]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
13 years agoRe-encode files from ISO-8859-1 to UTF-8
uqs [Sun, 22 May 2011 14:03:30 +0000 (14:03 +0000)]
Re-encode files from ISO-8859-1 to UTF-8

13 years agoRemove typos, tabs-after-spaces and EOL whitespace. Convert to UTF-8.
uqs [Sun, 22 May 2011 14:03:21 +0000 (14:03 +0000)]
Remove typos, tabs-after-spaces and EOL whitespace. Convert to UTF-8.

13 years agosh: Add test for r222173.
jilles [Sun, 22 May 2011 12:15:14 +0000 (12:15 +0000)]
sh: Add test for r222173.

13 years agosh: Fix bss-based buffer overflow in . builtin.
jilles [Sun, 22 May 2011 12:12:28 +0000 (12:12 +0000)]
sh: Fix bss-based buffer overflow in . builtin.

If the length of a directory in PATH together with the given filename
exceeded FILENAME_MAX (which may happen even for pathnames that work), a
static buffer was overflown.

The static buffer is unnecessary, we can use the stalloc() stack.

Obtained from: NetBSD
MFC after: 1 week

13 years agoFix typo in unused function name
uqs [Sun, 22 May 2011 09:58:48 +0000 (09:58 +0000)]
Fix typo in unused function name

Submitted by: arundel

13 years agoUpdate URLs and fix regex to work with current version of the files.
uqs [Sun, 22 May 2011 09:48:25 +0000 (09:48 +0000)]
Update URLs and fix regex to work with current version of the files.

13 years agoAdd a lock flags argument to the VFS_FHTOVP() file system
rmacklem [Sun, 22 May 2011 01:07:54 +0000 (01:07 +0000)]
Add a lock flags argument to the VFS_FHTOVP() file system
method, so that callers can indicate the minimum vnode
locking requirement. This will allow some file systems to choose
to return a LK_SHARED locked vnode when LK_SHARED is specified
for the flags argument. This patch only adds the flag. It
does not change any file system to use it and all callers
specify LK_EXCLUSIVE, so file system semantics are not changed.

Reviewed by: kib

13 years agosh: Expand aliases after assignments and redirections.
jilles [Sat, 21 May 2011 22:03:06 +0000 (22:03 +0000)]
sh: Expand aliases after assignments and redirections.

13 years agoRecognize HIO_FLUSH requests.
pjd [Sat, 21 May 2011 20:21:20 +0000 (20:21 +0000)]
Recognize HIO_FLUSH requests.

MFC after: 1 week

13 years ago1. Prior to r214782, UMA did not support multipage allocations before
alc [Sat, 21 May 2011 17:43:43 +0000 (17:43 +0000)]
1. Prior to r214782, UMA did not support multipage allocations before
uma_startup2() was called.  Thus, setting the variable "booted" to true in
uma_startup() was ok on machines with UMA_MD_SMALL_ALLOC defined, because
any allocations made after uma_startup() but before uma_startup2() could be
satisfied by uma_small_alloc().  Now, however, some multipage allocations
are necessary before uma_startup2() just to allocate zone structures on
machines with a large number of processors.  Thus, a Boolean can no longer
effectively describe the state of the UMA allocator.  Instead, make "booted"
have three values to describe how far initialization has progressed.  This
allows multipage allocations to continue using startup_alloc() until
uma_startup2(), but single-page allocations may begin using
uma_small_alloc() after uma_startup().

2. With the aforementioned change, only a modest increase in boot pages is
necessary to boot UMA on a large number of processors.

3. Retire UMA_MD_SMALL_ALLOC_NEEDS_VM.  It has only been used between
r182028 and r204128.

Reviewed by: attilio [1], nwhitehorn [3]
Tested by: sbruno

13 years agoFix typo, it is MPDU not MDPU.
bschmidt [Sat, 21 May 2011 16:41:41 +0000 (16:41 +0000)]
Fix typo, it is MPDU not MDPU.

13 years agoFix a typo, it is MPDU not MDPU.
bschmidt [Sat, 21 May 2011 16:34:53 +0000 (16:34 +0000)]
Fix a typo, it is MPDU not MDPU.

13 years agosh: Add test for positional parameters with more than one digit.
jilles [Sat, 21 May 2011 14:52:26 +0000 (14:52 +0000)]
sh: Add test for positional parameters with more than one digit.

This also passes on stable/8.

13 years agoThe Merlin analog register bank is from 0x7800 -> 0x78fc; fix the code
adrian [Sat, 21 May 2011 09:23:18 +0000 (09:23 +0000)]
The Merlin analog register bank is from 0x7800 -> 0x78fc; fix the code
to reflect this.

13 years agosh: Implement the cd -e flag proposed for the next POSIX issue.
jilles [Fri, 20 May 2011 22:55:18 +0000 (22:55 +0000)]
sh: Implement the cd -e flag proposed for the next POSIX issue.

This reflects failure to determine the pathname of the new directory in the
exit status (1). Normally, cd returns successfully if it did chdir() and the
call was successful.

In POSIX, -e only has meaning with -P; because our -L is not entirely
compliant and may fall back to -P mode, -e has some effect with -L as well.

13 years agoAdd a new knob to atkbd(4) to enable typematic rate detection on boot,
delphij [Fri, 20 May 2011 22:36:17 +0000 (22:36 +0000)]
Add a new knob to atkbd(4) to enable typematic rate detection on boot,
which is now disabled by default.  The detection is known to cause hangs
on boot with some new Lenovo laptops on FreeBSD/amd64.

Reported by: gnn
Discussed with: jkim
MFC after: 3 months

13 years agoThe statically configured (permanent) ARP entries are removed when an
qingli [Fri, 20 May 2011 19:12:20 +0000 (19:12 +0000)]
The statically configured (permanent) ARP entries are removed when an
interface is brought down, even though the interface address is still
valid. This patch maintains the permanent ARP entries as long as the
interface address (having the same prefix as that of the ARP entries)
is valid.

Reviewed by: delphij
MFC after: 5 days

13 years agoDatasheet says vge(4) controllers support DAC but it seems that's
yongari [Fri, 20 May 2011 18:27:13 +0000 (18:27 +0000)]
Datasheet says vge(4) controllers support DAC but it seems that's
not true on old PCI based controllers.  DAC configuration is read
from EEPROM in device reset phase and driver can override DAC
configuration.  However I guess there is an undocumented reason why
EEPROM configuration does not enable DAC so do not blindly override
DAC configuration.  Recent PCIe based controllers are supposed to
support 64bit DMA so allow 64bit DMA only on PCIe based controllers.

PR: kern/157184
MFC after: 1 week

13 years ago- add missing options and arguments to program's usage()
ru [Fri, 20 May 2011 17:35:10 +0000 (17:35 +0000)]
- add missing options and arguments to program's usage()
  and sync it with manpage's SYNOPSIS

- generally clean up a manpage's formatting

Reviewed by: edwin

13 years agoFix spelling errors.
alc [Fri, 20 May 2011 17:28:00 +0000 (17:28 +0000)]
Fix spelling errors.

13 years agoAdd missing header file.
pluknet [Fri, 20 May 2011 17:03:23 +0000 (17:03 +0000)]
Add missing header file.

MFC after: 1 week

13 years agoRemove unnecessary controller reinitialization by checking
yongari [Fri, 20 May 2011 17:01:22 +0000 (17:01 +0000)]
Remove unnecessary controller reinitialization by checking
IFF_DRV_RUNNING flag.  Previously running dhclient or adding alias
addresses reinitialized controller and it resulted in unnecessary
link flips.

Reviewed by: marius

13 years agosh: Allow terminating a heredoc with a terminator at EOF without a newline.
jilles [Fri, 20 May 2011 16:03:36 +0000 (16:03 +0000)]
sh: Allow terminating a heredoc with a terminator at EOF without a newline.

This is sometimes used with eval or old-style command substitution, and most
shells other than ash derivatives allow it.

It can also be used with scripts that violate POSIX's requirement on the
application that they end in a newline (scripts must be text files except
that line length is unlimited).

Example:
v=`cat <<EOF
foo
EOF`
echo $v

This commit does not add support for the similar construct with new-style
command substitution, like
v=$(cat <<EOF
foo
EOF)
This continues to require a newline after the terminator.

13 years agoEliminate a redundant #include. ("vm/vm_param.h" already includes
alc [Fri, 20 May 2011 15:26:31 +0000 (15:26 +0000)]
Eliminate a redundant #include.  ("vm/vm_param.h" already includes
"machine/vmparam.h".)

13 years agoSpelling in P_HADTHREADS.
pluknet [Fri, 20 May 2011 14:32:28 +0000 (14:32 +0000)]
Spelling in P_HADTHREADS.

MFC after: 1 week

13 years agoFix the description of the "paddr" keyword. It shows a process
pluknet [Fri, 20 May 2011 14:07:08 +0000 (14:07 +0000)]
Fix the description of the "paddr" keyword.  It shows a process
pointer just like a dublicating "uprocp".

MFC after: 1 week

13 years agoCorrect typos in comments, no functional changes.
bcr [Fri, 20 May 2011 11:29:09 +0000 (11:29 +0000)]
Correct typos in comments, no functional changes.

Found by: codespell
Reviewed by: kaiw
MFC after: 1 week

13 years agoDocument IPv6 support.
pjd [Fri, 20 May 2011 11:21:39 +0000 (11:21 +0000)]
Document IPv6 support.

MFC after: 3 weeks

13 years agoIf no listen address is specified, bind by default to:
pjd [Fri, 20 May 2011 11:16:25 +0000 (11:16 +0000)]
If no listen address is specified, bind by default to:

tcp4://0.0.0.0:8457
tcp6://[::]:8457

MFC after: 3 weeks

13 years agoRename ipv4/ipv6 to tcp4/tcp6.
pjd [Fri, 20 May 2011 11:15:27 +0000 (11:15 +0000)]
Rename ipv4/ipv6 to tcp4/tcp6.

MFC after: 3 weeks

13 years agoNow that hell is fully frozen it is good time to add IPv6 support to HAST.
pjd [Fri, 20 May 2011 11:14:05 +0000 (11:14 +0000)]
Now that hell is fully frozen it is good time to add IPv6 support to HAST.

MFC after: 3 weeks

13 years agoAllow [ ] characters in strings. They might be used in IPv6 addresses.
pjd [Fri, 20 May 2011 11:10:39 +0000 (11:10 +0000)]
Allow [ ] characters in strings. They might be used in IPv6 addresses.

MFC after: 3 weeks

13 years agoRename tcp4 to tcp in preparation for IPv6 support.
pjd [Fri, 20 May 2011 11:09:02 +0000 (11:09 +0000)]
Rename tcp4 to tcp in preparation for IPv6 support.

MFC after: 3 weeks

13 years agoRename proto_tcp4.c to proto_tcp.c in preparation for IPv6 support.
pjd [Fri, 20 May 2011 11:06:17 +0000 (11:06 +0000)]
Rename proto_tcp4.c to proto_tcp.c in preparation for IPv6 support.

MFC after: 2 weeks

13 years agoIn preparation for IPv6 support allow to specify multiple addresses to
pjd [Thu, 19 May 2011 23:18:42 +0000 (23:18 +0000)]
In preparation for IPv6 support allow to specify multiple addresses to
listen on.

MFC after: 3 weeks

13 years agoFix typo.
yongari [Thu, 19 May 2011 23:13:08 +0000 (23:13 +0000)]
Fix typo.

Submitted by: brad at OpenBSD

13 years agoFix typos in comments, no functional changes.
bcr [Thu, 19 May 2011 20:35:40 +0000 (20:35 +0000)]
Fix typos in comments, no functional changes.

Found by:       codespell
Reviewed by:    alfred
MFC after:      1 week

13 years agoSimplify t4_os_find_pci_capability.
np [Thu, 19 May 2011 19:37:41 +0000 (19:37 +0000)]
Simplify t4_os_find_pci_capability.

MFC after: 3 days

13 years agoStyle fixes:
jhb [Thu, 19 May 2011 17:22:48 +0000 (17:22 +0000)]
Style fixes:
- Sort forward declarations of structures.
- Prefer uint64_t to u_int64_t.

13 years agoFix clang warnings.
benl [Thu, 19 May 2011 15:18:35 +0000 (15:18 +0000)]
Fix clang warnings.

Approved by: philip (mentor)

13 years agoWhen adding examples to man-pages, try to make them at least
phk [Thu, 19 May 2011 13:42:19 +0000 (13:42 +0000)]
When adding examples to man-pages, try to make them at least
look like they might work.

Prodded by: Vadim Goncharov

13 years agoPut AN back after finding out that tzsetup(1) will complain that
edwin [Thu, 19 May 2011 13:09:39 +0000 (13:09 +0000)]
Put AN back after finding out that tzsetup(1) will complain that
it doesn't exist. It will be removed again once the tzdata distribution
files have been updated with the replacements for AN.

13 years agoAdd support for the SIIG Cyber 2S PCIe adapter. It is based on an
jhb [Thu, 19 May 2011 11:41:12 +0000 (11:41 +0000)]
Add support for the SIIG Cyber 2S PCIe adapter.  It is based on an
Oxford Semiconductor OX16PCI954 but uses only two ports with a non-default
clock rate.

PR: kern/152034
Tested by: Hans Fiedler  hans of hermes louisville edu
MFC after: 1 week

13 years agoImplement WITH{,OUT}_{GCC,BINUTILS} to provide finer-grained control
imp [Thu, 19 May 2011 05:13:25 +0000 (05:13 +0000)]
Implement WITH{,OUT}_{GCC,BINUTILS} to provide finer-grained control
over building gcc and binutils.  They default to true, unless
MK_TOOLCHAIN is no.

Reviewed by: ru@

13 years ago- Add support for AF_INET6 sockets for %S format character.
pjd [Wed, 18 May 2011 22:43:56 +0000 (22:43 +0000)]
- Add support for AF_INET6 sockets for %S format character.
- Use inet_ntop(3) instead of reimplementing it.
- Use %hhu for unsigned char instead of casting it to unsigned int and
  using %u.

MFC after: 1 week

13 years agoThe CDP_ACTIVE flag is cleared at the beginning of destroy_devl(),
kib [Wed, 18 May 2011 22:36:58 +0000 (22:36 +0000)]
The CDP_ACTIVE flag is cleared at the beginning of destroy_devl(),
and destroy_devl() drops dev_mtx. The protection against the race
with dev_rel(), introduced in r163328, should be extended to cover
destroy_devl() calls for the children of the destroyed dev.

Reported and tested by: joerg
MFC after: 1 week

13 years ago- Enable per-channel congestion notification.
np [Wed, 18 May 2011 22:09:04 +0000 (22:09 +0000)]
- Enable per-channel congestion notification.
- Enable PCIe relaxed ordering for all egress queues and rx data buffers.

MFC after: 3 days

13 years agoFix clang warnings.
benl [Wed, 18 May 2011 21:06:20 +0000 (21:06 +0000)]
Fix clang warnings.

Approved by: philip (mentor)

13 years agoFix clang warnings.
benl [Wed, 18 May 2011 21:04:29 +0000 (21:04 +0000)]
Fix clang warnings.

Note: possible minor security issues fixed (untrusted string used as
printf format string).

Approved by: philip (mentor)

13 years agoFix clang warnings.
benl [Wed, 18 May 2011 21:00:27 +0000 (21:00 +0000)]
Fix clang warnings.

Approved by: philip (mentor)

13 years agoFix clang warning (why is there nowhere yyparse() is declared?).
benl [Wed, 18 May 2011 20:57:23 +0000 (20:57 +0000)]
Fix clang warning (why is there nowhere yyparse() is declared?).

Approved by: philip (mentor)

13 years agoFix clang warnings.
benl [Wed, 18 May 2011 20:52:52 +0000 (20:52 +0000)]
Fix clang warnings.

Note: these were actually bugs (printf with no format string).

Approved by: philip (mentor)

13 years agoWipeout the end of disks, home to things like gmirror metadata, backup GPT tables,
jpaetzel [Wed, 18 May 2011 20:38:28 +0000 (20:38 +0000)]
Wipeout the end of disks, home to things like gmirror metadata, backup GPT tables,
and other potential evil.

Submitted by: Kris Moore <kris@pcbsd.org>
Approved by: kib (mentor)
Sponsored by: iXsystems

13 years agoExtracting optional components requires mounting devfs
jpaetzel [Wed, 18 May 2011 20:29:07 +0000 (20:29 +0000)]
Extracting optional components requires mounting devfs

Submitted by: Kris Moore <kris@pcbsd.org>
Approved by: kib (mentor)
Sponsored by: iXsystems

13 years agoUnbreak INET-less build.
tuexen [Wed, 18 May 2011 19:49:39 +0000 (19:49 +0000)]
Unbreak INET-less build.
Reported by bz@
MFC after: 1 week

13 years agoAdd a sanity check for the existence of an "addr" option
rmacklem [Wed, 18 May 2011 18:36:40 +0000 (18:36 +0000)]
Add a sanity check for the existence of an "addr" option
to both NFS clients. This avoids the crash reported by
Sergey Kandaurov (pluknet@gmail.com) to the freebsd-fs@
list with subject "[old nfsclient] different nmount()
args passed from mount vs mount_nfs" dated May 17, 2011.

Tested by: pluknet at gmail.com (old nfs client)
MFC after: 2 weeks

13 years agoAs requested by many people, with final prodding from Jason Hall, fix this
nwhitehorn [Wed, 18 May 2011 17:39:28 +0000 (17:39 +0000)]
As requested by many people, with final prodding from Jason Hall, fix this
so that running make release causes make obj to be run before doing
anything. This fixes a bug wherein, when run for the first time, and
without -DNOSRC, make release would attempt to recursively tar up the src
directory including its own output and enter an infinite loop.

While here, make the cross-building stuff work a little more the way it
should if only one of TARGET/TARGET_ARCH is specified.

13 years agoClean up a loose end from the conversion from gnu ar/ranlib to the BSD
imp [Wed, 18 May 2011 17:06:11 +0000 (17:06 +0000)]
Clean up a loose end from the conversion from gnu ar/ranlib to the BSD
one.  Without this, we don't have ar or randlib in the tool path,
leading to much pain for some users.  This pain is exposed by the
external toolchain enhancements that I'm working on.

Submitted by: John Hein (ages ago, and dropped on the floor by me: sorry)

13 years agoRevert r222069,222068 as they were intended to be committed to the
attilio [Wed, 18 May 2011 16:50:13 +0000 (16:50 +0000)]
Revert r222069,222068 as they were intended to be committed to the
largeSMP branch.

Reported by: pluknet

13 years agoFix warning spit out.
attilio [Wed, 18 May 2011 16:42:01 +0000 (16:42 +0000)]
Fix warning spit out.

Reported by: sbruno

13 years agoFix newly introduced code.
attilio [Wed, 18 May 2011 16:41:38 +0000 (16:41 +0000)]
Fix newly introduced code.

Reported by: sbruno

13 years agoMerge part of r221322 from largeSMP project:
attilio [Wed, 18 May 2011 16:07:30 +0000 (16:07 +0000)]
Merge part of r221322 from largeSMP project:
Sync XEN support with i386 about the usage of ipi_send_cpu()

Tested by: pluknet
MFC after: 2 weeks

13 years agousb: fix a missed use of use_generic in r222051
avg [Wed, 18 May 2011 11:38:36 +0000 (11:38 +0000)]
usb: fix a missed use of use_generic in r222051

Submitted by: gcooper
Pointyhat to: avg
MFC after: 1 month
X-MFC with: r222051

13 years agoThis isn't needed any longer, it's defined in ah_internal.h.
adrian [Wed, 18 May 2011 11:28:23 +0000 (11:28 +0000)]
This isn't needed any longer, it's defined in ah_internal.h.

13 years agoRelease allocated memory in procstat_close().
pluknet [Wed, 18 May 2011 10:04:54 +0000 (10:04 +0000)]
Release allocated memory in procstat_close().

Reviewed by: stass

13 years agousb: change to one-pass probing of device drivers
avg [Wed, 18 May 2011 07:40:12 +0000 (07:40 +0000)]
usb: change to one-pass probing of device drivers

This brings USB bus more in line with how newbus is supposed to be used.
Also, because of the two-pass probing the following message was produced
by devd in default configuration when almost any USB device was
connected:
Unknown USB device: vendor <> product <> bus <>
This should be fixed now.

Note that many USB device drivers pass some information from probe
method to attach method via ivars.  For this to continue working we rely
on the fact that the subr_bus code calls probe method of a winning driver
again before calling its attach method in the case where multiple
drivers claim to support a device.  This is done because device
description is set in successful probe methods and we want to get a correct
device description from a winning driver.  So now this logic is re-used
for setting ivars too.

Reviewed by: hselasky
MFC after: 1 month

13 years agoRestore old (v15) behaviour for a recursive snapshot destroy.
mm [Wed, 18 May 2011 07:37:02 +0000 (07:37 +0000)]
Restore old (v15) behaviour for a recursive snapshot destroy.
(zfs destroy -r pool/dataset@snapshot)

To destroy all descendent snapshots with the same name the top level
snapshot was not required to exist. So if the top level snapshot does
not exist, check permissions of the parent dataset instead.

Filed as Illumos Bug #1043

Reviewed by: delphij
Approved by: pjd
MFC after: together with v28

13 years agoModify the sample rate control algorithm to only select/sample HT
adrian [Wed, 18 May 2011 07:20:30 +0000 (07:20 +0000)]
Modify the sample rate control algorithm to only select/sample HT
rates for HT nodes.

13 years agoUpdate CPUID bits to reflect AMD Bulldozer and Intel Sandy Bridge features.
jkim [Tue, 17 May 2011 22:36:16 +0000 (22:36 +0000)]
Update CPUID bits to reflect AMD Bulldozer and Intel Sandy Bridge features.
Note AMD dropped SSE5 extensions in order to avoid ISA overlap with Intel
AVX instructions.  The SSE5 bit was recycled as XOP extended instruction
bit, CVT16 was deprecated in favor of F16C (half-precision float conversion
instructions for AVX), and the remaining FMA4 (4-operand FMA instructions)
gained a separate CPUID bit.  Replace non-existent references with today's
CPUID specifications.

13 years agoRemove an unused typedef.
attilio [Tue, 17 May 2011 22:15:53 +0000 (22:15 +0000)]
Remove an unused typedef.

Tested by: sbruno, pluknet

13 years agoMerge r221285 from largeSMP project:
attilio [Tue, 17 May 2011 22:14:00 +0000 (22:14 +0000)]
Merge r221285 from largeSMP project:
- Remove the following sysctl:
  kern.sched.ipiwakeup.onecpu
  kern.sched.ipiwakeup.htt2

  Because they are absolutely obsolete.  Probabilly the whole wakeup
  forward mechanism should be revisited for a better fitting in modern
  hw, in the future.
- As map2 variable is no longer used rename map3 to map2
- Fix a string by making more informative the msg and removing the
  arguments passing.

Reviewed by: julian
Tested by: several

13 years agoAdd support for "LED" enclosure management messages, defined by the AHCI.
mav [Tue, 17 May 2011 22:07:45 +0000 (22:07 +0000)]
Add support for "LED" enclosure management messages, defined by the AHCI.

When supported by hardware, this allows to control per-port activity, locate
and fault LEDs via the led(4) API for localization and status reporting
purposes. Supporting AHCI controllers may transmit that information to the
backplane controllers via SGPIO interface. Backplane controllers interpret
received statuses in some way (IBPI standard) to report them using present
indicators.

13 years agomq_setattr(2): Add missing const to man page.
jilles [Tue, 17 May 2011 21:03:34 +0000 (21:03 +0000)]
mq_setattr(2): Add missing const to man page.

The declaration in the header file is correct.

MFC after: 1 week

13 years agoOk, so we use different comment conventions in Symbol.map and Version.def
phk [Tue, 17 May 2011 19:20:32 +0000 (19:20 +0000)]
Ok, so we use different comment conventions in Symbol.map and Version.def
"#notimpressed" as my twittering friends would say.

13 years agoBackout libinstall.a -> libpkg commit.
flz [Tue, 17 May 2011 19:11:47 +0000 (19:11 +0000)]
Backout libinstall.a -> libpkg commit.

Discussed with: erwin, brooks, bapt

13 years agoTry to explain what sbufs do and add an example to show it.
phk [Tue, 17 May 2011 18:24:59 +0000 (18:24 +0000)]
Try to explain what sbufs do and add an example to show it.

Clarify return values.

13 years agoBump shlib version
phk [Tue, 17 May 2011 17:37:58 +0000 (17:37 +0000)]
Bump shlib version
Instigate symbol versioning
Pray that I don't break the build.

13 years agoFix a race in the SMP rendezvous code. Specifically, the write by the
jhb [Tue, 17 May 2011 16:39:08 +0000 (16:39 +0000)]
Fix a race in the SMP rendezvous code.  Specifically, the write by the
last CPU to to finish the rendezvous action may become visible to
different CPUs at different times.  As a result, the CPU that initiated
the rendezvous may exit the rendezvous and drop the lock allowing another
rendezvous to be initiated on the same CPU or a different CPU.  In that
case the exit sentinel may be cleared before all CPUs have noticed causing
those CPUs to hang forever.

Workaround this by using a generation count to notice when this race
occurs and to exit the rendezvous in that case.

The problem was independently diagnosted by mlaier@ and avg@ as well.

Submitted by: neel
Reviewed by: avg, mlaier
Obtained from: NetApp
MFC after: 1 week

13 years agoFix the debugging code path to correctly support HAL_DEBUG_UNMASKABLE.
adrian [Tue, 17 May 2011 16:30:34 +0000 (16:30 +0000)]
Fix the debugging code path to correctly support HAL_DEBUG_UNMASKABLE.

13 years agoCopy out the mtu when calling getsockopt() with SCTP_GET_PEER_ADDR_INFO.
tuexen [Tue, 17 May 2011 15:57:31 +0000 (15:57 +0000)]
Copy out the mtu when calling getsockopt() with SCTP_GET_PEER_ADDR_INFO.

MFC after: 1 week.

13 years agoFix whitespacing.
tuexen [Tue, 17 May 2011 15:46:28 +0000 (15:46 +0000)]
Fix whitespacing.
Reported by scf@

MFC after: 1 week.

13 years agoFix case, introduced in my previous commit.
adrian [Tue, 17 May 2011 15:03:39 +0000 (15:03 +0000)]
Fix case, introduced in my previous commit.

Pointy hat goes to: adrian, for having multiple build screens
open and checking the wrong one.

13 years agoFix manual section number for netstat(1).
pluknet [Tue, 17 May 2011 14:46:28 +0000 (14:46 +0000)]
Fix manual section number for netstat(1).

MFC after: 5 days

13 years agomdoc:
pluknet [Tue, 17 May 2011 12:58:19 +0000 (12:58 +0000)]
mdoc:
- use a proper macro for interface name ipfw0.
- add missing section number for bpf cross reference.

13 years agoUse the halMcastKeySrchSupport capability bit to selectively enable/disable
adrian [Tue, 17 May 2011 11:56:50 +0000 (11:56 +0000)]
Use the halMcastKeySrchSupport capability bit to selectively enable/disable
the multicast key search support for AR5212, AR5416 and later.

The general HAL routine ath_hal_getcapability() implement checking this
but it's overridden by a check in ar5212_misc:ar5212GetCapability().
This restores the later functionality in case it's found to be broken
in any of the 11n chipsets.

13 years agoSet this HAL capabilities flag correctly even though it isn't currently
adrian [Tue, 17 May 2011 11:52:53 +0000 (11:52 +0000)]
Set this HAL capabilities flag correctly even though it isn't currently
being used.

13 years agoDocument the supported hardware, somewhat.
ru [Tue, 17 May 2011 11:26:59 +0000 (11:26 +0000)]
Document the supported hardware, somewhat.

Reviewed by: hselasky
MFC after: 3 days

13 years agoRenamed PCI_INTERFACE_XHCI to PCIP_SERIALBUS_USB_XHCI and moved it
ru [Tue, 17 May 2011 11:23:43 +0000 (11:23 +0000)]
Renamed PCI_INTERFACE_XHCI to PCIP_SERIALBUS_USB_XHCI and moved it
to <dev/pci/pcireg.h>.

Reviewed by: hselasky
MFC after: 3 days

13 years agoRegen for WITH_OFED.
ru [Tue, 17 May 2011 11:08:15 +0000 (11:08 +0000)]
Regen for WITH_OFED.

13 years agoAdded description for WITH_OFED.
ru [Tue, 17 May 2011 11:06:41 +0000 (11:06 +0000)]
Added description for WITH_OFED.

13 years agoUse memset() instead of bzero() and memcpy() instead of bcopy(), there
phk [Tue, 17 May 2011 11:04:50 +0000 (11:04 +0000)]
Use memset() instead of bzero() and memcpy() instead of bcopy(), there
is no relevant difference for sbufs, and it increases portability of
the source code.

Split the actual initialization of the sbuf into a separate local
function, so that certain static code checkers can understand
what sbuf_new() does, thus eliminating on silly annoyance of
MISRA compliance testing.

Contributed by: An anonymous company in the last business I
expected sbufs to invade.

13 years agoRevert change to "MF" I made in r189767. I bet that at the time of r189767
ru [Tue, 17 May 2011 10:44:11 +0000 (10:44 +0000)]
Revert change to "MF" I made in r189767.  I bet that at the time of r189767
I checked with http://www.iso.org/iso/country_codes/iso_3166_code_lists.htm
and "MF" was officially spelled in English as "Saint Martin" there, but now
that "SX" exists (for "Sint Maarten (Dutch part)") (nice official "English"
spelling!) they seem to have added a "(French part)" suffix to "MF".  Since
this is also in line with Newsletter VI-1 (2007-09-21), catch up.

13 years agoAdd missing section number for .Xr jail.
pluknet [Tue, 17 May 2011 10:38:44 +0000 (10:38 +0000)]
Add missing section number for .Xr jail.

MFC after: 3 days

13 years agoMore thorough mdoc and language fixes.
uqs [Tue, 17 May 2011 09:51:02 +0000 (09:51 +0000)]
More thorough mdoc and language fixes.

Submitted by: ru

13 years agoISO3166: Update for newsletters VI-7 and VI-8 from 2010
uqs [Tue, 17 May 2011 08:13:27 +0000 (08:13 +0000)]
ISO3166: Update for newsletters VI-7 and VI-8 from 2010

- Name change for SH
- BQ, CW, and SX replace AN

Reviewed by: ru

13 years agoWhitespace and typo fixes.
uqs [Tue, 17 May 2011 08:13:18 +0000 (08:13 +0000)]
Whitespace and typo fixes.

Found by: codespell
Reviewed by: ru

13 years agoTypo fix.
uqs [Tue, 17 May 2011 08:13:10 +0000 (08:13 +0000)]
Typo fix.

13 years agoTypos, wording and mdoc fixes.
uqs [Tue, 17 May 2011 08:12:59 +0000 (08:12 +0000)]
Typos, wording and mdoc fixes.

13 years agoRemove redundant keywords.
hrs [Tue, 17 May 2011 07:40:13 +0000 (07:40 +0000)]
Remove redundant keywords.

Submitted by: wxs

13 years agoDon't expect PAGE_SIZE to exist on all platforms (It is a pretty arbitrary
phk [Tue, 17 May 2011 06:36:32 +0000 (06:36 +0000)]
Don't expect PAGE_SIZE to exist on all platforms (It is a pretty arbitrary
choice of default size in the first place)

Reverse the order of arguments to the internal static sbuf_put_byte()
function to match everything else in this file.

Move sbuf_putc_func() inside the kernel version of sbuf_vprintf
where it belongs.

sbuf_putc() incorrectly used sbuf_putc_func() which supress NUL
characters, it should use sbuf_put_byte().

Make sbuf_finish() return -1 on error.

Minor stylistic nits fixed.