]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
14 years agorwlock implemented from libthr need to fall through the 'hard path' and
Attilio Rao [Wed, 23 Sep 2009 21:38:57 +0000 (21:38 +0000)]
rwlock implemented from libthr need to fall through the 'hard path' and
query umtx also if the shared waiters bit is set on a shared lock.
The writer starvation avoidance technique, infact, can lead to shared
waiters on a shared lock which can bring to a missed wakeup and thus
to a deadlock if the right bit is not checked (a notable case is the
writers counterpart to be handled through expired timeouts).

Fix that by checking for the shared waiters bit also when unlocking the
shared locks.

That bug was causing a reported MySQL deadlock.
Many thanks go to Nick Esborn and his employer DesertNet which provided
time and machines to identify and fix this issue.

PR: thread/135673
Reported by: Nick Esborn <nick at desert dot net>
Tested by: Nick Esborn <nick at desert dot net>
Reviewed by: jeff

14 years agoMove sys/dev/x86bios to sys/compat/x86bios.
Jung-uk Kim [Wed, 23 Sep 2009 20:49:14 +0000 (20:49 +0000)]
Move sys/dev/x86bios to sys/compat/x86bios.

It may not be optimal but it is clearly better than the old place.

OK'ed by: delphij, paradox (ddkprog yahoo com)

14 years ago- Use spin lock instead of default mutex for safety. INT/IRET instructions
Jung-uk Kim [Wed, 23 Sep 2009 20:13:36 +0000 (20:13 +0000)]
- Use spin lock instead of default mutex for safety.  INT/IRET instructions
save/clear/restore flags but emulated flags have no effect on the host.
I believe BIOS writers never meant to run their code in emulated
environment with interrupt enabled. :-)
- Use memcpy(3) instead of copying individual members.  I believe struct
x86regs was intentionally copied from the first half of struct x86emu_regs
for this very purpose.
- Fix some style nits and consistencies.

Reviewed by: delphij, paradox (ddkprog yahoo com)

14 years agoReplace 'can not' with 'cannot' so that we use 'cannot' everywhere.
Rui Paulo [Wed, 23 Sep 2009 17:48:16 +0000 (17:48 +0000)]
Replace 'can not' with 'cannot' so that we use 'cannot' everywhere.

14 years agoExtract the code to find and map the MADT ACPI table during early kernel
John Baldwin [Wed, 23 Sep 2009 15:42:35 +0000 (15:42 +0000)]
Extract the code to find and map the MADT ACPI table during early kernel
startup and genericize it so it can be reused to map other tables as well:
- Add a routine to walk a list of ACPI subtables such as those used in the
  APIC and SRAT tables in the MI acpi(4) driver.
- Move the routines for mapping and unmapping an ACPI table as well as
  mapping the RSDT or XSDT and searching for a table with a given signature
  out into acpica_machdep.c for both amd64 and i386.

14 years agoUninline an instance of STAILQ_FOREACH_SAFE().
John Baldwin [Wed, 23 Sep 2009 15:39:54 +0000 (15:39 +0000)]
Uninline an instance of STAILQ_FOREACH_SAFE().

14 years agoUse %zu for size_t, not %zd.
Ed Maste [Wed, 23 Sep 2009 15:32:59 +0000 (15:32 +0000)]
Use %zu for size_t, not %zd.

Submitted by: ru
MFC after: 1 week

14 years agoAdd cross-filesystem regression tests for ACLs.
Edward Tomasz Napierala [Wed, 23 Sep 2009 15:12:20 +0000 (15:12 +0000)]
Add cross-filesystem regression tests for ACLs.

14 years agoIn VOP_SETACL(9) and VOP_GETACL(9), specifying wrong ACL type should result
Edward Tomasz Napierala [Wed, 23 Sep 2009 15:09:34 +0000 (15:09 +0000)]
In VOP_SETACL(9) and VOP_GETACL(9), specifying wrong ACL type should result
in EINVAL, not EOPNOTSUPP.

14 years agoAdd ACL fuzzer. It's not used by the regression tests right now,
Edward Tomasz Napierala [Wed, 23 Sep 2009 15:06:51 +0000 (15:06 +0000)]
Add ACL fuzzer.  It's not used by the regression tests right now,
but I'd prefert to have it here, so it won't get lost.

14 years agoImprove on r197334:
Max Laier [Wed, 23 Sep 2009 14:48:13 +0000 (14:48 +0000)]
Improve on r197334:
Find the most recently merged svn revision, too. If we get a svn revision
that matches HEAD use rXXX=GIT otherwise use rXXX+GIT.

Submitted by: avg
MFC after: 3 days
X-MFC: not stable/8 before 8.0

14 years ago- Remove a bogus test: setsockopt() doesn't return a length, getsockopt()
John Baldwin [Wed, 23 Sep 2009 14:23:09 +0000 (14:23 +0000)]
- Remove a bogus test: setsockopt() doesn't return a length, getsockopt()
  does.
- Use %z to printf a size_t to fix compile on 64-bit platforms.

14 years ago- Bump __FreeBSD_version to reflect the point when EVFILT_USER kevent filter
Stanislav Sedov [Wed, 23 Sep 2009 12:33:32 +0000 (12:33 +0000)]
- Bump __FreeBSD_version to reflect the point when EVFILT_USER kevent filter
  has been implemented.

14 years agoDocument osrel node for procfs.
Konstantin Belousov [Wed, 23 Sep 2009 12:10:13 +0000 (12:10 +0000)]
Document osrel node for procfs.

Approved by: des (procfs maintainer)
MFC after: 3 weeks

14 years agoAdd per-process osrel node to the procfs, to allow read and set p_osrel
Konstantin Belousov [Wed, 23 Sep 2009 12:08:08 +0000 (12:08 +0000)]
Add per-process osrel node to the procfs, to allow read and set p_osrel
value for the process.

Approved by: des (procfs maintainer)
MFC after: 3 weeks

14 years agoRestore BSD behaviour - when creating new directory entry use parent directory
Pawel Jakub Dawidek [Wed, 23 Sep 2009 09:18:16 +0000 (09:18 +0000)]
Restore BSD behaviour - when creating new directory entry use parent directory
gid to set group ownership and not process gid.

This was overlooked during v6 -> v13 switch.

PR: kern/139076
Reported by: Sean Winn <sean@gothic.net.au>
MFC after: 3 days

14 years agoIn netstat -x, do not try to print out tcp timer status for udp sockets.
Mike Silbersack [Wed, 23 Sep 2009 05:32:33 +0000 (05:32 +0000)]
In netstat -x, do not try to print out tcp timer status for udp sockets.

14 years agoInitialize registers to zero before calling the interrupt handlers inside
Xin LI [Wed, 23 Sep 2009 02:45:02 +0000 (02:45 +0000)]
Initialize registers to zero before calling the interrupt handlers inside
emulator.  This fixes VESA related freeze observed on some systems.

Submitted by: paradox <ddkprog yahoo com>

14 years agoMFp4:
Alexander Motin [Tue, 22 Sep 2009 22:23:52 +0000 (22:23 +0000)]
MFp4:
If on sense request device returns no sence, give up and return,
or we may loop forever.

14 years agoLock bus scan.
Alexander Motin [Tue, 22 Sep 2009 22:20:43 +0000 (22:20 +0000)]
Lock bus scan.

14 years agoMFp4:
Alexander Motin [Tue, 22 Sep 2009 22:17:14 +0000 (22:17 +0000)]
MFp4:
Reduce code duplication.

14 years agoTemporarily disable typematic retrieving code until we get a real fix,
Xin LI [Tue, 22 Sep 2009 20:12:10 +0000 (20:12 +0000)]
Temporarily disable typematic retrieving code until we get a real fix,
which currently causes hangs in some configurations.

Reported by: joel
Submitted by: swell.k at gmail.com

14 years agoThe elements in the component arrays may be direct Package objects rather
John Baldwin [Tue, 22 Sep 2009 18:34:23 +0000 (18:34 +0000)]
The elements in the component arrays may be direct Package objects rather
than references to objects.  In that case, simply use the Package directly.

I think a recent change to ACPI-CA is causing the interpreter to
automatically expand these references.

Reported by: Olivier Smedts  olivier gid0 org
MFC after: 3 days

14 years agoNote the D3.03 mesh changes.
Rui Paulo [Tue, 22 Sep 2009 18:19:18 +0000 (18:19 +0000)]
Note the D3.03 mesh changes.

MFC after: 1 week

14 years agoUpdate 802.11s mesh support to draft 3.03. This includes a revised frame
Rui Paulo [Tue, 22 Sep 2009 18:18:14 +0000 (18:18 +0000)]
Update 802.11s mesh support to draft 3.03. This includes a revised frame
format for peering and changes to the PERR frames.
Note that this is incompatible with the previous code.

Reviewed by: sam
MFC after: 1 week

14 years agoReserve events for XScale.
Rui Paulo [Tue, 22 Sep 2009 17:45:28 +0000 (17:45 +0000)]
Reserve events for XScale.

Reviewed by: jkoshy, gnn
MFC after: 1 week

14 years agoRemoved unused empty directory.
Xin LI [Tue, 22 Sep 2009 16:59:15 +0000 (16:59 +0000)]
Removed unused empty directory.

Submitted by: b. f. <bf1783 googlemail com>

14 years ago- Split the logic to parse an SMAP entry out into a separate function on
John Baldwin [Tue, 22 Sep 2009 16:51:00 +0000 (16:51 +0000)]
- Split the logic to parse an SMAP entry out into a separate function on
  amd64 similar to i386.  This fixes a bug on amd64 where overlapping
  entries would not cause the SMAP parsing to stop.
- Change the SMAP parsing code to do a sorted insertion into physmap[]
  instead of an append to support systems with out-of-order SMAP entries.

PR: amd64/138220
Reported by: James R. Van Artsdalen  james of jrv org
MFC after: 3 days

14 years agoChange the command argument to ioctl routines to u_long to avoid
Roman Divacky [Tue, 22 Sep 2009 16:28:07 +0000 (16:28 +0000)]
Change the command argument to ioctl routines to u_long to avoid
truncating the command to 32bit on 64bit archs where int is 32bit
(ie. amd64).

Approved by: scottl
Approved by: ed (mentor, implicit)

14 years agoDon't build ufs_gjournal.c at all if UFS_GJOURNAL option is not given
Roman Divacky [Tue, 22 Sep 2009 16:22:05 +0000 (16:22 +0000)]
Don't build ufs_gjournal.c at all if UFS_GJOURNAL option is not given
instead of building an almost empty C file.

Approved by: pjd
Approved by: ed (mentor, implicit)

14 years agoChange unsigned foo to u_foo as required by style(9).
Roman Divacky [Tue, 22 Sep 2009 16:16:02 +0000 (16:16 +0000)]
Change unsigned foo to u_foo as required by style(9).

Requested by: bde
Approved by: ed (mentor)

14 years agoDon't reread the command register to see if enabling I/O or memory
John Baldwin [Tue, 22 Sep 2009 15:43:03 +0000 (15:43 +0000)]
Don't reread the command register to see if enabling I/O or memory
decoding "took".  Other OS's that I checked do not do this and it breaks
some amdpm(4) devices.  Prior to 7.2 we did not honor the error returned
when this failed anyway, so this in effect restores previous behavior.

PR: kern/137668
Tested by: Aurelien Mere  aurelien.mere  amc-os.com
MFC after: 3 days

14 years agoAdd pieces of infrastructure required for NFSv4 ACL support in UFS.
Edward Tomasz Napierala [Tue, 22 Sep 2009 15:15:03 +0000 (15:15 +0000)]
Add pieces of infrastructure required for NFSv4 ACL support in UFS.

Reviewed by: rwatson

14 years agoMove es137x.c and es137x.h to a 2-clause BSD license. Also move a few
Joel Dahl [Tue, 22 Sep 2009 13:23:59 +0000 (13:23 +0000)]
Move es137x.c and es137x.h to a 2-clause BSD license.  Also move a few
comments.

Submitted by: Joachim Kuebart
Approved by: core, Russell Cattelan <cattelan@thebarn.com>

14 years agoThe ndis_scan function may be started after ndis vap have been destroyed
Coleman Kane [Tue, 22 Sep 2009 12:36:51 +0000 (12:36 +0000)]
The ndis_scan function may be started after ndis vap have been destroyed

PR: kern/138632
Submitted by: Paul B. Mahol <onemda at gmail.com>
MFC after: 3 days

14 years ago- Add missing bus_dmamap_sync(9) calls for the work DMA map. Previously
Marius Strobl [Tue, 22 Sep 2009 11:47:21 +0000 (11:47 +0000)]
- Add missing bus_dmamap_sync(9) calls for the work DMA map. Previously
  the work area was totally unsynchronized which means this driver only
  had a chance of working on x86 when no bounce buffers were involved,
  which isn't that likely given that support for 64-bit DMA is currently
  broken throughout ata(4).
- Add necessary little-endian conversion of accesses to the work area,
  making this driver work on big-endian hosts. While at it, use the
  alignment-agnostic byte order encoders in order to be on the safe side.
- Clear the reserved member of the SG list entries in order to be on the
  safe side. [1]

Submitted by: yongari [1]
Reviewed by: yongari
MFC after: 3 days

14 years ago- According to Linux, the ALi M5451 can do 31-bit DMA instead of just
Marius Strobl [Tue, 22 Sep 2009 11:38:45 +0000 (11:38 +0000)]
- According to Linux, the ALi M5451 can do 31-bit DMA instead of just
  30-bit like the reset of the controllers supported by this driver.
  Actually ALi M5451 can be setup up to generate 32-bit addresses by
  setting the 31st bit via the accompanying ISA bridge, which allows
  it to work in sparc64 machines whose IOMMU require at least 32-bit
  DMA. Even though other architectures would also benefit from 32-bit
  DMA, enabling this bit is limited to sparc64 as bus_dma(9) doesn't
  generally guarantee that a low address of BUS_SPACE_MAXADDR_32BIT
  results in a buffer in the 32-bit range.
- According to Tatsuo YOKOGAWA's ali(4), the the DMA transfer size of
  ALi M5451 is fixed to 64k and in fact using the default size of 4k
- The 4DWAVE DX and NX require the recording buffer to be 8-byte
  aligned so adjust the bus_dma_tag_create(9) accordingly.
- Unlike the rest of the controllers supported by this driver, the
  ALi M5451 only has 32 hardware channels instead of 64 so limit the
  loop in tr_intr() accordingly. [1]

Submitted by: yongari [1]
Reviewed by: yongari (superset of what is committed)
MFC after: 3 days

14 years agoUse an unsigned integer for storing the key code.
Ed Schouten [Tue, 22 Sep 2009 11:29:11 +0000 (11:29 +0000)]
Use an unsigned integer for storing the key code.

It seems Clang breaks when checking for SPCLKEY, which is now
0x80000000. Using an unsigned integer fixes this. This is also
consistent with other pieces of kbd/syscons code, because these also use
u_int.

Submitted by: rdivacky

14 years agoHide x86bios stuff in i386/amd64 specific files as atkbdc would get
Xin LI [Tue, 22 Sep 2009 07:10:23 +0000 (07:10 +0000)]
Hide x86bios stuff in i386/amd64 specific files as atkbdc would get
these stuff into build.

14 years agoBuild x86bios only for i386/amd64 for now. More work is required
Xin LI [Mon, 21 Sep 2009 23:58:29 +0000 (23:58 +0000)]
Build x86bios only for i386/amd64 for now.  More work is required
to make these functional on other architectures, and the current
code breaks sparc64 and powerpc.

Spotted by: tinderbox via des

14 years agoImprove mxge watchdog routine's ability to reliably reset a failed NIC:
Andrew Gallatin [Mon, 21 Sep 2009 20:16:10 +0000 (20:16 +0000)]
Improve mxge watchdog routine's ability to reliably reset a failed NIC:

- Mark the link as down, so if watchdog reset fails, link watching
    failover software can notice it
- Don't send MXGEFW_CMD_ETHERNET_DOWN if the NIC has been reset, it is
    not needed, and will fail on a freshly reset NIC.
- Ensure the transmit routines aren't attempting to PIO write to doorbells
    while the NIC is being reset.
- Download the correct f/w, rather than using the EEPROM f/w after reset.
- Export a count of the number of watchdog resets via sysctl
- Zero all f/w stats at reset.  This will lead to less confusing
    diagnostic output when investigating NIC failures.

MFC after: 3 days

14 years agoCorrect a section name; specifically, "AUTHOR" should be "AUTHORS". [1]
Alan Cox [Mon, 21 Sep 2009 18:02:02 +0000 (18:02 +0000)]
Correct a section name; specifically, "AUTHOR" should be "AUTHORS". [1]

Turn the "AUTHORS" section into a full sentence.

Pointed out by: ru [1]

14 years agoAdd FreeBSD 7.2 and 7.3.
Alan Cox [Mon, 21 Sep 2009 17:19:36 +0000 (17:19 +0000)]
Add FreeBSD 7.2 and 7.3.

Discussed with: ru

14 years ago- List newsletters VI-2, VI-3 and VI-4 that were already applied.
Ruslan Ermilov [Mon, 21 Sep 2009 14:43:05 +0000 (14:43 +0000)]
- List newsletters VI-2, VI-3 and VI-4 that were already applied.
- Apply newsletters VI-5 (BO's name change) and VI-6 (VE's name change).

14 years agoAdd support for throttling transmit bandwidth. This is most commonly
Andrew Gallatin [Mon, 21 Sep 2009 14:41:07 +0000 (14:41 +0000)]
Add support for throttling transmit bandwidth.  This is most commonly
used to reduce packet loss on high delay (WAN) paths with a
slow link.

14 years agoRemove forward_roundrobin(), it is unused for quite some time.
Konstantin Belousov [Mon, 21 Sep 2009 13:09:56 +0000 (13:09 +0000)]
Remove forward_roundrobin(), it is unused for quite some time.

Reviewed by: jhb
MFC after: 1 week

14 years agoIf CPU happens to be in usermode when a T_RESERVED trap occured,
Konstantin Belousov [Mon, 21 Sep 2009 09:41:51 +0000 (09:41 +0000)]
If CPU happens to be in usermode when a T_RESERVED trap occured,
then trapsignal is called with ksi.ksi_signo = 0. For debugging kernels,
that should end up in panic, for non-debugging kernels behaviour is
undefined.

Do panic regardeless of execution mode at the moment of trap.

Reviewed by: jhb
MFC after: 1 month

14 years agoTo be safe, make this module initialize earlier by specifying SI_SUB_CPU
Xin LI [Mon, 21 Sep 2009 09:09:55 +0000 (09:09 +0000)]
To be safe, make this module initialize earlier by specifying SI_SUB_CPU
instead of SI_SUB_KLD.  Note that we don't believe that there is any
consumers of x86bios that would be triggered in the early stage of boot,
so this is more or less just a safebelt.

Submitted by: swell.k at gmail.com

14 years agoStyle(9) fixes.
Xin LI [Mon, 21 Sep 2009 08:53:26 +0000 (08:53 +0000)]
Style(9) fixes.

Submitted by: swell.k at gmail.com

14 years agoUse __FBSDID to embed RCS ID.
Xin LI [Mon, 21 Sep 2009 08:37:57 +0000 (08:37 +0000)]
Use __FBSDID to embed RCS ID.

14 years agoFixed markup (mostly) errors.
Ruslan Ermilov [Mon, 21 Sep 2009 08:30:52 +0000 (08:30 +0000)]
Fixed markup (mostly) errors.

14 years agoAllow atkbd to obtain keyboard repeat rate from BIOS on amd64.
Xin LI [Mon, 21 Sep 2009 08:24:22 +0000 (08:24 +0000)]
Allow atkbd to obtain keyboard repeat rate from BIOS on amd64.

Submitted by: swell.k at gmail.com

14 years agoCollapase interrupt supporting functions to a new module, and switch from
Xin LI [Mon, 21 Sep 2009 08:17:57 +0000 (08:17 +0000)]
Collapase interrupt supporting functions to a new module, and switch from
x86emu to this new module.

This changeset also brings a fix for bugs introduced with the initial
x86emu commit, which prevents the user from using some display mode or
cause instant reboots during mode switch.

Submitted by: paradox <ddkprog yahoo com>

14 years agoRevert part that should not be in my previous commit.
Xin LI [Mon, 21 Sep 2009 07:38:50 +0000 (07:38 +0000)]
Revert part that should not be in my previous commit.

Pointy hat to: delphij

14 years agoAutomatically depend on x86emu when vesa or dpms is being built into
Xin LI [Mon, 21 Sep 2009 07:08:20 +0000 (07:08 +0000)]
Automatically depend on x86emu when vesa or dpms is being built into
kernel.  With this change the user no longer need to remember building
this option.

Submitted by: swell.k at gmail.com

14 years agoEnable s3pci on amd64 which works on top of VESA, and allow
Xin LI [Mon, 21 Sep 2009 07:05:48 +0000 (07:05 +0000)]
Enable s3pci on amd64 which works on top of VESA, and allow
static building it into kernel on i386 and amd64.

Submitted by: swell.k at gmail.com

14 years agoDisable a check on a disk size because it's too strict. This change is
Yoshihiro Takahashi [Mon, 21 Sep 2009 04:04:02 +0000 (04:04 +0000)]
Disable a check on a disk size because it's too strict.  This change is
to avoid using incorrect geometry.

It seems that this is the same problem in g_part_bsd_read()@g_part_bsd.c.

Reviewed by: rink
MFC after: 3 days

14 years ago(semiforced commit to add comment missed in last delta)
Matt Jacob [Mon, 21 Sep 2009 01:41:19 +0000 (01:41 +0000)]
(semiforced commit to add comment missed in last delta)
Add a maximum response length for FCP RSPNS IUs.

Clarify some of the FC option words for setting parameters
and try and disable automatic PRLI when in target mode- this
should correct some cases of N-port topologies with 23XX cards
where we put out an illegal PRLI (in target mode only we're
not supposed to put out a PRLI).

14 years agoRemove file unused in freebsd.
Matt Jacob [Mon, 21 Sep 2009 01:38:22 +0000 (01:38 +0000)]
Remove file unused in freebsd.

14 years agoMention that NUL characters are not allowed in sh(1) input.
Jilles Tjoelker [Sun, 20 Sep 2009 21:42:38 +0000 (21:42 +0000)]
Mention that NUL characters are not allowed in sh(1) input.

I do not consider this a bug because POSIX permits it and argument strings
and environment variables cannot contain '\0' anyway.

PR: bin/25542
MFC after: 2 weeks

14 years ago- Update the list of known-working machines based on feedback for 7.2.
Marius Strobl [Sun, 20 Sep 2009 19:14:34 +0000 (19:14 +0000)]
- Update the list of known-working machines based on feedback for 7.2.
- Update the V440 entry regarding added support for the on-board NICs
  in 8.0.

Approved by: blackend
MFC after: 3 days

14 years agoA wrong variable is used when setting up the interface
Qing Li [Sun, 20 Sep 2009 17:22:19 +0000 (17:22 +0000)]
A wrong variable is used when setting up the interface
address route, which broke source address selection in
some code paths.

Submitted by: noted by bz
Reviewed by: hrs
MFC after: immediately

14 years agoUpdate find(1) man page for -L/-delete interaction.
Jilles Tjoelker [Sun, 20 Sep 2009 16:47:56 +0000 (16:47 +0000)]
Update find(1) man page for -L/-delete interaction.

It is a bit unfortunate that the example to delete broken symlinks now uses
rm(1), but allowing this with -delete would require fixing fts(3) to not
imply FTS_NOCHDIR if FTS_LOGICAL is given (or hacks in the -delete option).

PR: bin/90687
MFC after: 2 weeks

14 years agoIEEE Std 1003.1, 2004 Edition states:
Diomidis Spinellis [Sun, 20 Sep 2009 15:47:31 +0000 (15:47 +0000)]
IEEE Std 1003.1, 2004 Edition states:

"The escape sequence '\n' shall match a <newline> embedded in
the pattern space."

It is unclear whether this also applies to a \n embedded in a
character class.  Disable the existing handling of \n in a character
class following Mac OS X, GNU sed version 4.1.5 with --posix, and
SunOS 5.10 /usr/bin/sed.

Pointed by: Marius Strobl
Obtained from: Mac OS X

14 years agoFollow POSIX (IEEE Std 1003.1, 2004 Edition) in the implementation
Diomidis Spinellis [Sun, 20 Sep 2009 15:17:40 +0000 (15:17 +0000)]
Follow POSIX (IEEE Std 1003.1, 2004 Edition) in the implementation
of the y (translate) command.

"If a backslash character is immediately followed by a backslash
character in string1 or string2, the two backslash characters shall
be counted as a single literal backslash character"

Pointed by: Marius Strobl
Obtained from: Mac OS X

14 years agoAdd correct test results.
Diomidis Spinellis [Sun, 20 Sep 2009 15:04:50 +0000 (15:04 +0000)]
Add correct test results.

14 years agoDescribe how other systems treat this case.
Diomidis Spinellis [Sun, 20 Sep 2009 14:20:00 +0000 (14:20 +0000)]
Describe how other systems treat this case.

14 years agoAllow [ to be used as a delimiter.
Diomidis Spinellis [Sun, 20 Sep 2009 14:11:33 +0000 (14:11 +0000)]
Allow [ to be used as a delimiter.

Pointed by: Marius Strobl
Obtained from: Apple

14 years agoThe transition to Subversion allows us to rename files without
Diomidis Spinellis [Sun, 20 Sep 2009 13:39:44 +0000 (13:39 +0000)]
The transition to Subversion allows us to rename files without
repo-copy hacks.
Remove the test-number prefix from the name of the output files,
so that new test cases can be easily added.

14 years agoPurge namecache in the same place OpenSolaris does.
Pawel Jakub Dawidek [Sun, 20 Sep 2009 13:28:29 +0000 (13:28 +0000)]
Purge namecache in the same place OpenSolaris does.

14 years agoRe-remove the IBM0057 ID used for PS/2 mouse controllers. The asl for the
John Baldwin [Sun, 20 Sep 2009 13:28:19 +0000 (13:28 +0000)]
Re-remove the IBM0057 ID used for PS/2 mouse controllers.  The asl for the
61p includes the hotkey device as IBM0068 and the mouse as IBM0057 similar
to other systems.

Submitted by: marius
MFC after: 3 days

14 years agoOld (a.out) rtld attempts to mmap zero-length region, e.g. when bss
Konstantin Belousov [Sun, 20 Sep 2009 12:40:56 +0000 (12:40 +0000)]
Old (a.out) rtld attempts to mmap zero-length region, e.g. when bss
of the linked object is zero-length. More old code assumes that mmap
of zero length returns success.

For a.out and pre-8 ELF binaries, allow the mmap of zero length.

Reported by: tegge
Reviewed by: tegge, alc, jhb
MFC after: 3 days

14 years agoFix handling of sctp_drain().
Michael Tuexen [Sun, 20 Sep 2009 11:33:39 +0000 (11:33 +0000)]
Fix handling of sctp_drain().

Approved by: rrs (mentor)
MFC after: 2 month

14 years agoFix errnos.
Michael Tuexen [Sun, 20 Sep 2009 11:32:22 +0000 (11:32 +0000)]
Fix errnos.

Approved by: rrs(mentor)
MFC after: 3 days.

14 years agoExtract version information from git repositories. Try to find a
Max Laier [Sat, 19 Sep 2009 21:46:12 +0000 (21:46 +0000)]
Extract version information from git repositories. Try to find a
corresponding svn version if this is a git-svn clone'ed repo.
This prefers svn version information over git in case a working
directory has both .svn and .git directories.

MFC after: 3 days
X-MFC: not stable/8 before 8.0

14 years agoTell about the full name we opened, not the unit.
Matt Jacob [Sat, 19 Sep 2009 20:36:43 +0000 (20:36 +0000)]
Tell about the full name we opened, not the unit.

Obtained from: Sean Bruno
MFC after: 1 month

14 years agoRemember to unlock the peripheral prior to notifying the user. Make some
Matt Jacob [Sat, 19 Sep 2009 20:25:54 +0000 (20:25 +0000)]
Remember to unlock the peripheral prior to notifying the user. Make some
allocations M_NOWAIT so that we don't try and sleep with a nested non-sleepable
lock.

This makes the userland scsi_target begin to function again.

Obtained from: Sean Bruno
MFC after: 1 month

14 years agoAdd getpagesizes(3). This functions either the number of supported page
Alan Cox [Sat, 19 Sep 2009 18:01:32 +0000 (18:01 +0000)]
Add getpagesizes(3).  This functions either the number of supported page
sizes or some number of the sizes themselves.  It is functionally
compatible with a function by the same name under Solaris.

Reviewed by: jhb

14 years agoMake the keyboard layer Unicode aware.
Ed Schouten [Sat, 19 Sep 2009 17:56:26 +0000 (17:56 +0000)]
Make the keyboard layer Unicode aware.

Just take keyent_t to use an u_int to store the Unicode codepoints.
Unfortunately the keymap is now too big to be loaded using an ioctl
argument, so change the ioctl to pick a pointer.

This change breaks kbdcontrol ABI. It doesn't break X11, because X11
doesn't do anything with syscons keymaps. It just switches the device
out of K_XLATE.

Obtained from: //depot/user/ed/newcons/...

14 years agoUse appropriate locking when using interface list.
Michael Tuexen [Sat, 19 Sep 2009 14:55:12 +0000 (14:55 +0000)]
Use appropriate locking when using interface list.

Approved by: rrs (mentor)
MFC after: 1 month.

14 years agoFix the disabling of sctp_drain().
Michael Tuexen [Sat, 19 Sep 2009 14:18:42 +0000 (14:18 +0000)]
Fix the disabling of sctp_drain().

Approved by: rrs (mentor)
MFC after: 1 month.

14 years agoGet SCTP working in combination with VIMAGE.
Michael Tuexen [Sat, 19 Sep 2009 14:02:16 +0000 (14:02 +0000)]
Get SCTP working in combination with VIMAGE.
Contains code from bz.
Approved by: rrs (mentor)
MFC after: 1 month.

14 years agointpm/sb700: force polling mode if configured interrupt is SMI
Andriy Gapon [Sat, 19 Sep 2009 08:56:28 +0000 (08:56 +0000)]
intpm/sb700: force polling mode if configured interrupt is SMI

instead of failing to attach

14 years agoChange cache attribute for VESA frame buffer from UC to write-combining.
Jung-uk Kim [Sat, 19 Sep 2009 04:36:38 +0000 (04:36 +0000)]
Change cache attribute for VESA frame buffer from UC to write-combining.
This improves scrolling speed for high resolution graphics mode console.

14 years agoCorrect BIOS header sanitizing on pc98.
Yoshihiro Takahashi [Sat, 19 Sep 2009 03:33:18 +0000 (03:33 +0000)]
Correct BIOS header sanitizing on pc98.

MFC after: 3 days

14 years agoWhen superpages are enabled, add the 2 or 4MB page size to the array of
Alan Cox [Fri, 18 Sep 2009 17:09:33 +0000 (17:09 +0000)]
When superpages are enabled, add the 2 or 4MB page size to the array of
supported page sizes.

Reviewed by: jhb
MFC after: 3 weeks

14 years agoAdd a new sysctl for reporting all of the supported page sizes.
Alan Cox [Fri, 18 Sep 2009 17:04:57 +0000 (17:04 +0000)]
Add a new sysctl for reporting all of the supported page sizes.

Reviewed by: jhb
MFC after: 3 weeks

14 years agoJust use ttydisc_rint_simple() instead of doing it ourselves.
Ed Schouten [Fri, 18 Sep 2009 15:39:09 +0000 (15:39 +0000)]
Just use ttydisc_rint_simple() instead of doing it ourselves.

This code seems to do exactly the same as ttydisc_rint_simple() does
nowadays. Just remove it.

Obtained from: //depot/user/ed/newcons/sys/dev/syscons/syscons.c

14 years agoReturn ENOBUFS consistently if user attempts to exceed
Bruce M Simpson [Fri, 18 Sep 2009 15:12:31 +0000 (15:12 +0000)]
Return ENOBUFS consistently if user attempts to exceed
in_mcast_maxsocksrc resource limit.

Submitted by: syrinx
MFC after: 3 days

14 years agoHave sparc64 join the separate livefs crowd. Including the docs packages
Ken Smith [Fri, 18 Sep 2009 14:53:01 +0000 (14:53 +0000)]
Have sparc64 join the separate livefs crowd.  Including the docs packages
on disc1 (needed now by sysinstall) overflows the target CDROM iso size.

MFC after: 3 days

14 years agoFix setfib(1) section number.
Christian Brueffer [Fri, 18 Sep 2009 14:17:00 +0000 (14:17 +0000)]
Fix setfib(1) section number.

PR: 133765
Submitted by: Konstantin Zolotukhin <erebus@gorodok.net>
MFC after: 3 days

14 years agoFix mdoc, typos, contractions.
Christian Brueffer [Fri, 18 Sep 2009 14:05:56 +0000 (14:05 +0000)]
Fix mdoc, typos, contractions.

This includes:
PR: 135520
Submitted by: Nobuyuki Koganemaru
Patch by: gavin
MFC after: 3 days

14 years agoDon't allocate new unnecessary pages when devstat_alloc() looses the
Attilio Rao [Fri, 18 Sep 2009 13:48:38 +0000 (13:48 +0000)]
Don't allocate new unnecessary pages when devstat_alloc() looses the
run for re-acuiring the lock, but recheck if new pages are allocatable
from the pool and free the previously allocated ones.

Tested by: pho, Giovanni Trematerra
<giovanni dot trematerra at gmail dot com>

14 years agoSpell Israel correctly.
Ed Schouten [Fri, 18 Sep 2009 13:44:39 +0000 (13:44 +0000)]
Spell Israel correctly.

Submitted by: Alexey Savartsov <asavartsov gmail com>
PR: bin/138580

14 years agoCall mwl_rxdma_cleanup() if there was an error setting up the RX DMA.
Rui Paulo [Fri, 18 Sep 2009 12:25:31 +0000 (12:25 +0000)]
Call mwl_rxdma_cleanup() if there was an error setting up the RX DMA.

14 years agoStyle fix - break too long a line in two.
Marko Zec [Fri, 18 Sep 2009 09:03:23 +0000 (09:03 +0000)]
Style fix - break too long a line in two.

Spotted by: bz
MFC after: 3 days

14 years agoo DragonflyBSD 2.4.0 added.
Maxim Konovalov [Fri, 18 Sep 2009 07:07:34 +0000 (07:07 +0000)]
o DragonflyBSD 2.4.0 added.

14 years agoVarious mdoc, spelling etc fixes.
Christian Brueffer [Fri, 18 Sep 2009 00:33:47 +0000 (00:33 +0000)]
Various mdoc, spelling etc fixes.

MFC after: 3 days

14 years agoFixed markup.
Ruslan Ermilov [Thu, 17 Sep 2009 19:15:49 +0000 (19:15 +0000)]
Fixed markup.

14 years agoChange the default transport protocol for use by the Mount protocol
Rick Macklem [Thu, 17 Sep 2009 19:08:15 +0000 (19:08 +0000)]
Change the default transport protocol for use by the Mount protocol
from UDP to TCP, so that it is consistent with TCP for NFS, which
became the default at r176198. Without this change, doing an NFS mount
against a server that only supports UDP would result in an unusable
mount point if a transport protocol option wasn't specified for the
mount.

Approved by: kib (mentor)
MFC after: 3 days