1. The pageout daemon used to block under certain
circumstances, and we needed to add new functionality
that would cause the pageout daemon to block more often.
Now, the pageout daemon mostly just gets rid of pages
and kills processes when the system is out of swap.
The swapping, rss limiting and object cache trimming
have been folded into a new daemon called "vmdaemon".
This new daemon does things that need to be done for
the VM system, but can block. For example, if the
vmdaemon blocks for memory, the pageout daemon
can take care of it. If the pageout daemon had
blocked for memory, it was difficult to handle
the situation correctly (and in some cases, was
impossible).
2. The collapse problem has now been entirely fixed.
It now appears to be impossible to accumulate unnecessary
vm objects. The object collapsing now occurs when ref counts
drop to one (where it is more likely to be more simple anyway
because less pages would be out on disk.) The original
fixes were incomplete in that pathological circumstances
could still be contrived to cause uncontrolled growth
of swap. Also, the old code still, under steady state
conditions, used more swap space than necessary. When
using the new code, users will generally notice a
significant decrease in swap space usage, and theoretically,
the system should be leaving fewer unused pages around
competing for memory.
Submitted by: John Dyson
and into ether_input(). It was silly to have bpf want this one way and
ether_input want it another way. Ripped out trailer support from the few
remaining drivers that still had it.
1) make #includes correct
2) fix bugs in address check macros
3) fixed bugs in, and enabled, recopy if heavily fragmented code
4) moved call to bpf tap to be before enqueing packet (probably gratuitous)
5) fixed bug that caused "abnormal interrupt" at boot time/first use
6) added support for reading Zynx address ROM
7) fixed bug that caused broadcasts to not work shortly after booting (only
manifested if not using multicast - e.g. not in FreeBSD 2.0)
8) fixed spelling errors in comments
Submitted by: Matt Thomas
Subject: Mea culpa -- small fix for netboot fixes
In accordance with the unavoidable principle sof Murphy's Law, I discovered
that the fixes I recently contributed for the netboot code had some small
flaws in them. Two of them were just typos and had no effect on how the
program functioned. The other one was a missing line from the rootopts and
swapopts functions I created in bootmenu.c, which was supposed to initialize
the NFS sotype flag. It defaults to UDP, and you can change it to TCP with
the rootopts or swapopts commands, but then you can't change it back again.
I originally had a line at the top of each function to reinitialize this
flag, but somehow it got lost in the shuffle, probably because I don't
actually have a need for that flag yet.
Submitted by: wpaul
That was the good news. The bad news is that bad144 is a proper mess,
and I don't have time to fix it now, so you will probably not be able to
use it anyway.
Sorry guys, go out and buy a 100Mb IDE drive and a paddleboard :-(
If somebody wants to pick up on this: bad144 needs to learn how to
stay inside our slice of the disk. That's the trick.
Go to a single dependancy in files.i386. Using a .c file for the
sequencer code won't work since I need to know the size of the program,
so we just include the generated .h file as:
"../../sys/gnu/misc/aic7770/aic7770_seq.h"
Reviewed by:
Submitted by:
Obtained from:
instead. The entire scheme just doesn't work as envisioned (hint: think
about make depend as well as all). Those extremely rare individuals who
actually hack on the sequencer code will know how to keep stuff in sync,
I *do* get the feeling!
Somehow, I don't think this stuff was tested at all! :-(
I really hope that it actually works, though my hopes are steadily diminishing.
Anyone with 27xx/28xx boards in -current is *strongly encouraged* to give this
stuff a shot! Otherwise, I suspect that we'll be punting this out of
2.0. I haven't found a single part of Justin's commit that wasn't broken
in some way.
of config so YOU MUST RECOMPILE CONFIG. Modifying config was the cleanest
solution to integrating this driver into the tree which will become more
obvious in the next commit.
a discussion going on about removing this code from the burden of the
GPL, but it won't happen before Beta, and this code should be tested
before release.
Supports 27/2842 class adaptec cards and is almost capable of supporting
aic7870 based adapters (294X series cards). It does not support Wide
controllers or the second channel on Twin boards although I have work in
progress on getting both channels and running.
I have also added a few performance improvements to this version that give
us approximately a 25% boost over the original driver. These patches have
been submitted to the author.
Obtained from: Linux aic7770 driver (John Aycock - aycock@cpsc.ucalgary.ca)
Smack the netboot program around so that it will allow the user to
specify mount options. [So that you can boot from a privileged port]
Change the default boot image name in netboot to /kernel, then strip
the leading slash when actually going out to get the NFS file handle.
Added support for 3Com 3c503 cards. Also added another command to
the (trans) that allows you to switch the 3Com's on-board transceiver
on and off. (ether.c, ether.h, bootmenu.c)
Modified the Makefile to support new compile-time options for 3c503
cards:
-DINCLUDE_3COM Include support for 3c503
-D_3COM_BASE=0x300 Define 3c503 base i/o address (if not
specified, 0x300 is the default)
-D_3COM_USE_AUI Disable the 3c503's transceiver by
default (without this flag the transceiver
is on by default)
buffering scheme and make it more in tune with FreeBSD's vfs_bio
implementation. The filesystem seems fairly stable, but I wouldn't recommend
it to anyone not willing to experience problems. This is very green code and
has the limitation that YOU CAN ONLY HAVE ONE LFS PARTITION MOUNTED AT A TIME.
What LFS is good for:
Non fsynced writes FASTER THAN FFS
Large deletions Increadibly fast
Reads are a little bit slower than FFS right now, but that is a factor of
how under optimized this code is. LFS should in theory perform at least as
well as FFS under fsync (iozone) type loads, and this is what I'm currently
working on.
Reviewed by: Justin Gibbs
Submitted by: John Dyson
Obtained from:
stages of debugging LFS:
* if we can't bmap, use old VOP code
*/
! if (/* (vp->v_mount && vp->v_mount->mnt_stat.f_type == MOUNT_LFS) || */
! VOP_BMAP(vp, foff, &dp, 0, 0)) {
for (i = 0; i < count; i++) {
if (i != reqpage) {
vnode_pager_freepage(m[i]);
--- 804,810 ----
/*
* if we can't bmap, use old VOP code
*/
! if (VOP_BMAP(vp, foff, &dp, 0, 0)) {
Reviewed by: gibbs
Submitted by: John Dyson
was supposed to have already been made, but got botched somewhere.
Don't clobber the last page of memory (where the message buffer is). Some
BIOS don't gratuitously wipe it out on reboot.
Remove bogus declaration of Debugger(). Call Debugger() even if DDB is
not defined, but still call panic() after Debugger() returns, although
most other SCSI drivers just call Debugger().
printf() is inconsistent with the prototype for the library printf() and
gets declared if DIAGNOSTIC is defined because <vm/vm_page.h> includes
<sys/systm.h>.
Change the sc_link structure from 8 targets with 8 lun's to
16 targets with 8 lun's. Wide scsi and the ncr 825 does support 16 targets.
scsiconf.c:
move the addition of the trailing NUL's out of a then clause, they are also
needed in the else clause for an unknown device.
Alphabetize.
Write all i/o functions in sleep so that we don't use anything from
NetBSD.
Restore the correct type of u_int for ports. This saves a whole cycle
per i/o on 486's.
Change `inline' back to __inline to avoid compiler warnings with
-Wreally-all.
Don't implement bdb() unless BDE_DEBUGGER is defined. Declare bdb_exists
outside the function to avoid hundreds of compiler warnings.
Let the compiler pick the register in asms if possible.
Implement ffs() using inline asm(). gcc provides a slightly different
one. It was broken in gcc-2.4.5 but works now. Declaring a correct
version inline ensures getting a correct version. FreeBSD-1.1.5 has
an slow inline version but FreeBSD-2.0 has a library version (which
probably never gets used).
Do inb() and outb() without using %edx for constant ports below 0x100.
Remove casts to the same type in queue functions.
Declare prototypes for everything implemented i386/*.s and also for
everything that is normally implemented as an inline here (I don't
like the current complete dependency on gcc). Ifdef out the prototypes
that are declared elsewhere. THere should be a separate header to
declare things implemented in i386/*.s, but then it would be harder
to override declarations with inlines.
${UII}
with the current default exception (un)mask. There should be no such
processes unless you change the mask. Someday the mask should be
changed to the IEEE default of everything masked. The npx state
gets saved so that it can be checked and this may have the side effect
of fixing a bug that was reported for 1.1.5. (npx exceptions may
sometimes leak across exits and clobber another process. I can't see
how this can happen.)
Get some missing/wrong declarations from headers now that the headers
have them.
I couldn't find a better way to avoid compiler warnings about
redundant and/or inconsistent declaration of ffs(). I'd like to
be able to declare prototypes in general headers without committing
to implementing them as `static inline' or `extern', but there
seems to be no way to do this with gcc-2.6.1. E.g.,
int foo(void);
static __inline int foo(void) { return 1; }
causes a warning about the linkage mismatch, while the opposite
order causes a warning about the redundant declaration.
the following.
Move declarations to and from <machine/segments.h>. Make segment stuff
static if possible.
Remove unused (although initialized) global variables _default_ldt,
currentldt, _gsel_tss (rename the latter to the auto variable
gtel_tss).
Use "correct" and consistent types for interrupt handlers.
Remove a mailing address from the code.
Fix type mismatches found by adding prototypes.
Partly support BDE_DEBUGGER. Still broken by conflict with APM. Does
nothing if BDE_DEBUGGER is not defined.
Clean up prototypes and data declarations. Declare most of the segment
functions that are implemented in support.s. Make data private in
machdep.c if possible.
Parenthesize expressions in macros properly!
${Uniformize idempotency ifdef}.
to avoid compiler warnings.
Clean up prototypes: alphabetize; don't use redundant `extern' or
meaningless `extern inline'.
Uniformize idempotency ifdef.
(These functions are implemented in assembler so the compiler can't
check the declarations.)
Clean up prototypes: restore CSRG's alphabetical order, arg names in
prototypes, formatting to fit in 80 columns
Declare a complete prototype for the function pointer *ifa_rtrequest.
radix.h:
Declare a complete prototype for the function pointer *rnh_walktree
and for the function rn_walktree.
Uniformize idempotency ifdef.
Disable the bogus declaration of pmap_bootstrap(). Since its arg list
is machine-dependent, it must be declared in a machine-dependent header.
vm_page.h:
Change `inline' to `__inline' and old-style function parameter lists for
inlined functions to new-style.
`inline' and old-style function parameter lists should never be used in
system headers, even in very machine-dependent ones, because they cause
warnings from gcc -Wreally-all.
get truncated to LONG_MAX. Don't lobotomize the merged library source.
Make all private data static.
Use int_parms for the i/o "address" since the "address" is really a number
and is represented as an int.
Add command `flags' to allow changing device flags.
Fix scrolling of device listing. Only scrolling of the current devtab
was controlled. Reprint the header after scrolling.
Rename commands and change strings to match their config(8) keywords:
io -> port
IOaddr -> port
mem -> iomem (abbreviation is io :-()
MemAddr -> iomem
case changes
Don't use NULL for ASCII NUL.
Call strtoul() with base 0 for both numbers and addresses so that input
is consistent and hex and octal can be used for numbers.
Fix entry of irq number. Check the range at no extra cost. It wasn't
possible to enter irq -1.
Format device listing better. Large numbers (such as 0xffffffff for the
GENERIC lpt0 port) messed up the formatting.
Show the unit number in the device listing. Comment about the fields
that aren't shown.
fault was at offset 0 in the object. This resulted in more overhead but
was othewise benign. Added incore() check in vnode_pager_has_page()
to work around a problem with LFS...other than slightly higher overhead,
this change has no affect on UFS.
put the stuff into the right "distribution". As default things end up
in "bindist".
Normal (ie: most) makefiles know naught of this.
More commits will follow, which will direct various parts of the tree
into the distribution we want them in.
Some of the grief of being release-engineer is supposed to go away with this.
of the 1.1.5 driver, a recent version of the NetBSD driver, Andres'
transmission start threshold code, and all other relavent changes to the driver
since it was brought into 2.0. The multicast support from NetBSD has not be
folded in yet. I've tested it under high loads for two weeks and it is now
robust enough to be included in the GENERIC kernel.
Reviewed by: gibbs
Submitted by: vega@sophia.inria.fr (Andres Vega Garcia)
scsi adapter. Let it work on every adapter. Someone want to rewrite
this to be coupled with the adapter specification in the config file,
so that you can say which adapters need it ?
cosmetique) because we already have right things there or his changes
are incorrect.
Fix mcd_subchan to return position, inspired by idea from
bugress@s069.infonet.net, but different implementation.
Here is the improved probe for the mse (Bus Mouse) device driver. I
have been running with this under 1.1.5.1 as well as 2.0 without a hitch for
quite a while.
Submitted by: lars
device announcement; assume a sector size of 512 instead (likely to be
right at all).
This case happens when booting with a removable disk device attached
(e.g. an MOD), but no medium inserted.
I think it's time for Ugen to get a freefall account, just so I can
direct mail at him directly and let him drop off patches for us here. Ugen?
Done!
Submitted by: ugen
Somebody should make a mib variable for it.
Just now it is pointless to dump the kernel, since we have nothing which
can read the dump.
Furthermore is should never be the default to dump.
options DODUMP
will enable dumps.
Enabled via REL2_1.
Added support for doing object collapses "on the fly". Enabled via REL2_1a.
Improved object collapses so that they can happen in more cases. Improved
sensing of modified pages to fix an apparant race condition and improve
clustered pageout opportunities. Fixed an "oops" with not restarting page
scan after a potential block in vm_pageout_clean() (not doing this can result
in strange behavior in some cases).
Submitted by: John Dyson & David Greenman
for all reasonable HZ's. HZ > 1000 doesn't work because of sloppy
conversions in hzto() (division by (tick / 1000) == 0). This was
fixed in 1.1.5.
Eliminate some extern declarations by including the appropriate header
files that now contain appropriate declarations.
tsleep()). Try `dd if=/dev/zero of=/dev/pcaudio bs=640k count=1'. The
write takes a few hundred seconds to drain, and if it is killed by a
signal, it still takes a few hundred seconds to drain and all of those
seconds are spent busy-waiting.
Clean up includes and declarations. Remove bogus casts of args to
timeout functions.
doesn't have to calculate it every call.
Rename `timer0_prescale' to `timer0_prescaler_count' and maintain it
correctly. Previously we lost a few 8253 cycles for every "prescaled"
clock interrupt, and the lossage grows rapidly at 16 KHz. Now we
only lose a few cycles for every standard clock interrupt.
Rename `*_divisor' to `*_max_count'.
Do the calculation of TIMER_DIV(rate) only once instead of 3 times each
time the rate is changed.
Don't allow preposterously large interrupt rates. Bug fixes elsewhere
should allow the system to survive rates that saturate the system, however.
Clean up declarations.
Include <machine/clock.h> to check our own declarations.
outside the critical region.
Make it work with 2.0. It wasn't designed to be called at splclock().
Make it work with prescaling. The overflow threshold was bogus.
Make it work for any HZ. Side effect of fixing prescaling.
Speed it up. Allocate registers better. Reduce multiplication and
division to multiplication and a shift. Speed is now 5-6 usec on a
486DX/33, was about 3 usec more.
Optimize for the non-pentium case. The pentium code got moved around
a bit and hasn't been tested.
Change #include's to 2.0 style.
for it is incomplete and buggy. There is no problem unless Xintr0()
is reentered or should be reentered, but high clock interrupt
frequencies for pcaudio cause Xintr0() to be reentered (or clock
ticks to be lost when Xintr0() should have been reentered but
wasn't), and we lose little by delaying the call to softclock().
Move declarations related to the clock driver to clock.h.
Move declarations related to the npx driver to npx.h.
Clean up the remaining declarations.
Find enclosed a short bugfix to get the union filesystem up and running
in FreeBSD-current. We don't think we've got all the problems yet but
these fixes sort out the major ones (which mostly concert bad locking
of vnodes), no doubt we'll post others as necessary. Known problems
include the inability of the umount command (not the system call) to unmount
unions in certain circumstances (this is due the way "realpath" works),
and the failure of direntries to always get all available files in
unioned subdirectories. We are, as they say, working on it.
Submitted by: tim@cs.city.ac.uk (Tim Wilkinson)
I know that many of these entries are bogus and need to be revisited,
but let's get the tree working again for now and then do a pass through
looking at all the __FreeBSD__ entries, shall we?
different types, and with the 'local cache', what is freed isn't necessarily
what was originally malloced. This screws malloc's statistics and type
allocation limits, resulting eventually in a deadlock when one of the
limits is bogusly reached. Recent performance tests on a Pentium machine
indicate no improvement with this optimization anyway (this is something
to be looked at further).
NB: You will have to recompile programs which use the `rt_use' member in
order to get the correct values. This should not cause incorrect operation,
but the statistics may look a little confusing.
floppies now. I'm not sure why, but things hang when it gets to the
`changing root to fd0c' part. Without your latest commit, everything works
fine. Maybe you can figure out what you broke after ALPHA! :)
errors at a lower ipl. clist starvation problems can cause hundreds of
tty buffer overflows per second and logging them all amplified the
problems. This problem was less serious in 1.1.5.
Avoid a race in the check for starting a new block of output. com_events
was sometimes messed up and siopoll() looped endlessly. This bug was
introduced in 2.0.
Clean up previous 2 commits. Rename sio_registerdev() to sioregisterdev()
to match the (bad) surrounding naming conventions. There should be a
generic_registerdev().
a clist return with an error. There are some clist starvation/deadlock
bugs elsewhere and killing clist hogs didn't help because the breaks
only exited from the inner loops.
I'm not sure if this is just masking another problem (like, should
ap->a_eofflag EVER be NULL?), but if it prevents a panic for now then
it may save an ALPHA customer.
Submitted by: jhay
created by Amancio Hasty (specificly, this, in conjunction with his sound
driver mods for dual-mode DMA will allow VAT compiled for BSD/386 1.1 to
run under FreeBSD 2.x.)
Otherwise clean up the includes. Don't include anything included by
param.h. Do include systm.h and cons.h to avoid satisfy -Wimplicit.
Don't include console.h or use NOKEY because these are for syscons
and we use generic consoles.
Don't follow null pointer for command "ls -lrt" - don't allow extra
args but do allow trailing blanks.
Check for invalid device numbers. strtol() failures are now checked
for in all cases, but not carefully enough. We should check for
trailing junk, allow any base in all cases (just like config) and
handle signs better . (Use strtoul not strtol and cast by assignment
to the correct type - always an integral type, PARM_ADDR is bogus.
Hex numbers > 0x7fffffff can't be entered now. 0xffffffff has to
be entered as -1.)
Cosmetic.
Return from trap() if trap_fatal() returns. trap_fatal() isn't
fatal if you have ddb. Returning from trap() is usually the right
thing to do and much better than falling through.
Build a dummy frame at the top of tmpstk to help debuggers trace the stack
when the system is idle.
swtch.s: idle():
Initialize the frame pointer so that debuggers don't try to trace a bogus
stack.
Load the frame pointer, load the stack pointer and switch out the old
stack in the unique order that never leaves one of the pointers pointers
invalid so that debuggers can trace idle(). Disabling interrupts
provides sufficient validity for normal operation, but debuggers use
(trace) traps.
. avoid resetting the FDC every time the last motor is going off;
instead, give it a 60-second period for possible later reactivation.
This prevents us from needing to recalibrate the FDC too often,
but still allows for an ``automagic error recovery', just in case the
controller is absolutely stuck. (Simply wait for 60 seconds, and
try it again.)
. made the floppy head settle time after a seek a constant
that might be overridden by a config option. (Well, actually the
divisor of the settle time). Pepople often reported problems with
their floppies, so i need a simply mechanism that allows them
to play with that value. (I personally cannot find any problem
on *my* drives.)
. implement the FD_DEBUG ioctl command, in case the driver
is compiled with DEBUG turned on.
. removed a bogus parameter from a printf; the remaining warnings
from gcc -Wall seem to be warnings about the %b format gcc cannot
understand
. rearrange Garett's code to fit better in the existing structure
of #define/type/function ordering.
. make everything fit into 79 columns again.
This way, it's possible for a user to activate/deactivate floppy driver
debugging, even if (s)he doesn't like the overhead of running DDB.
Since some ppl tend to have problems with their drives, this might be
valuable for investigations.
in the far pointers are multiples of 4K (as is normal when the video
BIOS is at seg 0xc000). Disallow mode switching if the pointer is bad.
Use a new pa_to_va() macro for all BIOS and video addresses in syscons.
