]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoAdd support PLANEX COMMUNICATIONS Inc. GN-1200TC (8169S)
sanpei [Thu, 4 Mar 2004 06:42:46 +0000 (06:42 +0000)]
Add support PLANEX COMMUNICATIONS Inc. GN-1200TC (8169S)

20 years agoFix an off-by-one error and rework our EC space handler. Writing to address
njl [Thu, 4 Mar 2004 05:58:50 +0000 (05:58 +0000)]
Fix an off-by-one error and rework our EC space handler.  Writing to address
0xFF would fail previously as AE_BAD_PARAMETER.  It's unknown if this caused
any actual problems.

20 years agoPart 2 of Project Evil: Pretend to be Windows 2000 for buggy ASL that
njl [Thu, 4 Mar 2004 05:57:41 +0000 (05:57 +0000)]
Part 2 of Project Evil:  Pretend to be Windows 2000 for buggy ASL that
always expects to be running on some MS OS.  A survey of ASL shows that
this is the 2nd most common expected OS value.  (1st is Win98 and we don't
emulate its buggy ACPI support.)  Our ACPI support is similar to Win2k,
also.  Put this behavior under ACPICA_PEDANTIC so we can get back to our
previous behavior for OSV testing.

20 years agoAdd NetBSD 1.6.2.
maxim [Thu, 4 Mar 2004 05:34:29 +0000 (05:34 +0000)]
Add NetBSD 1.6.2.

20 years agoDon't disable Cx support and throttling on machines with a P_BLK_LEN != 6
njl [Thu, 4 Mar 2004 05:17:52 +0000 (05:17 +0000)]
Don't disable Cx support and throttling on machines with a P_BLK_LEN != 6
even though the spec mandates this.  Some have a value of 5 to indicate
throttling + C2 and some have 7 to indicate an extra C3 state.  Support
throttling if the value is >= 4, C2 for >= 5, and C3 for >= 6.

20 years agoAdd a "quirks" value to disable quirks handling for a given boot.
njl [Thu, 4 Mar 2004 04:42:59 +0000 (04:42 +0000)]
Add a "quirks" value to disable quirks handling for a given boot.
Also, disable quirks if booting with a custom DSDT.  Add a quirk
to disable loading ACPI so known bad systems can be completely
blacklisted.

20 years agoAdd a new option to mountd(8), -p <port>. This allows the user to specify
bms [Thu, 4 Mar 2004 04:42:52 +0000 (04:42 +0000)]
Add a new option to mountd(8), -p <port>. This allows the user to specify
a known port for use in firewall rulesets; otherwise the port is chosen
at run-time by bindresvport().

MFC after: 1 week

20 years agoRemove some long unused definitions.
alc [Thu, 4 Mar 2004 04:26:14 +0000 (04:26 +0000)]
Remove some long unused definitions.

20 years agoRemove unneeded label 'done2' from socket(). We now grab Giant
rwatson [Thu, 4 Mar 2004 01:57:48 +0000 (01:57 +0000)]
Remove unneeded label 'done2' from socket().  We now grab Giant
only around socreate(), and don't need it for file descriptor
accesses.

Submitted by: sam

20 years agos/considred/considered/
brueffer [Thu, 4 Mar 2004 00:52:16 +0000 (00:52 +0000)]
s/considred/considered/

20 years agoAdd sanity checks to the ndis_packet and ndis_buffer pool handling
wpaul [Thu, 4 Mar 2004 00:17:14 +0000 (00:17 +0000)]
Add sanity checks to the ndis_packet and ndis_buffer pool handling
routines to guard against problems caused by (possibly) buggy drivers.

The RealTek 8180 wireless driver calls NdisFreeBuffer() to release
some of its buffers _after_ it's already called NdisFreeBufferPool()
to destroy the pool to which the buffers belong. In our implementation,
this error causes NdisFreeBuffer() to touch stale heap memory.

If you are running a release kernel, and hence have INVARIANTS et al
turned off, it turns out nothing happens. But if you're using a
development kernel config with INVARIANTS on, the malloc()/free()
sanity checks will scribble over the pool memory with 0xdeadc0de
once it's released so that any attempts to touch it will cause a
trap, and indeed this is what happens. It happens that I run 5.2-RELEASE
on my laptop, so when I tested the rtl8180.sys driver, it worked fine
for me, but people trying to run it with development systems checked
out or cvsupped from -current would get a page fault on driver load.

I can't find any reason why the NDISulator would cause the RealTek
driver to do the NdisFreeBufferPool() prematurely, and the same driver
obviously works with Windows -- or at least, it doesn't cause a crash:
the Microsoft documentation for NdisFreeBufferPool() says that failing
to return all buffers to the pool before calling  NdisFreeBufferPool()
causes a memory leak.

I've written to my contacts at RealTek asking them to check if this
is indeed a bug in their driver. In the meantime, these new sanity checks
will catch this problem and issue a warning rather than causing a trap.
The trick is to keep a count of outstanding buffers for each buffer pool,
and if the driver tries to call NdisFreeBufferPool() while there are still
buffers outstanding, we mark the pool for deletion and then defer
destroying it until after the last buffer has been reclaimed.

20 years agoUse different dummy wait channels to avoid panic in msleep().
des [Wed, 3 Mar 2004 23:03:18 +0000 (23:03 +0000)]
Use different dummy wait channels to avoid panic in msleep().

Reviewed by: jhb

20 years ago[this is just a forced commit to say:] The time_t-specific safety measure
gad [Wed, 3 Mar 2004 22:56:41 +0000 (22:56 +0000)]
[this is just a forced commit to say:]  The time_t-specific safety measure
added by the sparc64_installcheck target is mostly from Marcel, although
it includes some adjustments of my own...

20 years ago- Remove d_valid field, we can use d_consumer field to check if disk
pjd [Wed, 3 Mar 2004 22:29:24 +0000 (22:29 +0000)]
- Remove d_valid field, we can use d_consumer field to check if disk
  is valid.
- Use SYSCTL_DECL() instead of using own, ugly extern.

20 years agoAdd regression tests for geom_concat class.
pjd [Wed, 3 Mar 2004 21:52:49 +0000 (21:52 +0000)]
Add regression tests for geom_concat class.

20 years agoMake this WARNS=2 clean by:
johan [Wed, 3 Mar 2004 20:21:15 +0000 (20:21 +0000)]
Make this WARNS=2 clean by:
- including <timeconv.h> to get _long_to_time prototype
- removing an unused variable

Bump WARNS to it keep clean.

Approved by: ume on ipfw@

20 years ago(that should be: "assuming NO major problems are found", of course...)
gad [Wed, 3 Mar 2004 19:48:52 +0000 (19:48 +0000)]
(that should be: "assuming NO major problems are found", of course...)

Also note that these three files are intended to be short-term aids for
making this specific transition.  They will probably disappear after
5.3-release (although the installworld_*k scripts do include some
general-purpose ideas that we might want to consider at a later date).

20 years agoCommit the first set of files for changing time_t on freebsd/sparc64
gad [Wed, 3 Mar 2004 19:36:20 +0000 (19:36 +0000)]
Commit the first set of files for changing time_t on freebsd/sparc64
from a 32-bit value to a 64-bit value.  This commit does not actually
change anything.  It merely provides instructions, scripts, and a safety
measure in Makefile.inc1 for people who want to make the change.

The real change to 64-bit time_t's on sparc64 is scheduled to happen
on March 10th, assuming that so major problems are found between now
and then by early-adopters.

Reviewed by: freebsd-sparc64

20 years agoChange to acpi_{Get,Set}Integer to provide both methods. Convert all
njl [Wed, 3 Mar 2004 18:34:42 +0000 (18:34 +0000)]
Change to acpi_{Get,Set}Integer to provide both methods.  Convert all
callers to the new API.

Submitted by: Mark Santcroos <marks@ripe.net>

20 years agoAdd proper support for DbgPrint(): only print messages if bootverbose
wpaul [Wed, 3 Mar 2004 17:57:05 +0000 (17:57 +0000)]
Add proper support for DbgPrint(): only print messages if bootverbose
is set, since some drivers with debug info can be very chatty.

Also implement DbgBreakPoint(), which is the Windows equivalent of
Debugger(). Unfortunately, this forces subr_ntoskrnl.c to include
opt_ddb.h.

20 years agoFixed "make clean".
ru [Wed, 3 Mar 2004 17:04:49 +0000 (17:04 +0000)]
Fixed "make clean".

20 years agoAdd acpi_video(4) man page.
njl [Wed, 3 Mar 2004 16:34:03 +0000 (16:34 +0000)]
Add acpi_video(4) man page.

PR:
Submitted by: Mark Santcroos <marks@ripe.net>
Reviewed by:
Approved by:
Obtained from:
MFC after:

20 years agoSome (most?) processes don't cleanup their pid files when
mtm [Wed, 3 Mar 2004 15:28:44 +0000 (15:28 +0000)]
Some (most?) processes don't cleanup their pid files when
they exit. Sometimes they simply can't (core-dump, for example).
So, when searching for a running program send standard error
output from ps(1) to never-never land.

# This should quite those
# ps: kvm_getprocs: No such user
# errors. Since ports use a duplicate rc.subr(8) you might
# still see this error from ports startup scripts.

20 years agoFrom the PR:
mtm [Wed, 3 Mar 2004 15:21:01 +0000 (15:21 +0000)]
From the PR:
Certain MTA configurations mean that the notifications from
virecover keep bouncing; so here's a patch to allow administrators
to turn them off.

PR: conf/54910
Submitted by: bms (with a minor cleanup)

20 years agoDocument missing EFAULT errno value.
dds [Wed, 3 Mar 2004 14:51:11 +0000 (14:51 +0000)]
Document missing EFAULT errno value.

MFC after: 2 weeks

20 years ago- stlye and comments
ume [Wed, 3 Mar 2004 14:33:16 +0000 (14:33 +0000)]
- stlye and comments
- variable name change (scopeid -> zoneid)
- u_short -> u_int16_t, u_char -> u_int8_t

Obtained from: KAME

20 years ago1. Add ct.4 manual page entry.
rik [Wed, 3 Mar 2004 13:26:18 +0000 (13:26 +0000)]
1. Add ct.4 manual page entry.
2. Add ctau.4 link to ct.4.

Approved by: imp (mentor)

20 years agoAdd manual page for Cronyx/Tau-ISA device driver.
rik [Wed, 3 Mar 2004 13:23:18 +0000 (13:23 +0000)]
Add manual page for Cronyx/Tau-ISA device driver.

Approved by: imp (mentor)

20 years agoAdd ctau (Cronyx/Tau-ISA) device driver entry.
rik [Wed, 3 Mar 2004 12:43:59 +0000 (12:43 +0000)]
Add ctau (Cronyx/Tau-ISA) device driver entry.

Approved by: imp (mentor)

20 years agoAdd rules for font.h atkbdmap.h ukbdmap.h so more of LINT can be built.
obrien [Wed, 3 Mar 2004 09:37:44 +0000 (09:37 +0000)]
Add rules for font.h atkbdmap.h ukbdmap.h so more of LINT can be built.

20 years agoAdd missing working options from i386 file.
obrien [Wed, 3 Mar 2004 09:35:47 +0000 (09:35 +0000)]
Add missing working options from i386 file.

20 years agoLower i386's INLINE_LIMIT to AMD64's level.
obrien [Wed, 3 Mar 2004 09:35:06 +0000 (09:35 +0000)]
Lower i386's INLINE_LIMIT to AMD64's level.

20 years agomga_state & r128_state need an inline-limit larger than amd64's default
obrien [Wed, 3 Mar 2004 09:33:03 +0000 (09:33 +0000)]
mga_state & r128_state need an inline-limit larger than amd64's default
in order to compile.  AMD64's default could be reduced below 8k, if
dev/aic7xxx/aic79xx.c got this same treatment.

20 years agoSync with i386/NOTES rev. 1.1131.
obrien [Wed, 3 Mar 2004 09:26:40 +0000 (09:26 +0000)]
Sync with i386/NOTES rev. 1.1131.

20 years agoAdd a note explaining how to use libmap.conf to work around problems
des [Wed, 3 Mar 2004 09:17:06 +0000 (09:17 +0000)]
Add a note explaining how to use libmap.conf to work around problems
with the libc_r -> libpthread transition.

20 years agoTypo fix.
blackend [Wed, 3 Mar 2004 09:15:07 +0000 (09:15 +0000)]
Typo fix.

PR: docs/63677
Submitted by: Volker Stolz <stolz@i2.informatik.rwth-aachen.de>
MFC after: 1 week

20 years agoPeter prefers it this way, bde might also[*]. I just want to have a chance
obrien [Wed, 3 Mar 2004 08:33:34 +0000 (08:33 +0000)]
Peter prefers it this way, bde might also[*].  I just want to have a chance
of working on amd64 for vmware use.
[*] bde will probably not like either version...

20 years agoMake return code in noarg case GNU-compatible
ache [Wed, 3 Mar 2004 08:29:00 +0000 (08:29 +0000)]
Make return code in noarg case GNU-compatible

20 years agoPrefer uintptr_t to intptr_t.
obrien [Wed, 3 Mar 2004 08:27:33 +0000 (08:27 +0000)]
Prefer uintptr_t to intptr_t.

20 years agoUse a long as the opaque type so that it matches the size of a pointer
obrien [Wed, 3 Mar 2004 08:24:31 +0000 (08:24 +0000)]
Use a long as the opaque type so that it matches the size of a pointer
on both 32-bit and 64-bit platforms.

20 years agoChange __FUNCTION__ to __func__
scottl [Wed, 3 Mar 2004 07:43:34 +0000 (07:43 +0000)]
Change __FUNCTION__ to __func__

Submitted by: Stefan Farfeleder

20 years agoBlah, s/newbus/busdma/ in the last commit.
obrien [Wed, 3 Mar 2004 07:33:43 +0000 (07:33 +0000)]
Blah, s/newbus/busdma/ in the last commit.

20 years agoFinish the removal of the 'nv' device reference.
scottl [Wed, 3 Mar 2004 07:16:42 +0000 (07:16 +0000)]
Finish the removal of the 'nv' device reference.

20 years agoOpps, nv(4) isn't committed yet. Comment out.
obrien [Wed, 3 Mar 2004 07:09:40 +0000 (07:09 +0000)]
Opps, nv(4) isn't committed yet.  Comment out.

20 years agoAMD64 versions.
obrien [Wed, 3 Mar 2004 07:00:48 +0000 (07:00 +0000)]
AMD64 versions.

20 years agoAdjust lnc(4) for 64-bit platforms should it get newbus'ified.
obrien [Wed, 3 Mar 2004 06:54:26 +0000 (06:54 +0000)]
Adjust lnc(4) for 64-bit platforms should it get newbus'ified.

20 years agoAdjust ed(4) for 64-bit platforms should it get newbus'ified.
obrien [Wed, 3 Mar 2004 06:48:42 +0000 (06:48 +0000)]
Adjust ed(4) for 64-bit platforms should it get newbus'ified.

20 years agoUse a long as the opaque type so that it matches the size of a pointer
obrien [Wed, 3 Mar 2004 06:20:36 +0000 (06:20 +0000)]
Use a long as the opaque type so that it matches the size of a pointer
on both 32-bit and 64-bit platforms.

20 years agoAdd memory barrier routines for AMD64.
obrien [Wed, 3 Mar 2004 06:19:03 +0000 (06:19 +0000)]
Add memory barrier routines for AMD64.

20 years agoCast thru intptr_t on the way to void* for success on 64-bit platforms.
obrien [Wed, 3 Mar 2004 06:18:29 +0000 (06:18 +0000)]
Cast thru intptr_t on the way to void* for success on 64-bit platforms.

20 years agoFinish fixing style problems in my previous commit.
njl [Wed, 3 Mar 2004 05:51:48 +0000 (05:51 +0000)]
Finish fixing style problems in my previous commit.

Pointed out by:  bde

20 years agoThis commit was generated by cvs2svn to compensate for changes in r126520,
njl [Wed, 3 Mar 2004 05:47:09 +0000 (05:47 +0000)]
This commit was generated by cvs2svn to compensate for changes in r126520,
which included commits to RCS files with non-trunk default branches.

20 years agoIn the descrption of network interfaces:
bde [Wed, 3 Mar 2004 05:26:03 +0000 (05:26 +0000)]
In the descrption of network interfaces:
- use consistent formatting (no tabs)
- improved wording for cx and oltr.

20 years agoBe more GNU-compatible in diagnostics
ache [Wed, 3 Mar 2004 03:05:21 +0000 (03:05 +0000)]
Be more GNU-compatible in diagnostics

20 years agoAdd support for quirks for acpi tables. Key off OEM vendor and revision.
njl [Wed, 3 Mar 2004 03:02:17 +0000 (03:02 +0000)]
Add support for quirks for acpi tables.  Key off OEM vendor and revision.
Sort acpi debug values.  Change "disable" to "disabled" to match rest of
the kernel.  Remove debugging from acpi_toshiba since it was only used for
probe/attach.

20 years agoMinor style(9) fixes. remove a register keyword, correct two indents.
gad [Wed, 3 Mar 2004 02:44:52 +0000 (02:44 +0000)]
Minor style(9) fixes.  remove a register keyword, correct two indents.
Noticed while comparing to OpenBSD version.

20 years agoConvert K&R-style routine definitions to C89 style, partially to reduce
gad [Wed, 3 Mar 2004 02:41:21 +0000 (02:41 +0000)]
Convert K&R-style routine definitions to C89 style, partially to reduce
diffs with OpenBSD.  This causes no changes to the object produced.

20 years agoFix a couple of camcontrol issues that popped up on sparc64:
ken [Wed, 3 Mar 2004 01:51:24 +0000 (01:51 +0000)]
Fix a couple of camcontrol issues that popped up on sparc64:

 - bzero the CCB header in getdevtree() and set the path properly, to
   avoid having random garbage in the CCB header.
 - if the lun isn't specified in a device specifier, it should default to
   0, not whatever random data happens to be in the lun variable.
 - move the prototype for getdevtree() out from under #ifndef MINIMALISTIC,
   since it is used in both cases.

Submitted by: Marius Strobl <marius@alchemy.franken.de> (mostly)
MFC After: 2 weeks

20 years agoAdd some missing DUMMYNET_UNLOCK() in config_pipe().
mlaier [Wed, 3 Mar 2004 01:33:22 +0000 (01:33 +0000)]
Add some missing DUMMYNET_UNLOCK() in config_pipe().

Noticed by: Simon Coggins
Approved by: bms(mentor)

20 years agoForced commit to note that files have been repocopied from
cperciva [Wed, 3 Mar 2004 00:45:30 +0000 (00:45 +0000)]
Forced commit to note that files have been repocopied from
src/sbin/nologin to src/usr.sbin/nologin in preparation for
moving nologin(8) from /sbin to /usr/sbin.

20 years agoRemove reference to cx.4. A manpage doen't have to reference itself :-)
brueffer [Tue, 2 Mar 2004 20:42:21 +0000 (20:42 +0000)]
Remove reference to cx.4.  A manpage doen't have to reference itself :-)

20 years agoMove PFIL_HOOKS and ipfw past the scope checks to allow easy redirection to
mlaier [Tue, 2 Mar 2004 20:29:55 +0000 (20:29 +0000)]
Move PFIL_HOOKS and ipfw past the scope checks to allow easy redirection to
linklocal.

Obtained from: OpenBSD
Reviewed by: ume
Approved by: bms(mentor)

20 years agoMake swapbacked md(4) devices respect the -x and -y emulation arguments.
phk [Tue, 2 Mar 2004 20:13:23 +0000 (20:13 +0000)]
Make swapbacked md(4) devices respect the -x and -y emulation arguments.

20 years agoIf handed a file pointer we can't write to, set errno properly to EBADF
jkh [Tue, 2 Mar 2004 19:43:51 +0000 (19:43 +0000)]
If handed a file pointer we can't write to, set errno properly to EBADF
in order to get SUSv2 conformant behavior in higher level calls like
fputs() and puts().

Reviewed by: bde

20 years agoWhitespace fixes.
obrien [Tue, 2 Mar 2004 19:15:13 +0000 (19:15 +0000)]
Whitespace fixes.

20 years agoAdd new Matrix Orbital LCD panel id's so that they are recognized and
peter [Tue, 2 Mar 2004 19:03:26 +0000 (19:03 +0000)]
Add new Matrix Orbital LCD panel id's so that they are recognized and
attached via uftdi->ucom.

20 years agoRegen
peter [Tue, 2 Mar 2004 19:01:56 +0000 (19:01 +0000)]
Regen

20 years agoAdd some device id's for Matrix Orbital's newer LCD panels. These use
peter [Tue, 2 Mar 2004 19:01:30 +0000 (19:01 +0000)]
Add some device id's for Matrix Orbital's newer LCD panels.  These use
another ftdi usb->serial bridge with different ID's.

20 years agoFix plural usage and a grammar nit
brueffer [Tue, 2 Mar 2004 18:38:32 +0000 (18:38 +0000)]
Fix plural usage and a grammar nit

20 years agoo remove stray 'the'
brueffer [Tue, 2 Mar 2004 18:30:25 +0000 (18:30 +0000)]
o remove stray 'the'
o remove instance of wrong apostrophe usage

MFC after: 3 days

20 years agoLanguage cleanup
brueffer [Tue, 2 Mar 2004 18:20:42 +0000 (18:20 +0000)]
Language cleanup

20 years ago1. Add modules makefile for Cronyx Tau/ISA driver (ctau).
rik [Tue, 2 Mar 2004 16:51:34 +0000 (16:51 +0000)]
1. Add modules makefile for Cronyx Tau/ISA driver (ctau).

Approved by: imp (mentor)

20 years ago1. Connect Cronyx Tau/ISA driver (ctau) to kernel.
rik [Tue, 2 Mar 2004 16:48:19 +0000 (16:48 +0000)]
1. Connect Cronyx Tau/ISA driver (ctau) to kernel.

Approved by: imp (mentor)

20 years ago1. Fix compilation and panic while system boot problem after makedev was
rik [Tue, 2 Mar 2004 16:44:07 +0000 (16:44 +0000)]
1. Fix compilation and panic while system boot problem after makedev was
changed to unde2dev.

Approved by: imp (mentor)

20 years ago1. Renames NCT constant to NCTAU. This will help while MFC to 4 branch.
rik [Tue, 2 Mar 2004 16:39:40 +0000 (16:39 +0000)]
1. Renames NCT constant to NCTAU. This will help while MFC to 4 branch.
2. Fix compilation and panic while system boot problem after makedev
was changed to unde2dev.

Approved by: imp (mentor)

20 years agoIf being verbose in the autosense code, print the original error.
sos [Tue, 2 Mar 2004 16:16:54 +0000 (16:16 +0000)]
If being verbose in the autosense code, print the original error.

20 years agoscope awareness of ff01:: is not merged, yet. So, clear
ume [Tue, 2 Mar 2004 16:01:27 +0000 (16:01 +0000)]
scope awareness of ff01:: is not merged, yet.  So, clear
embeded form of scopeid for ff01:: for now.

Pointed out by: mlaier

20 years agoAlways assert that the passed in lock is the same as the saved lock in the
jhb [Tue, 2 Mar 2004 15:02:08 +0000 (15:02 +0000)]
Always assert that the passed in lock is the same as the saved lock in the
sleep queue now that the one abnormal case has been fixed.

20 years agoCorrect handling of PDROP in msleep() to just skip the mtx_lock() rather
jhb [Tue, 2 Mar 2004 14:58:33 +0000 (14:58 +0000)]
Correct handling of PDROP in msleep() to just skip the mtx_lock() rather
than clear the lock pointer so that sleepq_add() still gets the correct
lock pointer and doesn't bogusly trip an assertion.

20 years agoTwo minor follow-ups on the MT_TAG removal:
mlaier [Tue, 2 Mar 2004 14:37:23 +0000 (14:37 +0000)]
Two minor follow-ups on the MT_TAG removal:
ifp is now passed explicitly to ether_demux; no need to look it up again.
Make mtag a global var in ip_input.

Noticed by: rwatson
Approved by: bms(mentor)

20 years agoProtect shell expansions.
ru [Tue, 2 Mar 2004 14:32:23 +0000 (14:32 +0000)]
Protect shell expansions.

20 years agoElf_Phdr.p_type 7 is "PT_TLS".
jake [Tue, 2 Mar 2004 14:06:27 +0000 (14:06 +0000)]
Elf_Phdr.p_type 7 is "PT_TLS".

20 years agoReport the original command on failures that causes auto sense.
sos [Tue, 2 Mar 2004 14:05:12 +0000 (14:05 +0000)]
Report the original command on failures that causes auto sense.
Keep the ATA_R_QUIET flag if set during autosense.

20 years agoFix getting progress data for some device in yet another way.
sos [Tue, 2 Mar 2004 14:03:43 +0000 (14:03 +0000)]
Fix getting progress data for some device in yet another way.
Take advantage of the new autosense logic.

20 years agoModify contigmalloc1() so that the free page queues lock is not held when
alc [Tue, 2 Mar 2004 08:25:58 +0000 (08:25 +0000)]
Modify contigmalloc1() so that the free page queues lock is not held when
vm_page_free() is called.  The problem with holding this lock is that it is
a spin lock and vm_page_free() may attempt the acquisition of a different
default-type lock.

20 years agoIncrease kernel VA from 256Mb to 512Mb by shifting the segment used
grehan [Tue, 2 Mar 2004 06:49:21 +0000 (06:49 +0000)]
Increase kernel VA from 256Mb to 512Mb by shifting the segment used
for user copyinout down to 12, and keeping segments 13/14 for
kernel VA.

It would be nice to have more available, but segments lower than
this are reserved for either memory or 1:1 mapped device i/o,
and seg 15 is OpenFirmware ROM. Also, the effort to keep OpenFirmware
available for callbacks limits the use of VA-mapped segments.
Fortunately UMA_MD_SMALL_ALLOC takes away a lot of VM pressure.

Obtained from:  NetBSD

20 years agoHook libthr to the build for PowerPC
grehan [Tue, 2 Mar 2004 06:26:37 +0000 (06:26 +0000)]
Hook libthr to the build for PowerPC

20 years agolibthr powerpc support.
grehan [Tue, 2 Mar 2004 06:25:20 +0000 (06:25 +0000)]
libthr powerpc support.

Submitted by:  Suleiman Souhlal <refugee@segfaulted.com>
Tested with: most libpthread tests, Apache 'worker' MDM

20 years agomakecontext lib call.
grehan [Tue, 2 Mar 2004 06:22:24 +0000 (06:22 +0000)]
makecontext lib call.

Submitted by: Suleiman Souhlal <refugee@segfaulted.com>

20 years agoKernel changes for libthr (and probably libpthread).
grehan [Tue, 2 Mar 2004 06:13:09 +0000 (06:13 +0000)]
Kernel changes for libthr (and probably libpthread).

include/ucontext.h
 - remove trapframe and switch over to 'generic' description of machine
   state. Include version field to help with future modifications.
   Include floating point and altivec state, and hopefully align
   correctly

powerpc/copyinout.c
 - fill out casuptr() sync primitive, required by kern_umtx.c

powerpc/machdep.c
 - shifted proc0/thread0/pcpu setup to before cninit, since
   syscons -> make_dev -> devlock requires a valid curthread
 - implemented get_mcontext/set_mcontext
 - recast sendsig/sigreturn to use get/set_mcontext and new
   ucontext struct. floating point now saved
 - TODO: save/restore altivec state

powerpc/vm_machdep.c
 - implemented cpu_thread_setup/cpu_set_upcall/cpu_set_upcall_kse
 - eliminated trailing whitespace

Submitted by:  Suleiman Souhlal <refugee@segfaulted.com>, ucontext by grehan

20 years agoWhen we get a packet error, move on, don't go into an infinite loop
julian [Tue, 2 Mar 2004 05:43:42 +0000 (05:43 +0000)]
When we get a packet error, move on, don't go into an infinite loop
looking at it.

fixes at least one cause of "hanging" due to this driver.

20 years agoAdd CTLTYPE_QUAD to the list of types for which we don't want an empty
dd [Tue, 2 Mar 2004 05:42:52 +0000 (05:42 +0000)]
Add CTLTYPE_QUAD to the list of types for which we don't want an empty
value.  All the other numeric types are doing it . . .

20 years agoWhitespace changes to match rest of file..
julian [Tue, 2 Mar 2004 01:46:34 +0000 (01:46 +0000)]
Whitespace changes to match rest of file..

20 years agoCheck for TDF_SINTR before calling sleepq_abort() as there is a narrow
jhb [Mon, 1 Mar 2004 23:07:58 +0000 (23:07 +0000)]
Check for TDF_SINTR before calling sleepq_abort() as there is a narrow
race in between sleepq_add() and sleepq_catch_signals() in that setting
td_wchan and TDF_SINTR is not atomic to sched_lock but only to the sleepq
lock.  This band-aid will stop assertion failures, but there is perhaps a
larger problem with the sleepq_add/sleepq_catch_signals race that I am not
sure how to solve.  For the signals case the race is harmless because we
always call cursig() after setting TDF_SINTR.  However, KSE doesn't do
anything in sleepq_catch_signals() to check that this race was lost, so I
am unsure if this race is harmful for this specific abort.

20 years ago- Fix typo.
simon [Mon, 1 Mar 2004 22:58:07 +0000 (22:58 +0000)]
- Fix typo.
- Use .Fx when I'm here anyway.

PR: docs/63622
Submitted by: Andrew Boothman <andrew@mux.org.uk>

20 years agoRename NET_PICKUP_GIANT() to NET_LOCK_GIANT(), and NET_DROP_GIANT()
rwatson [Mon, 1 Mar 2004 22:37:01 +0000 (22:37 +0000)]
Rename NET_PICKUP_GIANT() to NET_LOCK_GIANT(), and NET_DROP_GIANT()
to NET_UNLOCK_GIANT().  While they are used in similar ways, the
semantics are quite different -- NET_LOCK_GIANT() and NET_UNLOCK_GIANT()
directly wrap mutex lock and unlock operations, whereas drop/pickup
special case the handling of Giant recursion.  Add a comment saying
as much.

Add NET_ASSERT_GIANT(), which conditionally asserts Giant based
on the value of debug_mpsafenet.

20 years agoMove the "fast" but unused eventhandlers to pasture. They're in the
phk [Mon, 1 Mar 2004 22:36:58 +0000 (22:36 +0000)]
Move the "fast" but unused eventhandlers to pasture.  They're in the
CVS repo should we ever need them.

Objected to by:

20 years agoAdd command line option for chrootdir, all pkg_add(8) operations will be
kensmith [Mon, 1 Mar 2004 21:58:21 +0000 (21:58 +0000)]
Add command line option for chrootdir, all pkg_add(8) operations will be
done inside of chroot(2) to chrootdir.  Added to help with sysinstall(8)
support of install to alternate root but possibly useful for setting up
jails, etc.

No objection from: portmgr@
Style(9) abuse due to: entire program violates style(9)
Approved by: rwatson (mentor)

20 years agoThis is an initial whitespace cleanup commit, new content to follow.
kensmith [Mon, 1 Mar 2004 21:50:49 +0000 (21:50 +0000)]
This is an initial whitespace cleanup commit, new content to follow.
Adding hard line breaks at the end of sentences.

Approved by: rwatson (mentor)

20 years agoChange another pointer name that was missed in the previous commit.
scottl [Mon, 1 Mar 2004 21:45:49 +0000 (21:45 +0000)]
Change another pointer name that was missed in the previous commit.

Spotted by: njl