]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
14 years ago- Remove a redundant check for dpms(4).
jkim [Thu, 12 Nov 2009 18:16:35 +0000 (18:16 +0000)]
- Remove a redundant check for dpms(4).
- Test a cheaper function first.

14 years ago- Partially revert hackish r198964 and r199002.
jkim [Thu, 12 Nov 2009 17:56:56 +0000 (17:56 +0000)]
- Partially revert hackish r198964 and r199002.
- Add a proxy driver vgapm to help vgapci to save/load VGA state.
- Move device_set_desc() to the right place while we are here.

Reviewed by: jhb

14 years agoWe need to allocate space for the header in the create path also.
rnoland [Thu, 12 Nov 2009 16:28:39 +0000 (16:28 +0000)]
We need to allocate space for the header in the create path also.

This fixes a null pointer dereference with "gpart create -s GPT" after
the previous commit.

Reported by: Yuri Pankov
Pointyhat to: me
MFC after: 1 week

14 years agoAdd the possibility for vfs.root.mountfrom tunable to accept a list of
attilio [Thu, 12 Nov 2009 15:59:05 +0000 (15:59 +0000)]
Add the possibility for vfs.root.mountfrom tunable to accept a list of
items rather than a single one. The list is a space separated collection
of items defined as the current one accepted.

While there fix also a nit in a comment.

Obtained from: Sandvine Incorporated
Reviewed by: emaste
Tested by: Giovanni Trematerra
<giovanni dot trematerra at gmail dot com>
Sponsored by: Sandvine Incorporated
MFC: 2 weeks

14 years agoProvide a real fix to the too-many-translations problem when booting
nwhitehorn [Thu, 12 Nov 2009 15:19:09 +0000 (15:19 +0000)]
Provide a real fix to the too-many-translations problem when booting
from CD on 64-bit hardware to replace existing band-aids. This occurred
when the preloaded mdroot required too many mappings for the static
buffer.

Since we only use the translations buffer once, allocate a dynamic
buffer on the stack. This early in the boot process, the call chain
is quite short and we can be assured of having sufficient stack space.

Reviewed by: grehan

14 years ago- We are not guaranteed that we're not dropping a reference that
ume [Thu, 12 Nov 2009 14:48:36 +0000 (14:48 +0000)]
- We are not guaranteed that we're not dropping a reference that
  we did not add.  Call LLE_REMREF() only when callout_stop()
  actually canceled a pending callout.
- callout_reset() may cancel a pending callout.  When
  callout_reset() canceled a pending callout, call LLE_REMREF()
  to drop a reference for the canceled callout.

MFC after: 1 week

14 years agoRemove the explicit definition of inet_aton() as it was introduced as a
attilio [Thu, 12 Nov 2009 14:22:12 +0000 (14:22 +0000)]
Remove the explicit definition of inet_aton() as it was introduced as a
general function in r199208.

Reported by: np
Sponsored by: Sandvine Incorporated
MFC: 1 week

14 years agoAdd missing IEEE1394 support dropped during merge from NetBSD.
ume [Thu, 12 Nov 2009 11:54:12 +0000 (11:54 +0000)]
Add missing IEEE1394 support dropped during merge from NetBSD.

14 years agoMFi386: revision 199104
nyan [Thu, 12 Nov 2009 11:27:07 +0000 (11:27 +0000)]
MFi386: revision 199104

  Make isa_dma functions MPSAFE by introducing its own private lock.

14 years agoFix cpu model for PODP5V83. It is P24T, not P54T.
nyan [Thu, 12 Nov 2009 10:59:00 +0000 (10:59 +0000)]
Fix cpu model for PODP5V83.  It is P24T, not P54T.
Also remove redundant 'Overdrive' word.

Pointed out by: SATOU Tomokazu (tomo1770 at maple ocn ne jp)
MFC after: 1 week

14 years ago- Style nits.
kuriyama [Thu, 12 Nov 2009 03:31:19 +0000 (03:31 +0000)]
- Style nits.
- Remove unneeded TUNABLE_INT().

Suggested by: avg, kib

14 years agoAdd test-pidfile.c and increase warning level.
des [Thu, 12 Nov 2009 01:37:25 +0000 (01:37 +0000)]
Add test-pidfile.c and increase warning level.

14 years agoTest cases for pidfile(3) - including two designed to catch issues arising
des [Thu, 12 Nov 2009 01:37:02 +0000 (01:37 +0000)]
Test cases for pidfile(3) - including two designed to catch issues arising
from the incorrect use of fcntl(2) instead of flock(2).

14 years agoFix warnings and remove one unnecessary use of vfork(). The other could
des [Thu, 12 Nov 2009 01:34:55 +0000 (01:34 +0000)]
Fix warnings and remove one unnecessary use of vfork().  The other could
also be removed with a little more work.

14 years agoFix warnings
des [Thu, 12 Nov 2009 01:33:57 +0000 (01:33 +0000)]
Fix warnings

14 years agoIntroduce a new option (BOOT_PROMPT_123) that lets enter the boot prompt
attilio [Thu, 12 Nov 2009 01:30:17 +0000 (01:30 +0000)]
Introduce a new option (BOOT_PROMPT_123) that lets enter the boot prompt
only when typing the sequence "123" (opposite to the standard 'push any
button' approach).
That results useful when using serial lines sending garbage and leading
to unwilling boot prompt appearence.

Obtained from: Sandvine Incorporated
Reviewed by: emaste, jhb
Sponsored by: Sandvine Incorporated
MFC: 1 week

14 years agoThe building the dev nameunit string, in devclass_add_device() is based
attilio [Thu, 12 Nov 2009 00:52:14 +0000 (00:52 +0000)]
The building the dev nameunit string, in devclass_add_device() is based
on the assumption that the unit linked with the device is invariant but
that can change when calling devclass_alloc_unit() (because -1 is passed
or, more simply, because the unit choosen is beyond the table limits).
This results in a completely bogus string building.

Fix this by reserving the necessary room for all the possible characters
printable by a positive integer (we do not allow for negative unit
number).

Reported by: Sandvine Incorporated
Reviewed by: emaste
Sponsored by: Sandvine Incorporated
MFC: 1 week

14 years agoMove inet_aton() (specular to inet_ntoa(), already present in libkern)
attilio [Thu, 12 Nov 2009 00:46:28 +0000 (00:46 +0000)]
Move inet_aton() (specular to inet_ntoa(), already present in libkern)
into libkern in order to made it usable by other modules than alias_proxy.

Obtained from: Sandvine Incorporated
Sponsored by: Sandvine Incorporated
MFC: 1 week

14 years agosh: Use sigaction instead of signal/siginterrupt combination.
jilles [Wed, 11 Nov 2009 23:13:24 +0000 (23:13 +0000)]
sh: Use sigaction instead of signal/siginterrupt combination.

14 years agoAdd interface description capability as inspired by OpenBSD.
delphij [Wed, 11 Nov 2009 21:30:58 +0000 (21:30 +0000)]
Add interface description capability as inspired by OpenBSD.

MFC after: 3 months

14 years agoMore consistent whitespace.
jhb [Wed, 11 Nov 2009 20:29:40 +0000 (20:29 +0000)]
More consistent whitespace.

14 years agoUse a dedicated callout to drive the transmit watchdog timer instead of
jhb [Wed, 11 Nov 2009 20:27:53 +0000 (20:27 +0000)]
Use a dedicated callout to drive the transmit watchdog timer instead of
using if_watchdog and if_timer.

Tested by: gavin

14 years ago- Remove trailing ";" after if statement
antoine [Wed, 11 Nov 2009 19:39:45 +0000 (19:39 +0000)]
- Remove trailing ";" after if statement
- Remove #if 0 section that was never needed/used

Reviewed by: raj@
MFC after: 1 month

14 years agoWith an i386 kernel the igb driver can cause a
jfv [Wed, 11 Nov 2009 19:13:40 +0000 (19:13 +0000)]
With an i386 kernel the igb driver can cause a
page fault panic on initialization due to a large
number of bounce pages being allocated. This is due
to the dma tag requiring page alignment on mbuf mapping.
This was removed some time back from the ixgbe driver
and is not needed here either.

14 years agoSynchronize with C.msg revision 199083 and improve some existing messages.
rene [Wed, 11 Nov 2009 18:28:12 +0000 (18:28 +0000)]
Synchronize with C.msg revision 199083 and improve some existing messages.

Reviewed by: remko
Approved by: remko

14 years agoCreate verifier used by FreeBSD NFS client is suboptimal because the
jh [Wed, 11 Nov 2009 15:43:07 +0000 (15:43 +0000)]
Create verifier used by FreeBSD NFS client is suboptimal because the
first part of a verifier is set to the first IP address from
V_in_ifaddrhead list. This address is typically the loopback address
making the first part of the verifier practically non-unique. The second
part of the verifier is initialized to zero making its initial value
non-unique too.

This commit changes the strategy for create verifier initialization:
just initialize it to a random value. Also move verifier handling into
its own function and use a mutex to protect the variable.

This change is a candidate for porting to sys/nfsclient.

Reviewed by: jhb, rmacklem
Approved by: trasz (mentor)

14 years agoANSIfy.
ume [Wed, 11 Nov 2009 15:21:06 +0000 (15:21 +0000)]
ANSIfy.

MFC after: 1 week

14 years agoRemove trailing ";" in struct ieee80211_beacon_offsets declaration
antoine [Wed, 11 Nov 2009 15:00:56 +0000 (15:00 +0000)]
Remove trailing ";" in struct ieee80211_beacon_offsets declaration

Found by: phk's FlexeLint in September
Reviewed by: rpaulo@
MFC after: 1 month

14 years agoFix off by one in ieee80211_send_action_register
antoine [Wed, 11 Nov 2009 14:58:48 +0000 (14:58 +0000)]
Fix off by one in ieee80211_send_action_register

Found by: phk's FlexeLint in September
Reviewed by: rpaulo@
MFC after: 1 month

14 years agoreflect that pg_ps_enabled is a tunable, not just a read-only sysctl
avg [Wed, 11 Nov 2009 14:21:31 +0000 (14:21 +0000)]
reflect that pg_ps_enabled is a tunable, not just a read-only sysctl

Nod from: jhb

14 years agoRevert r198873. Having different VAPPEND semantics for VOP_ACCESS(9)
trasz [Wed, 11 Nov 2009 13:49:22 +0000 (13:49 +0000)]
Revert r198873.  Having different VAPPEND semantics for VOP_ACCESS(9)
and VOP_ACCESSX(9) is not a good idea.

14 years agoAdd links to zfs(8) and zpool(8) to mount(8) manual page.
trasz [Wed, 11 Nov 2009 12:55:58 +0000 (12:55 +0000)]
Add links to zfs(8) and zpool(8) to mount(8) manual page.

14 years agoCorrect the information about the doceng@ team members - Murray Stokely
roam [Wed, 11 Nov 2009 11:37:43 +0000 (11:37 +0000)]
Correct the information about the doceng@ team members - Murray Stokely
stepped down some time ago, about the same time as Giorgos Keramidas
joined the team.

PR: 140465
Submitted by: Denny Lin <dennylin93@cnmc32.hs.ntnu.edu.tw>
MFC after: 2 weeks

14 years agoFix the grammar as in the PR, and then some.
roam [Wed, 11 Nov 2009 11:31:02 +0000 (11:31 +0000)]
Fix the grammar as in the PR, and then some.

PR: 140454
Submitted by: Jeremy Huddleston <jeremyhu@apple.com>
MFC after: 2 weeks

14 years agoAdd unit to the short month names for Japanese locales.
ume [Wed, 11 Nov 2009 11:24:02 +0000 (11:24 +0000)]
Add unit to the short month names for Japanese locales.
Without unit, the output of the application like ls(1)
is complicated.

Reviewed by: nork
MFC after: 1 week

14 years agoMFp4:
mav [Wed, 11 Nov 2009 11:10:36 +0000 (11:10 +0000)]
MFp4:
- Move tagged queueing control from ADA to ATA XPT. It allows to control
  device command queue length correctly. First step to support < 32 tags.
- Limit queue for non-tagged devices by 2 slots for ahci(4) and siis(4).
- Implement quirk matching for ATA devices.
- Move xpt_schedule_dev_sendq() from header to source file.
- Move delayed queue shrinking to the more expected place - element freeing.
- Remove some SCSIsms in ATA.

14 years agoAdded option NETGRAPH_VLAN.
ru [Wed, 11 Nov 2009 11:07:30 +0000 (11:07 +0000)]
Added option NETGRAPH_VLAN.

Submitted by: pluknet

14 years agoMFp4:
mav [Wed, 11 Nov 2009 10:44:09 +0000 (10:44 +0000)]
MFp4:
Add set of chip IDs, known to support AHCI.

14 years agoPlace home and end before insert and delete.
ed [Wed, 11 Nov 2009 09:43:26 +0000 (09:43 +0000)]
Place home and end before insert and delete.

These keys have different sequences when using cursorkeys, while insert
and delete stay the same. If they are placed like this, libteken will
return NULL instead of a proper sequence for these characters.

14 years agoAdd a new flag to vidcontrol, -T, that allows terminal mode switching.
ed [Wed, 11 Nov 2009 08:39:57 +0000 (08:39 +0000)]
Add a new flag to vidcontrol, -T, that allows terminal mode switching.

This will make it more easy for people to experiment with TERM=xterm.
Instead of echoing these strange escape sequences, I can just instruct
them to run `vidcontrol -T xterm'.

14 years agoCURVNET_RESTORE() was not called in certain cases.
ume [Wed, 11 Nov 2009 08:28:18 +0000 (08:28 +0000)]
CURVNET_RESTORE() was not called in certain cases.

MFC after: 3 days

14 years agoAllow Syscons terminal emulators to provide function key strings.
ed [Wed, 11 Nov 2009 08:20:19 +0000 (08:20 +0000)]
Allow Syscons terminal emulators to provide function key strings.

xterm and cons25 have some incompatibilities when it comes to escape
sequences for special keys, such as F1 to F12, home, end, etc. Add a new
te_fkeystr() that can be used to override the strings.

scterm-sck won't do anything with this, but scterm-teken will use
teken_get_sequences() to obtain the proper sequence.

14 years agoAlways home the cursor when changing the scrolling region.
ed [Wed, 11 Nov 2009 08:11:21 +0000 (08:11 +0000)]
Always home the cursor when changing the scrolling region.

I thought this only had to be done when in origin mode, to ensure that
the cursor is not placed outside the origin, but it seems this is also
done when not in origin mode.

This fixes some artifacts when pressing ^L while running irssi in tmux.
(Almost) nobody noticed this, because cons25 doesn't have scrolling
regions.

14 years agoReduce probe priority of USB input devices to BUS_PROBE_GENERIC from
nwhitehorn [Wed, 11 Nov 2009 03:17:51 +0000 (03:17 +0000)]
Reduce probe priority of USB input devices to BUS_PROBE_GENERIC from
BUS_PROBE_SPECIFIC. This allows device-specific drivers like atp to
attach reliably.

Reviewed by: hps

14 years agoBe careful which vattr fields are set during setattr replay.
pjd [Tue, 10 Nov 2009 22:27:33 +0000 (22:27 +0000)]
Be careful which vattr fields are set during setattr replay.
Without this fix strange things can appear after unclean shutdown like
files with mode set to 07777.

Reported by: des
MFC after: 3 days

14 years agoAvoid passing invalid mountpoint to getnewvnode().
pjd [Tue, 10 Nov 2009 22:25:46 +0000 (22:25 +0000)]
Avoid passing invalid mountpoint to getnewvnode().

Reported by: rwatson
Tested by: rwatson
MFC after: 3 days

14 years agoAdd ixgb(4) to NOTES.
jhb [Tue, 10 Nov 2009 22:07:37 +0000 (22:07 +0000)]
Add ixgb(4) to NOTES.

Approved by: jfv

14 years ago- Locking fixes to not do silly things like drop the lock only to call a
jhb [Tue, 10 Nov 2009 22:04:19 +0000 (22:04 +0000)]
- Locking fixes to not do silly things like drop the lock only to call a
  function that immediately reacquires the lock.  Also removes recursive
  locking.
- Use the statistics timer to drive the transmit watchdog instead of using
  if_watchdog and if_timer.

Tested by: gavin

14 years agoController does not update Tx descriptors(send BDs) after sending
yongari [Tue, 10 Nov 2009 20:29:20 +0000 (20:29 +0000)]
Controller does not update Tx descriptors(send BDs) after sending
frames so remove unnecessary BUS_DMASYNC_PREREAD and
BUS_DMASYNC_POSTREAD of bus_dmamap_sync(9).

14 years agos/a default/the default/
dougb [Tue, 10 Nov 2009 19:50:28 +0000 (19:50 +0000)]
s/a default/the default/

Submitted by: remko

14 years agoAdd support for the touchpads found in later models of iBook and
nwhitehorn [Tue, 10 Nov 2009 19:14:06 +0000 (19:14 +0000)]
Add support for the touchpads found in later models of iBook and
Powerbook.

Reviewed by: Rohit Grover <rgrover1 at gmail.com>

14 years agoPurge some useless mergeinfo.
jhb [Tue, 10 Nov 2009 14:32:09 +0000 (14:32 +0000)]
Purge some useless mergeinfo.

14 years agoWhen rename("a", "b/.") is performed, target namei() call returns
kib [Tue, 10 Nov 2009 11:50:37 +0000 (11:50 +0000)]
When rename("a", "b/.") is performed, target namei() call returns
dvp == vp. Rename syscall does not check for the case, and at least
ufs_rename() cannot deal with it. POSIX explicitely requires that both
rename(2) and rmdir(2) return EINVAL when any of the pathes end in "/.".

Detect the slashdot lookup for RENAME or REMOVE in lookup(), and return
EINVAL.

Reported by: Jim Meyering <jim meyering net>
Tested by: simon, pho
MFC after: 1 week

14 years agoIn r198506, kern_sigsuspend() started doing cursig/postsig loop to make
kib [Tue, 10 Nov 2009 11:46:53 +0000 (11:46 +0000)]
In r198506, kern_sigsuspend() started doing cursig/postsig loop to make
sure that a signal was delivered to the thread before returning from
syscall. Signal delivery puts new return frame on the user stack, and
modifies trap frame to enter signal handler. As a consequence, syscall
return code sets EINTR as error return for signal frame, instead of the
syscall return.

Also, for ia64, due to different registers layout for those two kind of
frames, usermode sigsegfaulted when returned from signal handler.

Use newly-introduced cpu_set_syscall_retval(9) to set syscall result,
and return EJUSTRETURN from kern_sigsuspend() to prevent syscall return
code from modifying this frame [1].

Another issue is that pending SIGCONT might be cancelled by SIGSTOP,
causing postsig() not to deliver any catched signal [2]. Modify
postsig() to return 1 if signal was posted, and 0 otherwise, and use
this in the kern_sigsuspend loop.

Proposed by: marcel [1]
Noted by: davidxu [2]
Reviewed by: marcel, davidxu
MFC after: 1 month

14 years agoExtract the code that records syscall results in the frame into MD
kib [Tue, 10 Nov 2009 11:43:07 +0000 (11:43 +0000)]
Extract the code that records syscall results in the frame into MD
function cpu_set_syscall_retval().

Suggested by: marcel
Reviewed by: marcel, davidxu
PowerPC, ARM, ia64 changes: marcel
Sparc64 tested and reviewed by: marius, also sunv reviewed
MIPS tested by: gonzo
MFC after: 1 month

14 years agoCheck fork() return value
des [Tue, 10 Nov 2009 10:42:48 +0000 (10:42 +0000)]
Check fork() return value

14 years agoRemove a bunch of code used to detect SMP on ((i386 && !pc98) || amd64) and
des [Tue, 10 Nov 2009 10:34:44 +0000 (10:34 +0000)]
Remove a bunch of code used to detect SMP on ((i386 && !pc98) || amd64) and
offer to install an SMP kernel.  The way this worked was: on supported
platforms, code to read ACPI tables and BIOS MP tables was compiled into
sysinstall, and if an SMP kernel config was present in the source tree when
sysinstall was built, code that called it was also compiled.  Since we
haven't had SMP kernel configs in years, the latter was never compiled and
the former never ran.

This only removes dead and unreachable code; it does *not* remove the NCpus
variable, nor the code that sets it to 1, nor the code that asks the user to
select a kernel from a list.

Discussed with: re@, randi@ and others

14 years agoMFp4:
mav [Tue, 10 Nov 2009 09:46:52 +0000 (09:46 +0000)]
MFp4:
Organize device IDs and add some more of them.

14 years agoFix globbing
des [Tue, 10 Nov 2009 09:45:43 +0000 (09:45 +0000)]
Fix globbing

Noticed by: delphij, David Cornejo <dave@dogwood.com>
Forgotten by: des

14 years agoMore rational usage()
des [Tue, 10 Nov 2009 09:44:55 +0000 (09:44 +0000)]
More rational usage()

14 years agoAdd Japanese catalogue entries for newer errnos: EBADMSG, EMULTIHOP,
ume [Tue, 10 Nov 2009 03:56:51 +0000 (03:56 +0000)]
Add Japanese catalogue entries for newer errnos: EBADMSG, EMULTIHOP,
ENOLINK, EPROTO, ENOTCAPABLE.

14 years agoAdd a note about no hostname leading to "Amnesiac" on the console
dougb [Tue, 10 Nov 2009 03:18:49 +0000 (03:18 +0000)]
Add a note about no hostname leading to "Amnesiac" on the console

The text is inspired by the PR, but more in line with the existing text

PR: docs/140434
Submitted by: Jason Helfman <jhelfman@e-e.com>

14 years agoAdd a minimal change to prevent NULL deference in ee(1).
delphij [Tue, 10 Nov 2009 00:48:24 +0000 (00:48 +0000)]
Add a minimal change to prevent NULL deference in ee(1).

To repeat the problem, one can press "Ctrl+C" and then enter "0".

Submitted by: Alexander Best <alexbestms wwu de>

14 years agoZero out Tx/Rx descriptors before using them. Also add missing
yongari [Mon, 9 Nov 2009 23:09:18 +0000 (23:09 +0000)]
Zero out Tx/Rx descriptors before using them. Also add missing
bus_dmamap_sync(9) after Tx descriptor initialization.

14 years agoAdd missing bus_dmamap_sync(9) before issuing kick command.
yongari [Mon, 9 Nov 2009 22:58:30 +0000 (22:58 +0000)]
Add missing bus_dmamap_sync(9) before issuing kick command.

14 years agoUnbreak booting of FreeBSD/mips by merging r195429 from projects/mips:
gonzo [Mon, 9 Nov 2009 22:01:58 +0000 (22:01 +0000)]
Unbreak booting of FreeBSD/mips by merging r195429 from projects/mips:
- Move dpcpu initialization to mips_proc0_init. It's
    more appropriate place for it. Besides dpcpu_init
    requires pmap module to be initialized and calling it
    int pmap.c hangs the system

14 years agoSpell sz correctly.
nwhitehorn [Mon, 9 Nov 2009 21:12:28 +0000 (21:12 +0000)]
Spell sz correctly.

Pointed out by: jmallett

14 years agoMFV of r199105, tzdata2009r:
edwin [Mon, 9 Nov 2009 20:49:12 +0000 (20:49 +0000)]
MFV of r199105, tzdata2009r:

- Three Australian stations in Antarctica have changed their time zone:
  Casey moved from UTC+8 to UTC+11
  Davis moved from UTC+7 to UTC+5
  Mawson moved from UTC+6 to UTC+5
  The changes occurred on 2009-10-18 at 02:00 (local times).

MFC after: 3 days

14 years agoMake isa_dma functions MPSAFE by introducing its own private lock. These
rdivacky [Mon, 9 Nov 2009 20:29:10 +0000 (20:29 +0000)]
Make isa_dma functions MPSAFE by introducing its own private lock. These
functions are selfcontained (ie. they touch only isa_dma.c static variables
and hardware) so a private lock is sufficient to prevent races. This changes
only i386/amd64 while there are also isa_dma functions for ia64/sparc64.
Sparc64 are ones empty stubs and ia64 ones are unused as ia64 does not
have isa (says marcel).

This patch removes explicit locking of Giant from a few drivers (there
are some that requires this but lack ones - this patch fixes this) and
also removes the need for implicit locking of Giant from attach routines
where it's provided by newbus.

Approved by: ed (mentor, implicit)
Reviewed by: jhb, attilio (glanced by)
Tested by: Giovanni Trematerra <giovanni.trematerra gmail com>
IA64 clue: marcel

14 years agoRemove ifdefed out part of code, which seems to have originated a decade ago
trasz [Mon, 9 Nov 2009 19:53:34 +0000 (19:53 +0000)]
Remove ifdefed out part of code, which seems to have originated a decade ago
in OpenBSD.  As it is now, there is no way for this to be useful, since IPsec
is free to forward packets via whatever interface it wants, so checking
capabilities of the interface passed from ip_output (fetched from the routing
table) serves no purpose.

Discussed with: sam@

14 years agoFix variable type.
mav [Mon, 9 Nov 2009 19:47:46 +0000 (19:47 +0000)]
Fix variable type.

14 years agoPurge some duplicate mergeinfo.
jhb [Mon, 9 Nov 2009 18:50:34 +0000 (18:50 +0000)]
Purge some duplicate mergeinfo.

14 years agoAdd gai_strerror() catalog for ja_JP.UTF-8 and ja_JP.eucJP.
ume [Mon, 9 Nov 2009 17:26:16 +0000 (17:26 +0000)]
Add gai_strerror() catalog for ja_JP.UTF-8 and ja_JP.eucJP.

14 years agoMention the layout change of ieee80211req_scan_result.
rpaulo [Mon, 9 Nov 2009 16:05:32 +0000 (16:05 +0000)]
Mention the layout change of ieee80211req_scan_result.

14 years agoDriver for the Apple Touchpad present on MacBook (non-Pro & Pro).
rpaulo [Mon, 9 Nov 2009 15:59:09 +0000 (15:59 +0000)]
Driver for the Apple Touchpad present on MacBook (non-Pro & Pro).

Submitted by: Rohit Grover <rgrover1 at gmail.com>
MFC after: 2 months

14 years agoIncrease the size of the OFW translations buffer to handle G5 systems
nwhitehorn [Mon, 9 Nov 2009 14:26:23 +0000 (14:26 +0000)]
Increase the size of the OFW translations buffer to handle G5 systems
that use many translation regions in firmware, and add bounds checking
to prevent buffer overflows in case even the new value is exceeded.

Reported by: Jacob Lambert
MFC after: 3 days

14 years agoAdd NLS catalogs support to gai_strerror(3).
ume [Mon, 9 Nov 2009 12:46:59 +0000 (12:46 +0000)]
Add NLS catalogs support to gai_strerror(3).
Controlled by NLS define.

14 years agoFix comment.
ume [Mon, 9 Nov 2009 12:38:13 +0000 (12:38 +0000)]
Fix comment.

Pointed out by: nyan
MFC after: 1 week

14 years agoAdd ja_JP.eucJP catalog.
ume [Mon, 9 Nov 2009 12:33:47 +0000 (12:33 +0000)]
Add ja_JP.eucJP catalog.

Reviewed by: hrs, nork, takawata
MFC after: 1 week

14 years agoAdd ja_JP.UTF-8 catalog.
ume [Mon, 9 Nov 2009 12:28:59 +0000 (12:28 +0000)]
Add ja_JP.UTF-8 catalog.

Reviewed by: hrs, nork, takawata
MFC after: 1 week

14 years agoAdd support for ATA Power Management.
mav [Mon, 9 Nov 2009 11:39:51 +0000 (11:39 +0000)]
Add support for ATA Power Management.

14 years agoThe isr_intval in ieee80211req_scan_result structure should be 16 bit.
rpaulo [Mon, 9 Nov 2009 11:23:37 +0000 (11:23 +0000)]
The isr_intval in ieee80211req_scan_result structure should be 16 bit.
This makes ifconfig list scan display the correct beacon interval
(previously it would int overflow). As a side effect, this makes the
ieee80211req_scan_result word aligned.

Submitted by: Paul B Mahol <onemda at gmail.com>

14 years agoAdd more ICH10 chip IDs.
mav [Mon, 9 Nov 2009 09:27:09 +0000 (09:27 +0000)]
Add more ICH10 chip IDs.

Submitted by: Dmitry S. Luhtionov <mitya@cabletv.dp.ua>

14 years agostyle(9): add missing parentheses
oleg [Mon, 9 Nov 2009 09:12:45 +0000 (09:12 +0000)]
style(9): add missing parentheses

14 years agoInitialize the whole message unit's DMA buffer to zero, this fixes a panic
delphij [Mon, 9 Nov 2009 07:28:29 +0000 (07:28 +0000)]
Initialize the whole message unit's DMA buffer to zero, this fixes a panic
during boot when ARC1200 is being used with certain motherboard models.

This commit brings the driver to the same state of vendor's 1.20.00.16
release.  Many thanks to Areca for their continued support to FreeBSD.

Reported by: Jirka Mikulas <jiri mikulas com>
Submitted by: Erich Chen (Areca)
Obtained from: ftp://ftp.areca.com.tw/RaidCards/AP_Drivers/FreeBSD/DRIVER/SourceCode/arcmsr-freebsd-1.20.00.16-91010.zip
MFC after: 3 days

14 years ago- Add hw.clflush_disable loader tunable to avoid panic (trap 9) at
kuriyama [Mon, 9 Nov 2009 02:54:16 +0000 (02:54 +0000)]
- Add hw.clflush_disable loader tunable to avoid panic (trap 9) at
  map_invalidate_cache_range() even if CPU is not Intel.
- This tunable can be set to -1 (default), 0 and 1.  -1 is same as
  current behavior, which automatically disable CLFLUSH on Intel CPUs
  without CPUID_SS (should be occured on Xen only).  You can specify 1
  when this panic happened on non-Intel CPUs (such as AMD's).  Because
  disabling CLFLUSH may reduce performance, you can try with setting 0
  on Intel CPUs without SS to use CLFLUSH feature.

Reviewed by: kib
Reported by: karl, kuriyama
Related to: kern/138863

14 years agoApply a NetBSD fix (revision 1.12) to handle multi-session bzip2 files
delphij [Mon, 9 Nov 2009 02:37:02 +0000 (02:37 +0000)]
Apply a NetBSD fix (revision 1.12) to handle multi-session bzip2 files
as created by pbzip2.

Submitted by: mrg (NetBSD.org)
MFC after: 1 week

14 years agoCorrect disabling checksum offloading for BCM5700 B0.
yongari [Mon, 9 Nov 2009 00:16:50 +0000 (00:16 +0000)]
Correct disabling checksum offloading for BCM5700 B0.

14 years agoCorrect Olympus quirk.
thompsa [Sun, 8 Nov 2009 21:08:50 +0000 (21:08 +0000)]
Correct Olympus quirk.

Submitted by: Pavel Gubin

14 years agoAdd missing mtx_destroy().
thompsa [Sun, 8 Nov 2009 21:07:47 +0000 (21:07 +0000)]
Add missing mtx_destroy().

Submitted by: Sebastian Huber

14 years agoImprove support for High-speed USB audio devices.
thompsa [Sun, 8 Nov 2009 21:00:50 +0000 (21:00 +0000)]
Improve support for High-speed USB audio devices.
- fix issues regarding the mixer, where the interface number was not set in
  time.
- fix wrong use of resolution parameter.

Submitted by: Hans Petter Selasky

14 years agoimprove support for high speed isochronous endpoints which does not run 1:1,
thompsa [Sun, 8 Nov 2009 20:54:03 +0000 (20:54 +0000)]
improve support for high speed isochronous endpoints which does not run 1:1,
but needs intervalling 1:2, 1:4 or 1:8

Submitted by: Hans Petter Selasky

14 years agoIntegrate lost interrupts patch from the old USB stack.
thompsa [Sun, 8 Nov 2009 20:51:15 +0000 (20:51 +0000)]
Integrate lost interrupts patch from the old USB stack.

Some EHCI chips from VIA / ATI seem to trigger interrupts before writing back
the qTD status, or miss signalling occasionally under heavy load.  If the host
machine is too fast, we can miss transaction completion - when we scan the
active list the transaction still seems to be active. This generally exhibits
itself as a umass stall that never recovers.

We work around this behaviour by setting up this callback after any softintr
that completes with transactions still pending, giving us another chance to
check for completion after the writeback has taken place

Submitted by: Alexander Nedotsuko
MFC after: 3 days

14 years agoehci_init() will do reset and set the usbrev flag. Fix problem where
thompsa [Sun, 8 Nov 2009 20:44:55 +0000 (20:44 +0000)]
ehci_init() will do reset and set the usbrev flag.  Fix problem where
ehci_reset() was called before ehci_init().

PR: usb/140242
Submitted by: Sebastian Huber

14 years ago- fix refcounting error during data transfer
thompsa [Sun, 8 Nov 2009 20:03:52 +0000 (20:03 +0000)]
- fix refcounting error during data transfer
- fix a memory leak on the USB backend
- fix invalid pointer computations (in one case memory outside the allocated
  area was written in LibUSB v1.0)
- make sure memory is always initialised, also in failing cases
- add missing functions from v1.0.4

PR: usb/140325
Reported by: Robert Jenssen
Submitted by: Hans Petter Selasky
MFC After: 3 days

14 years agoPartially revert r199035.
yongari [Sun, 8 Nov 2009 19:59:54 +0000 (19:59 +0000)]
Partially revert r199035.
Revision 1.158 says only lower ten bits of
BGE_RXLP_LOCSTAT_IFIN_DROPS register is valid. For BCM5761 case it
seems the controller maintains 16bits value for the register.
However 16bits are still too small to count all dropped packets
happened in a second. To get a correct counter we have to read the
register in bge_rxeof() which would be too expensive.

Pointed out by: bde

14 years agoAdd a check for the connection being shut down to the krpc
rmacklem [Sun, 8 Nov 2009 19:02:13 +0000 (19:02 +0000)]
Add a check for the connection being shut down to the krpc
client just before queuing a request for the connection. The
code already had a check for the connection being shut down
while the request was queued, but not one for the shut down
having been initiated by the server before the request was
in the queue. This appears to fix the problem of slow reconnects
against an NFS server that drops inactive connections reported
by Olaf Seibert, but does not fix the case
where the FreeBSD client generates RST segments at about the
same time as ACKs. This is still a problem that is being
investigated. This patch does not cause a regression for this
case.

Tested by: Olaf Seibert, Daniel Braniss
Reviewed by: dfr
MFC after: 5 days

14 years agoIntroduce define and kernel option ATA_REQUEST_TIMEOUT to control ATA(4)
mav [Sun, 8 Nov 2009 14:33:19 +0000 (14:33 +0000)]
Introduce define and kernel option ATA_REQUEST_TIMEOUT to control ATA(4)
command timeout.

Submitted by: keramida

14 years agoFix a copy+paste error by checking the correct variable against MM_NULLACT.
brueffer [Sun, 8 Nov 2009 14:02:54 +0000 (14:02 +0000)]
Fix a copy+paste error by checking the correct variable against MM_NULLACT.

PR: 140386
Submitted by: soulcatcher <soulcatcher13@gmail.com
MFC after: 1 week

14 years ago- Strip trailing CRs
gabor [Sun, 8 Nov 2009 11:55:03 +0000 (11:55 +0000)]
- Strip trailing CRs

Requested by: Alex Kozlov <spam@rm-rf.kiev.ua> (via private mail)