2 which would hang the drive since it didn't expect to ever get something so low. We now do a MIN(maxoffset, given_offset). 2) In the case of Wide cards, we were turning on sync transfers after a sucessfull wide negotiation. Now we leave the offset alone in the per target scratch space (which implies asyncronous transfers since we initialize it that way) until a syncronous negotation occurs. 3) We were advertizing a max offset of 15 instead of 8 for wide devices. 4) If the upper level SCSI code sent down a "SCSI_RESET", it would hang the system because we would end up sending a null command to the sequencer. Now we handle SCSI_RESET correctly by having the sequencer interrupt us when it is about to fill the message buffer so that we can fill it in ourselves. The sequencer will also "simulate" a command complete for these "message only" SCBs so that the kernel driver can finish up properly. The cdplay utility will send a "SCSI_REST" to the cdplayer if you use the reset command. 5) The code that handles SCSIINTs was broken in that if more than one type of error was true at once, we'd do outbs without the card being paused. The else clause after the busfree case was also an accident waiting to happen. I've now turned this into an if, else if, else type of thing, since in most cases when we handle one type of error, it should be okay to ignore the rest (ie if we have a SELTO, who cares if there was a parity error on the transaction?), but the section should really be rewritten after 2.0.5. This fix was the least obtrusive way to patch the problem. 6) Only tag either SDTR or WDTR negotiation on an SCB. The real problem is that I don't account for the case when an SCB that is tagged to do a particular type of negotiation completes or SELTOs (selection timeout) without the negotiation taking place, so the accounting of sdtrpending and wdtrpending gets screwed up. In the wide case, if we tag it to do both wdtr and sdtr, it only performs wdtr (since wdtr must occur first and we spread out the negotiation over two commands) so we always have sdtrpending set for that target and we never do a real SDTR. I fill properly fix the accounting after 2.0.5 goes out the door, but this works (as confirmed by Dan) on wide targets. Other stuff that is also included: 1) Don't do a bzero when recycling SCBs. The only thing that must explicitly be set to zero is the scb control byte which is done in ahc_get_scb. We also need to set the SG_list_pointer and SG_list_count to 0 for commands that do not transfer data. 2) Mask the interrupt type printout for the aic7870 case. The bit we were using to determine interrupt type is only valid for the aic7770. Submitted by: Justin Gibbs
1995-05-17 07:06:02 +00:00
scbp->control = 0;
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
scbp->flags = SCB_ACTIVE;
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
#ifdef AHC_DEBUG
ahc->activescbs++;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if((ahc_debug & AHC_SHOWSCBCNT)
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
&& (ahc->activescbs == ahc->maxscbs))
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
printf("ahc%d: Max SCBs active\n", ahc->unit);
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
#endif
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
}
Don't arbitrarily set SCSI_NOSLEEP. It is now handled correctly by the higher level scsi code. Spls should never be conditionalized, so don't do so here. Restructure the get_scb routine so that we can't get into an infinite loop if the ccbs are exhausted and we are are called with SCSI_NOSLEEP set. Other driver maintainer's that based their scb allocation routines on Julian's code should look at these changes and implement them for their driver. The aic7xxx driver inspired these changes because early revs of the aic7770 chips have so few SCBs that you can actually run out. If you have a rev C or aic7770 (as is reported by the driver probe) and had more than 2 drives, you could get into an infinite loop when using up all of the SCBs. Since the driver will only allow two SCBs per device and I only had two devices, I never saw this problem on my Rev C card. Bzero only 19 bytes of the scb instead of 2k (ack!). This was a hold over from when a struct SCB only contained the information downloaded to the board, but we now store kernel driver data in there as well. This greatly lowers the overhead for small transactions (I get ~1MB/sec for dds with a 512 byte block size). Submitted by: John Dyson with the aic7xxx specific optimization by me
1995-04-23 22:04:58 +00:00
splx(opri);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
return (scbp);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void ahc_loadseq(iobase)
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
u_long iobase;
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
{
Change the include for the sequencer program to point back to the right place: include "../../sys/gnu/misc/aic7770/aic7770_seq.h"
1994-11-18 09:14:14 +00:00
static unsigned char seqprog[] = {
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
# include "aic7xxx_seq.h"
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
};
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(SEQCTL + iobase, PERRORDIS|SEQRESET|LOADRAM);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Generic support for the entire aic7x70 line of adaptors. No one else came up with a directory place ment more me, so now we have i386/scsi.
1995-01-13 02:24:31 +00:00
outsb(SEQRAM + iobase, seqprog, sizeof(seqprog));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(SEQCTL + iobase, FASTMODE|SEQRESET);
do {
outb(SEQCTL + iobase, SEQRESET|FASTMODE);
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
} while (inb(SEQADDR0 + iobase) != 0 &&
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
inb(SEQADDR1 + iobase != 0));
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
/*
* Function to poll for command completion when
* interrupts are disabled (crash dumps)
*/
static int
ahc_poll(ahc, wait)
struct ahc_data *ahc;
int wait; /* in msec */
{
u_long iobase = ahc->baseport;
u_long stport = INTSTAT + iobase;
while (--wait) {
DELAY(1000);
if (inb(stport) & INT_PEND)
break;
} if (wait == 0) {
printf("ahc%d: board not responding\n", ahc->unit);
return (EIO);
}
ahc_intr((void *)ahc);
return (0);
}
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_timeout(arg)
void *arg;
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
{
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
struct scb *scb = (struct scb *)arg;
struct ahc_data *ahc;
int s, h, found;
u_char bus_state;
u_long iobase;
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
s = splbio();
h = splhigh();
if (!(scb->flags & SCB_ACTIVE)) {
/* Previous timeout took care of me already */
splx(h);
splx(s);
return;
}
ahc = (struct ahc_data *)scb->xs->sc_link->adapter_softc;
if (ahc->in_timeout) {
/*
* Some other SCB has started a recovery operation
* and is still working on cleaning things up.
*/
if (scb->flags & SCB_TIMEDOUT) {
/*
* This SCB has been here before and is not the
* recovery SCB. Cut our losses and panic. Its
* better to do this than trash a filesystem.
*/
panic("ahc%d: Timed-out command times out "
"again\n", ahc->unit);
}
else if (!(scb->flags & SCB_ABORTED))
{
/*
* This is not the SCB that started this timeout
* processing. Give this scb another lifetime so
* that it can continue once we deal with the
* timeout.
*/
scb->flags |= SCB_TIMEDOUT;
timeout(ahc_timeout, (caddr_t)scb,
(scb->xs->timeout * hz) / 1000);
splx(h);
splx(s);
return;
}
}
ahc->in_timeout = TRUE;
splx(h);
/*
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
* Ensure that the card doesn't do anything
* behind our back.
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
*/
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
PAUSE_SEQUENCER(ahc);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
sc_print_addr(scb->xs->sc_link);
printf("timed out ");
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
/*
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
* Take a snapshot of the bus state and print out
* some information so we can track down driver bugs.
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
*/
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
iobase = ahc->baseport;
bus_state = inb(iobase + LASTPHASE);
switch(bus_state & PHASE_MASK)
{
case P_DATAOUT:
printf("in dataout phase");
break;
case P_DATAIN:
printf("in datain phase");
break;
case P_COMMAND:
printf("in command phase");
break;
case P_MESGOUT:
printf("in message out phase");
break;
case P_STATUS:
printf("in status phase");
break;
case P_MESGIN:
printf("in message in phase");
break;
default:
printf("while idle, LASTPHASE == 0x%x",
bus_state);
/*
* We aren't in a valid phase, so assume we're
* idle.
*/
bus_state = 0;
break;
}
printf(", SCSISIGI == 0x%x\n", inb(iobase + SCSISIGI));
/* Decide our course of action */
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
if(scb->flags & SCB_ABORTED)
{
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
/*
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
* Been down this road before.
* Do a full bus reset.
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
*/
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
char channel = (scb->target_channel_lun & SELBUSB)
? 'B': 'A';
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
found = ahc_reset_channel(ahc, channel, scb->position,
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
XS_TIMEOUT, /*Initiate Reset*/TRUE);
printf("ahc%d: Issued Channel %c Bus Reset #1. "
"%d SCBs aborted\n", ahc->unit, channel, found);
ahc->in_timeout = FALSE;
Log ABORT_TAG messages to the console. When attempting to abort a command, don't assume that just because the sequecer happens to have SCBPTR pointing at the scb we want that it is an active command.
1995-04-01 19:53:04 +00:00
}
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
else {
Remove trailing whitespace.
1995-05-30 08:16:23 +00:00
/*
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
* Send a Bus Device Reset Message:
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
* The target that is holding up the bus may not
* be the same as the one that triggered this timeout
* (different commands have different timeout lengths).
* It is also impossible to get a message to a target
* if we are in a "frozen" data transfer phase. Our
* strategy here is to queue a bus device reset message
* to the timed out target if it is disconnected.
* Otherwise, if we have an active target we stuff the
* message buffer with a bus device reset message and
* assert ATN in the hopes that the target will let go
* of the bus and finally disconnect. If this fails,
* we'll get another timeout 2 seconds later which will
* cause a bus reset.
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
*/
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
u_char active_scb, control;
struct scb *active_scbp;
active_scb = inb(SCBPTR + iobase);
active_scbp = ahc->scbarray[active_scb];
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
control = inb(SCB_CONTROL + iobase);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
/* Test to see if scb is disconnected */
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
outb(SCBPTR + iobase, scb->position);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
if(inb(SCB_CONTROL + iobase) & DISCONNECTED) {
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
scb->flags |= SCB_DEVICE_RESET|SCB_ABORTED;
scb->SG_segment_count = 0;
scb->SG_list_pointer = 0;
scb->data = 0;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
scb->datalen = 0;
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_send_scb(ahc, scb);
ahc_add_waiting_scb(iobase, scb);
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
timeout(ahc_timeout, (caddr_t)scb, (2 * hz));
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
sc_print_addr(scb->xs->sc_link);
printf("BUS DEVICE RESET message queued.\n");
outb(SCBPTR + iobase, active_scb);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
UNPAUSE_SEQUENCER(ahc);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
}
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
/* Is the active SCB really active? */
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
else if((active_scbp->flags & SCB_ACTIVE) && bus_state) {
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(MSG_LEN + iobase, 1);
outb(MSG0 + iobase, MSG_BUS_DEVICE_RESET);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
outb(SCSISIGO + iobase, bus_state|ATNO);
sc_print_addr(active_scbp->xs->sc_link);
printf("asserted ATN - device reset in message buffer\n");
active_scbp->flags |= SCB_DEVICE_RESET|SCB_ABORTED;
if(active_scbp != scb) {
untimeout(ahc_timeout, (caddr_t)active_scbp);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
/* Give scb a new lease on life */
timeout(ahc_timeout, (caddr_t)scb,
(scb->xs->timeout * hz) / 1000);
}
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
timeout(ahc_timeout, (caddr_t)active_scbp, (2 * hz));
outb(SCBPTR + iobase, active_scb);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
UNPAUSE_SEQUENCER(ahc);
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
}
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
else {
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
/* No active target? Try reseting the bus */
char channel = (scb->target_channel_lun & SELBUSB)
? 'B': 'A';
found = ahc_reset_channel(ahc, channel, scb->position,
XS_TIMEOUT, /*Initiate Reset*/TRUE);
printf("ahc%d: Issued Channel %c Bus Reset #2. "
"%d SCBs aborted\n", ahc->unit, channel,
found);
ahc->in_timeout = FALSE;
Handle both channels of Twin channeled devices. Respect the configuration values for syncronous negotiation. The 284x series adaptors can now be supported without the Bios being enabled. If you disable the Bios on the 274x series adaptors, all configuration parameters revert to the default since there is no way to retrieve them.
1994-12-31 19:31:56 +00:00
}
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
}
Properly deal with the Ultra series of adapters. We should now understand the new seeprom format and negotiate up to 20MHz sync if set in SCSI-Select. Reduce the complexity of the timeout code by running it at splhigh(). Fix a bug that caused rescheduled timeouts at 0 clock ticks in the future causing an infinite loop. Obtained from: Timeout bug noticed by David Greenman and wcarchive.
1995-10-26 23:57:18 +00:00
splx(s);
My port of Julian's 1742 driver to work with the Linux aic7770 sequencer code.
1994-11-17 20:22:31 +00:00
}
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
/*
* The device at the given target/channel has been reset. Abort
* all active and queued scbs for that target/channel.
*/
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static int
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_reset_device(ahc, target, channel, timedout_scb, xs_error)
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
struct ahc_data *ahc;
int target;
char channel;
u_char timedout_scb;
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
u_int32_t xs_error;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
{
u_long iobase = ahc->baseport;
struct scb *scbp;
u_char active_scb;
int i = 0;
int found = 0;
/* restore this when we're done */
active_scb = inb(SCBPTR + iobase);
/*
* Search the QINFIFO.
*/
{
int saved_queue[AHC_SCB_MAX];
int queued = inb(QINCNT + iobase);
for (i = 0; i < (queued - found); i++) {
saved_queue[i] = inb(QINFIFO + iobase);
scbp = ahc->scbarray[saved_queue[i]];
if (ahc_match_scb (scbp, target, channel)){
/*
* We found an scb that needs to be aborted.
*/
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
scbp->flags = SCB_ABORTED|SCB_QUEUED_FOR_DONE;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
scbp->xs->error |= xs_error;
if(scbp->position != timedout_scb)
untimeout(ahc_timeout, (caddr_t)scbp);
outb(SCBPTR + iobase, scbp->position);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
outb(SCBARRAY + iobase, 0);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
i--;
found++;
}
}
/* Now put the saved scbs back. */
for (queued = 0; queued < i; queued++) {
outb (QINFIFO + iobase, saved_queue[queued]);
}
}
/*
* Search waiting for selection list.
*/
{
u_char next, prev;
next = inb(WAITING_SCBH + iobase); /* Start at head of list. */
prev = SCB_LIST_NULL;
while (next != SCB_LIST_NULL) {
scbp = ahc->scbarray[next];
/*
* Select the SCB.
*/
if (ahc_match_scb(scbp, target, channel)) {
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
next = ahc_abort_wscb(ahc, scbp, prev,
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
iobase, timedout_scb, xs_error);
found++;
}
else {
outb(SCBPTR + iobase, scbp->position);
prev = next;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
next = inb(SCB_NEXT_WAITING + iobase);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
}
}
}
/*
* Go through the entire SCB array now and look for
* commands for this target that are active. These
* are other (most likely tagged) commands that
* were disconnected when the reset occured.
*/
for(i = 0; i < ahc->numscbs; i++) {
scbp = ahc->scbarray[i];
if((scbp->flags & SCB_ACTIVE)
&& ahc_match_scb(scbp, target, channel)) {
/* Ensure the target is "free" */
ahc_unbusy_target(target, channel, iobase);
outb(SCBPTR + iobase, scbp->position);
Changes to support Performance enhancements in the sequencer. Bring back the polling code (for dumps mostly), but protect it with splbio() to make it safe for reprobes.
1996-01-29 03:17:39 +00:00
outb(SCBARRAY + iobase, 0);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
scbp->flags = SCB_ABORTED|SCB_QUEUED_FOR_DONE;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
scbp->xs->error |= xs_error;
if(scbp->position != timedout_scb)
untimeout(ahc_timeout, (caddr_t)scbp);
found++;
}
}
outb(SCBPTR + iobase, active_scb);
return found;
}
/*
* Manipulate the waiting for selection list and return the
* scb that follows the one that we remove.
*/
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static u_char
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
ahc_abort_wscb (ahc, scbp, prev, iobase, timedout_scb, xs_error)
struct ahc_data *ahc;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
struct scb *scbp;
u_char prev;
u_long iobase;
u_char timedout_scb;
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
u_int32_t xs_error;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
{
u_char curscbp, next;
int target = ((scbp->target_channel_lun >> 4) & 0x0f);
char channel = (scbp->target_channel_lun & SELBUSB) ? 'B' : 'A';
/*
* Select the SCB we want to abort and
* pull the next pointer out of it.
*/
curscbp = inb(SCBPTR + iobase);
outb(SCBPTR + iobase, scbp->position);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
next = inb(SCB_NEXT_WAITING + iobase);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
/* Clear the necessary fields */
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
outb(SCB_CONTROL + iobase, 0);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(SCB_NEXT_WAITING + iobase, SCB_LIST_NULL);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
ahc_unbusy_target(target, channel, iobase);
/* update the waiting list */
if( prev == SCB_LIST_NULL )
/* First in the list */
outb(WAITING_SCBH + iobase, next);
else {
/*
* Select the scb that pointed to us
* and update its next pointer.
*/
outb(SCBPTR + iobase, prev);
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(SCB_NEXT_WAITING + iobase, next);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
}
/*
* Point us back at the original scb position
* and inform the SCSI system that the command
* has been aborted.
*/
outb(SCBPTR + iobase, curscbp);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
scbp->flags = SCB_ABORTED|SCB_QUEUED_FOR_DONE;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
scbp->xs->error |= xs_error;
if(scbp->position != timedout_scb)
untimeout(ahc_timeout, (caddr_t)scbp);
return next;
}
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
static void
ahc_busy_target(target, channel, iobase)
u_char target;
char channel;
u_long iobase;
{
u_char active;
u_long active_port = ACTIVE_A + iobase;
if(target > 0x07 || channel == 'B') {
/*
* targets on the Second channel or
* above id 7 store info in byte two
* of HA_ACTIVE
*/
active_port++;
}
active = inb(active_port);
active |= (0x01 << (target & 0x07));
outb(active_port, active);
}
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
ahc_unbusy_target(target, channel, iobase)
u_char target;
char channel;
u_long iobase;
{
u_char active;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
u_long active_port = ACTIVE_A + iobase;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
if(target > 0x07 || channel == 'B') {
/*
* targets on the Second channel or
* above id 7 store info in byte two
* of HA_ACTIVE
*/
active_port++;
}
active = inb(active_port);
active &= ~(0x01 << (target & 0x07));
outb(active_port, active);
}
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static void
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
ahc_reset_current_bus(iobase)
u_long iobase;
{
outb(SCSISEQ + iobase, SCSIRSTO);
DELAY(1000);
outb(SCSISEQ + iobase, 0);
}
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static int
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_reset_channel(ahc, channel, timedout_scb, xs_error, initiate_reset)
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
struct ahc_data *ahc;
char channel;
u_char timedout_scb;
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
u_int32_t xs_error;
u_char initiate_reset;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
{
u_long iobase = ahc->baseport;
u_char sblkctl;
char cur_channel;
u_long offset, offset_max;
int found;
/*
* Clean up all the state information for the
* pending transactions on this bus.
*/
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
found = ahc_reset_device(ahc, ALL_TARGETS, channel,
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
timedout_scb, xs_error);
if(channel == 'B'){
ahc->needsdtr |= (ahc->needsdtr_orig & 0xff00);
ahc->sdtrpending &= 0x00ff;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(ACTIVE_B + iobase, 0);
offset = TARG_SCRATCH + iobase + 8;
offset_max = TARG_SCRATCH + iobase + 16;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
}
else if (ahc->type & AHC_WIDE){
ahc->needsdtr = ahc->needsdtr_orig;
ahc->needwdtr = ahc->needwdtr_orig;
ahc->sdtrpending = 0;
ahc->wdtrpending = 0;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(ACTIVE_A + iobase, 0);
outb(ACTIVE_B + iobase, 0);
offset = TARG_SCRATCH + iobase;
offset_max = TARG_SCRATCH + iobase + 16;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
}
else{
ahc->needsdtr |= (ahc->needsdtr_orig & 0x00ff);
ahc->sdtrpending &= 0xff00;
The long awaited stability patch set for the aic7xxx driver: Simplify the initialization of adapters by pulling all card specific initialization to the card specific modules. Update comments and fix formating. Pass struct ahc_data*'s to functions instead of unit numbers. Take advantage of the quad word alignment of SCB fields. Adapt to new sequencer changes: 1) Waiting scb list no longer has a tail. 2) Fill the message buffer as appropriate during a parity error. 3) Count all of the SGs involved in a residual instead of just the current one. The reset/abort code still needs a lot of work. Reviewed by: David Greenman <davidg@FreeBSd.org>
1996-01-03 06:32:12 +00:00
outb(ACTIVE_A + iobase, 0);
offset = TARG_SCRATCH + iobase;
offset_max = TARG_SCRATCH + iobase + 8;
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
}
for(;offset < offset_max;offset++) {
/*
* Revert to async/narrow transfers
* until we renegotiate.
*/
u_char targ_scratch;
targ_scratch = inb(offset);
targ_scratch &= SXFR;
outb(offset, targ_scratch);
}
/*
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
* Reset the bus if we are initiating this reset and
* restart/unpause the sequencer
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
*/
/* Case 1: Command for another bus is active */
sblkctl = inb(SBLKCTL + iobase);
cur_channel = (sblkctl & SELBUSB) ? 'B' : 'A';
if(cur_channel != channel)
{
/*
* Stealthily reset the other bus
* without upsetting the current bus
*/
outb(SBLKCTL + iobase, sblkctl ^ SELBUSB);
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
if( initiate_reset )
{
ahc_reset_current_bus(iobase);
}
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(CLRSINT1 + iobase, CLRSCSIRSTI|CLRSELTIMEO);
outb(CLRINT + iobase, CLRSCSIINT);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
outb(SBLKCTL + iobase, sblkctl);
UNPAUSE_SEQUENCER(ahc);
}
/* Case 2: A command from this bus is active or we're idle */
else {
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
if( initiate_reset )
{
ahc_reset_current_bus(iobase);
}
Fix support for the aic7850 by looking only at the relavent bits of the QINCNT. The 7850 puts random garbage in the high bits and all my attempts to determine the cause of this failed. This approach does seem to work around the problem. Go back to relying on the SCSIPERR interrupt instead of having the sequencer interrupt at the beginning of ITloop after a parity error occured. Determine the number of SCBs on a card automatically and base the qcntmask on the number of SCBs. Add entries for 11.4MHz, 8.8MHz, 8.0MHz, and 7.2MHz to ULTRA portion of the syncrate table. They seem to work fine on the 2940UW I have here and will allow more non-ultra devices (like my tape drive) to run sync while the adapter is in ULTRA mode. Return XS_SELTIMEOUT instead of XS_TIMEOUT for selection timeouts. I was getting sick of waiting for the SCSI code to retry each non-existant unit multiple times during boot and XS_SELTIMEOUT bypasses all retries. Use new SLIST queue macros. This was inspired by NetBSD using TAILQs in their SCSI drivers. For optimum cache hits, the free scb list should be LIFO which is what the old and new code does. NetBSD implemented a FIFO queue for some reason. Spaces -> tabs.
1996-03-31 03:15:31 +00:00
outb(CLRSINT1 + iobase, CLRSCSIRSTI|CLRSELTIMEO);
outb(CLRINT + iobase, CLRSCSIINT);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
RESTART_SEQUENCER(ahc);
}
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
ahc_run_done_queue(ahc);
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
return found;
}
bt.c, aic7xxx.c: Cleanse the SCSI subsystem of its internally defined types u_int32, u_int16, u_int8, int32, int16, int8. Use the system defined *_t types instead. aic7xxx.c: Fix the reset code. Instead of queing up all of the SCBs that timeout during timeout processing, we take the first and have it champion the effort. Any other scbs that timeout during timeout handling are given another lifetime to complete in the hopes that once timeout handing is finished, they will complete normally. If one of these SCBs times out a second time, we panic and Justin tries again. The other major change is to queue flag aborted SCBs during timeout handling, and "ahc_done" them all at once as soon as we have the controller back into a sane state. Calling ahc_done any earlier will cause the SCSI subsystem to toss the command right back at us and the attempt to queue the command will conflict with what the timeout routine is trying to accomplish. The aic7xxx driver will now respond to bus resets initiated by other devices.
1996-03-10 07:11:45 +00:00
void
ahc_run_done_queue(ahc)
struct ahc_data *ahc;
{
int i;
struct scb *scbp;
for(i = 0; i < ahc->numscbs; i++) {
scbp = ahc->scbarray[i];
if(scbp->flags & SCB_QUEUED_FOR_DONE)
ahc_done(ahc, scbp);
}
}
Cleanup, make things static.
1995-10-31 18:41:49 +00:00
static int
Long overdue, more complete, reset code. These changes implement a BUS DEVICE RESET followed by BUS RESET failure recovery strategy including the necesary renegotiation of sync/wide transfers after recovery completes. Clean up debugging code to make it more finely selectable. Reset code debugging is enabled for now so I can get more feedback on how this code behaves in real life.
1995-07-31 08:25:36 +00:00
ahc_match_scb (scb, target, channel)
struct scb *scb;
int target;
char channel;
{
int targ = (scb->target_channel_lun >> 4) & 0x0f;
char chan = (scb->target_channel_lun & SELBUSB) ? 'B' : 'A';
if (target == ALL_TARGETS)
return (chan == channel);
else
return ((chan == channel) && (targ == target));
}
Reference in New Issue Copy Permalink