]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
11 years agoRedo r242889, now using the method from projects/amd64_xen_pv r240747.
Dimitry Andric [Mon, 12 Nov 2012 22:28:32 +0000 (22:28 +0000)]
Redo r242889, now using the method from projects/amd64_xen_pv r240747.

Reminded by: kib
MFC after: 3 days

11 years agoRedo r242842, now actually fixing the warnings, as follows:
Dimitry Andric [Mon, 12 Nov 2012 22:01:29 +0000 (22:01 +0000)]
Redo r242842, now actually fixing the warnings, as follows:
- In sys/ofed/drivers/infiniband/core/cma.c, an enum struct member is
  interpreted as an int, so cast it to an int.
- In sys/ofed/drivers/infiniband/core/ud_header.c, initialize the
  packet_length variable in ib_ud_header_init(), to prevent undefined
  behaviour.
- In sys/ofed/drivers/infiniband/ulp/sdp/sdp_rx.c, call rdma_notify()
  with the correct enum type and value.
- In sys/ofed/include/linux/pci.h, change the PCI_DEVICE and PCI_VDEVICE
  macros to use C99 struct initializers, so additional members can be
  overridden.

Reviewed by: delphij, Garrett Cooper <yanegomi@gmail.com>
MFC after: 1 week

11 years agoFix a minor warning in sys/i386/xen/clock.c.
Dimitry Andric [Mon, 12 Nov 2012 20:50:11 +0000 (20:50 +0000)]
Fix a minor warning in sys/i386/xen/clock.c.

MFC after: 3 days

11 years agoMerge of vendor import of 242925, tzdata2012i
Edwin Groothuis [Mon, 12 Nov 2012 20:44:59 +0000 (20:44 +0000)]
Merge of vendor import of 242925, tzdata2012i

Update to tzdata2012i:
- Cuba is changing back to Standard Time on 4 November 2012.

11 years agoVendor import of tzdata2012i
Edwin Groothuis [Mon, 12 Nov 2012 20:41:02 +0000 (20:41 +0000)]
Vendor import of tzdata2012i

- Cuba is changing back to Standard Time on 4 November 2012.

Obtained from: ftp://ftp.iana.org/tz/releases/

11 years ago - Fix a bug that has existed since the original softdep implementation.
Jeff Roberson [Mon, 12 Nov 2012 19:53:55 +0000 (19:53 +0000)]
 - Fix a bug that has existed since the original softdep implementation.
   When a background copy of a cg is written we complete any work associated
   with that bmsafemap.  If new work has been added to the non-background
   copy of the buffer it will be completed before the next write happens.
   The solution is to do the rollbacks when we make the copy so only those
   dependencies that were present at the time of writing will be completed
   when the background write completes.  This would've resulted in various
   bitmap related corruptions and panics.  It also would've expired journal
   entries early causing journal replay to miss some records.

MFC after: 2 weeks

11 years agoFix a stack leak in [unused] cycle_menuitem function while we're here
Devin Teske [Mon, 12 Nov 2012 18:38:54 +0000 (18:38 +0000)]
Fix a stack leak in [unused] cycle_menuitem function while we're here
(required misconfiguration and/or missing environment vars to occur).

Reviewed by: peterj, adrian (co-mentor)
Approved by: adrian (co-mentor)

11 years agoUpdate MAX_TASKS to scale a bit based on MAXCPU
Sean Bruno [Mon, 12 Nov 2012 18:09:25 +0000 (18:09 +0000)]
Update MAX_TASKS to scale a bit based on MAXCPU

This alleviates issues on newer Sandy/Ivy Bridge gear that seems to require
boatloads more ACPI resources than before.

Reviewed by: avg@
Obtained from: Yahoo! Inc.
MFC after: 2 weeks

11 years agoClarify where the authorized_key file lives.
Eitan Adler [Mon, 12 Nov 2012 14:16:39 +0000 (14:16 +0000)]
Clarify where the authorized_key file lives.

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

11 years agoAdd new USB IDs.
Hans Petter Selasky [Mon, 12 Nov 2012 07:25:51 +0000 (07:25 +0000)]
Add new USB IDs.

MFC after: 1 week
PR: usb/173503

11 years agoAllow this file to be used in LOCORE sections of the kernel.
Rui Paulo [Mon, 12 Nov 2012 06:15:54 +0000 (06:15 +0000)]
Allow this file to be used in LOCORE sections of the kernel.

11 years agoFix DDB command "show map XXX":
Attilio Rao [Mon, 12 Nov 2012 00:30:40 +0000 (00:30 +0000)]
Fix DDB command "show map XXX":
- Check that an argument is always available, otherwise current map
  printing before to recurse is garbage.
- Spit out a message if an argument is not provided.
- Remove unread nlines variable.
- Use an explicit recursive function, disassociated from the
  DB_SHOW_COMMAND() body, in order to make clear prototype and recursion
  of the above mentioned function.  The code results now much less
  obscure.

Submitted by: gianni

11 years agoTweak comments.
Attilio Rao [Sun, 11 Nov 2012 23:25:47 +0000 (23:25 +0000)]
Tweak comments.

In collabouration with: alc

11 years agoCorrect date of Stanley's encounter with Livingstone.
Greg Lehey [Sun, 11 Nov 2012 22:43:36 +0000 (22:43 +0000)]
Correct date of Stanley's encounter with Livingstone.

Obtained from: Henry Morton Stanley, "How I met Livingstone", http://www.gutenberg.org/dirs/5/1/5/5157/5157-h/5157-h.htm
MFC after: 14 days

11 years agoCorrectly fix the 'scan during STA mode' crash.
Adrian Chadd [Sun, 11 Nov 2012 21:58:18 +0000 (21:58 +0000)]
Correctly fix the 'scan during STA mode' crash.

11 years agoRemove this; i incorrectly committed the wrong (debug) changes in my
Adrian Chadd [Sun, 11 Nov 2012 21:57:18 +0000 (21:57 +0000)]
Remove this; i incorrectly committed the wrong (debug) changes in my
previous commit.

11 years agoPrevent including .zfs snapshot directories in the src.txz
Glen Barber [Sun, 11 Nov 2012 21:52:18 +0000 (21:52 +0000)]
Prevent including .zfs snapshot directories in the src.txz
distribution.  This can happen if the src/ tree checkout is
within its own ZFS dataset, and the 'snapdir' ZFS property
is set to 'visible.'

Approved by: hrs
MFC after: 3 days
X-MFC-To: stable/9 only

11 years agoAdd the standard exit status to the ssh-copy-id man page.
Eitan Adler [Sun, 11 Nov 2012 15:34:58 +0000 (15:34 +0000)]
Add the standard exit status to the ssh-copy-id man page.

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

11 years agosh: Forward-declare struct alias instead of giving up type safety via void *
Jilles Tjoelker [Sun, 11 Nov 2012 15:13:24 +0000 (15:13 +0000)]
sh: Forward-declare struct alias instead of giving up type safety via void *

11 years agoAdd an explanatory comment to lib/libc/gen/isnan.c about the fix to make
Dimitry Andric [Sun, 11 Nov 2012 13:28:04 +0000 (13:28 +0000)]
Add an explanatory comment to lib/libc/gen/isnan.c about the fix to make
static linking with libc and libm work.

Requested by: jilles
MFC after: 1 week
X-MFC-With: 242879

11 years agoFix a typo.
Roman Divacky [Sun, 11 Nov 2012 10:45:21 +0000 (10:45 +0000)]
Fix a typo.

11 years agoChange the XNB_ASSERT from a statement expression to do-while(0) as its
Roman Divacky [Sun, 11 Nov 2012 10:42:34 +0000 (10:42 +0000)]
Change the XNB_ASSERT from a statement expression to do-while(0) as its
result is never used.

11 years agoAdd "pid" to the help menu (sort keys section).
Rui Paulo [Sun, 11 Nov 2012 08:22:58 +0000 (08:22 +0000)]
Add "pid" to the help menu (sort keys section).

11 years agoAdd the PID column to the list of sort keys.
Rui Paulo [Sun, 11 Nov 2012 08:16:33 +0000 (08:16 +0000)]
Add the PID column to the list of sort keys.

11 years agoDon't call av_set_tim() if it's NULL.
Adrian Chadd [Sun, 11 Nov 2012 00:34:10 +0000 (00:34 +0000)]
Don't call av_set_tim() if it's NULL.

This happens during a scan in STA mode; any queued data frames will
be power save queued but as there's no TIM in STA mode, it panics.

This was introduced by me when I disabled my driver-aware power save
handling support.

11 years agoCorrect some rather weird and broken behaviour observed when doing
Adrian Chadd [Sat, 10 Nov 2012 22:37:06 +0000 (22:37 +0000)]
Correct some rather weird and broken behaviour observed when doing
actual traffic with an AR9380/AR9382/AR9485.

The sample rate control stats would show impossibly large numbers for
"successful packets transmitted."  The number was a tad under 2^^64-1.
So after a bit of digging, I found that the sample rate control code
was making 'tries' turn into a negative number.. and this was because
ts_longretry was too small.

The hardware returns "ts_longretry" at the current rate selection,
not overall for that TX descriptor.  So if you setup four TX rate
scenarios and the second one works, ts_longretry is only set for
the number of attempts at that second rate scenario.  The FreeBSD HAL
code does the correction in ath_hal_proctxdesc() - however, this isn't
possible with EDMA.

EDMA TX completion is done separate from the original TX descriptor.
So the real solution is to split out "find ts_rate and ts_longretry"
from "complete TX descriptor".  Until that's done, put a hack in
the EDMA TX path that uses the rate scenario information in the ath_buf.

Tested: AR9380, AR9382, AR9485 STA mode

11 years agoOnly define isnan, isnanf, __isnan and __isnanf in libc.so, not in
Dimitry Andric [Sat, 10 Nov 2012 21:22:10 +0000 (21:22 +0000)]
Only define isnan, isnanf, __isnan and __isnanf in libc.so, not in
libc.a and libc_p.a.  In addition, define isnan in libm.a and libm_p.a,
but not in libm.so.

This makes it possible to statically link executables using both isnan
and isnanf with libc and libm.

Tested by: kargl
MFC after: 1 week

11 years ago- Protect mnt_data and mnt_flags under the mount interlock
Attilio Rao [Sat, 10 Nov 2012 19:32:16 +0000 (19:32 +0000)]
- Protect mnt_data and mnt_flags under the mount interlock
- Move mp->mnt_stat manipulation where all of them happens

Reported by: davide
Discussed with: kib
Tested by: flo
MFC after: 2 months
X-MFC: 241519, 242536,242616, 242727

11 years agoWork around pc98 tinderbox failures in sys/boot/pc98, by making sure a
Dimitry Andric [Sat, 10 Nov 2012 16:08:21 +0000 (16:08 +0000)]
Work around pc98 tinderbox failures in sys/boot/pc98, by making sure a
cross gcc gets built during the cross-tools stage.

MFC after: 1 week
X-MFC-With: 242706

11 years agoFix all the inconsistent nve_os* function declarations and definitions
Dimitry Andric [Sat, 10 Nov 2012 16:02:12 +0000 (16:02 +0000)]
Fix all the inconsistent nve_os* function declarations and definitions
in nve(4).  The OS_API structure defined in os.h expects NV_API_CALL
attributes, effectively regparm(0), on its function pointers, but all
the functions were declared and defined without this attribute.

MFC after: 1 week

11 years agos/ATH_DEBUG/ATH_DEBUG_ALQ
Kevin Lo [Sat, 10 Nov 2012 15:21:39 +0000 (15:21 +0000)]
s/ATH_DEBUG/ATH_DEBUG_ALQ

11 years agoUse ANSI prototype to fix build with clang.
Yoshihiro Takahashi [Sat, 10 Nov 2012 14:58:06 +0000 (14:58 +0000)]
Use ANSI prototype to fix build with clang.

MFC after: 1 week

11 years agoMFi386: r232263
Yoshihiro Takahashi [Sat, 10 Nov 2012 14:56:35 +0000 (14:56 +0000)]
MFi386: r232263

  Fix build mpboot.s with clang.

11 years agoReduce diffs against i386.
Yoshihiro Takahashi [Sat, 10 Nov 2012 13:48:41 +0000 (13:48 +0000)]
Reduce diffs against i386.

11 years agoFix some KASSERTs.
Yoshihiro Takahashi [Sat, 10 Nov 2012 13:41:25 +0000 (13:41 +0000)]
Fix some KASSERTs.
They are missing changes from r208833, r227394 and r227442.

11 years agoMFi386: r211924
Yoshihiro Takahashi [Sat, 10 Nov 2012 13:25:46 +0000 (13:25 +0000)]
MFi386: r211924

  Register an interrupt vector for DTrace return probes.

11 years agoUse ANSI prototype to fix build with clang.
Yoshihiro Takahashi [Sat, 10 Nov 2012 13:24:33 +0000 (13:24 +0000)]
Use ANSI prototype to fix build with clang.

MFC after: 1 week

11 years agocosmetic changes to reduce diffs against i386.
Yoshihiro Takahashi [Sat, 10 Nov 2012 12:42:28 +0000 (12:42 +0000)]
cosmetic changes to reduce diffs against i386.

11 years agoMFi386: r241300
Yoshihiro Takahashi [Sat, 10 Nov 2012 12:38:06 +0000 (12:38 +0000)]
MFi386: r241300

  i386 comconsole: don't loop forever if hardware doesn't respond

    - clear capability flags when hw timeouts
    - retire comc_started status variable and directly use c_flags to see
      if comconsole is selected for use

11 years agoMFi386: r241301
Yoshihiro Takahashi [Sat, 10 Nov 2012 12:25:27 +0000 (12:25 +0000)]
MFi386: r241301

  add detection of serial console presence to btx and boot2-like blocks

11 years agozfs_ioc_destroy_snaps_nvl: remove disk device entries for zvol snapshots
Andriy Gapon [Sat, 10 Nov 2012 12:22:26 +0000 (12:22 +0000)]
zfs_ioc_destroy_snaps_nvl: remove disk device entries for zvol snapshots

... before trying to destroy the zvol snapshots themselves.

PR: kern/173442
Reported by: Petri Helenius <petri@helenius.fi>,
mm
Obtained from: Brian Behlendorf <behlendorf1@llnl.gov>,
Illumos Bug #3170
Tested by: Petri Helenius <petri@helenius.fi>
MFC after: 10 days

11 years agoMFi386: r241785
Yoshihiro Takahashi [Sat, 10 Nov 2012 11:55:52 +0000 (11:55 +0000)]
MFi386: r241785

  boot: use -march=i386 for both i386 and amd64 builds

11 years agoInitialize hdrlen to 0 to avoid clang warning in NOINET case.
Roman Divacky [Sat, 10 Nov 2012 10:41:00 +0000 (10:41 +0000)]
Initialize hdrlen to 0 to avoid clang warning in NOINET case.

11 years agoFix the build.
Kevin Lo [Sat, 10 Nov 2012 08:34:40 +0000 (08:34 +0000)]
Fix the build.

11 years agoSeveral optimizations to sched_idletd():
Alexander Motin [Sat, 10 Nov 2012 07:02:57 +0000 (07:02 +0000)]
Several optimizations to sched_idletd():
 - Do not try to steal load from other CPUs if there was no contest switches
on this CPU (i.e. it was idle all the time and woke up just for bus mastering
or TLB shutdown). If current CPU was idle, then it is quite unlikely that some
other CPU has load to steal.  Under high I/O rate, when TLB shutdowns cause
numerous CPU wakeups, on 24-CPU system load stealing code may consume up to
25% of all CPU time without giving any benefits.
 - Change code that implements spinning for load to restart spin in case of
context switch.  Previous code periodically called cpu_idle() even under
high interrupt/context switch rate.
 - Rise spinning threshold to 10KHz, where it gives at least some effect
that may worth consumed power.

Reviewed by: jeff@

11 years agoMore style(9) tabs vs. spaces:
Greg Lehey [Sat, 10 Nov 2012 06:10:07 +0000 (06:10 +0000)]
More style(9) tabs vs. spaces:
- tabs after #define
- Not in comments.

MFC after: 2 weeks

11 years agoCorrect date and spelling of encounter between Stanley and
Greg Lehey [Sat, 10 Nov 2012 04:49:09 +0000 (04:49 +0000)]
Correct date and spelling of encounter between Stanley and
Livingstone.

MFC after: 2 weeks

11 years agoBe a bit more paranoid.
Eitan Adler [Sat, 10 Nov 2012 03:44:08 +0000 (03:44 +0000)]
Be a bit more paranoid.
Use more portable constructs in order to allow upstream adoption
Add per-file error messages

Reviewed by: jilles
Approved by: cperciva
MFC after: 1 week

11 years agoAllow maxusers to scale on machines with large address space.
Alfred Perlstein [Sat, 10 Nov 2012 02:08:40 +0000 (02:08 +0000)]
Allow maxusers to scale on machines with large address space.

Some hooks are added to clamp down maxusers and nmbclusters for
small address space systems.

VM_MAX_AUTOTUNE_MAXUSERS - the max maxusers that will be autotuned based on
physical memory.
VM_MAX_AUTOTUNE_NMBCLUSTERS - max nmbclusters based on physical memory.

These are set to the old values on i386 to preserve the clamping that was
being done to all arches.

Another macro VM_AUTOTUNE_NMBCLUSTERS is provided to allow an override
for the calculation on a MD basis.  Currently no arch defines this.

Reviewed by: peter
MFC after: 2 weeks

11 years agoMFV r242729 (mm):
Xin LI [Sat, 10 Nov 2012 01:52:52 +0000 (01:52 +0000)]
MFV r242729 (mm):

Illumos r13840:97fd5cdf328a:

3145 single-copy arc
3212 ztest: race condition between vdev_online() and spa_vdev_remove()

Illumos r13849:3468a95b27cd:

3258 ztest's use of file descriptors is unstable

11 years agoImport jemalloc 3.2.0.
Jason Evans [Sat, 10 Nov 2012 01:46:13 +0000 (01:46 +0000)]
Import jemalloc 3.2.0.

11 years agoAttempt toward a buildable universe by silenting a few warnings for OFED.
Xin LI [Sat, 10 Nov 2012 00:32:47 +0000 (00:32 +0000)]
Attempt toward a buildable universe by silenting a few warnings for OFED.

11 years agoUse %s when calling make_dev with a string pointer. This makes
Xin LI [Fri, 9 Nov 2012 21:41:07 +0000 (21:41 +0000)]
Use %s when calling make_dev with a string pointer.  This makes
clang happy.

MFC after: 2 weeks

11 years agoUndo over-aggressive conversion of spaces to tabs. ie: those within
Peter Wemm [Fri, 9 Nov 2012 20:19:56 +0000 (20:19 +0000)]
Undo over-aggressive conversion of spaces to tabs. ie: those within
format strings, "period, space, space" in comment text, etc.

11 years agoReduce LLVM's default stack alignment for i386 from 16 to 4 bytes, as
Dimitry Andric [Fri, 9 Nov 2012 18:56:27 +0000 (18:56 +0000)]
Reduce LLVM's default stack alignment for i386 from 16 to 4 bytes, as
the FreeBSD ABI requires.  This is essentially a revert of upstream llvm
commit r126226, and it will be reverted by upstream too.

MFC after: 1 week

11 years agoSimplify sending keepalives.
Alexander V. Chernikov [Fri, 9 Nov 2012 18:23:38 +0000 (18:23 +0000)]
Simplify sending keepalives.
Prepare ipfw_tick() to be used by other consumers.

Reviewed by: ae(basically)
MFC after: 2 weeks

11 years agoComplete MPSAFE VFS interface and remove MNTK_MPSAFE flag.
Attilio Rao [Fri, 9 Nov 2012 18:02:25 +0000 (18:02 +0000)]
Complete MPSAFE VFS interface and remove MNTK_MPSAFE flag.
Porters should refer to __FreeBSD_version 1000021 for this change as
it may have happened at the same timeframe.

11 years agoDont compile some files in drm with -finline-limit=1350. GCC 4.2.1 doesnt
Roman Divacky [Fri, 9 Nov 2012 17:46:07 +0000 (17:46 +0000)]
Dont compile some files in drm with -finline-limit=1350. GCC 4.2.1 doesnt
need that and clang does not support that. This effectively reverts r126542.

11 years agoFix LOW and FULL speed USB INTERRUPT endpoint support for the
Hans Petter Selasky [Fri, 9 Nov 2012 16:28:58 +0000 (16:28 +0000)]
Fix LOW and FULL speed USB INTERRUPT endpoint support for the
DWC OTG driver. Fix a hang issue when using LOW and FULL speed
BULK traffic. Make sure we don't ask for data in the last
microframe. This allows using devices like USB mice and USB
keyboards connected to the RPI-B.

Suggested by: gonzo @

11 years agoDo not try to enable new features in the %cr4 if running under
Konstantin Belousov [Fri, 9 Nov 2012 16:00:30 +0000 (16:00 +0000)]
Do not try to enable new features in the %cr4 if running under
hypervisor.  Apparently, hypervisors failed to filter out 'Standard
Extended Features' report from CPUID, but deliver #gp when
corresponding bit in %cr4 is toggled.

This shall be reconsidered later, after hypervisors correct the bug.

Reported and tested by: joel
Reviewed by: avg
MFC after: 2 weeks

11 years agoGive tw_osl_dbg_printf format string to pacify clang warning.
Roman Divacky [Fri, 9 Nov 2012 15:29:52 +0000 (15:29 +0000)]
Give tw_osl_dbg_printf format string to pacify clang warning.

11 years agoGive device_printf format string to pacify clang warning.
Roman Divacky [Fri, 9 Nov 2012 14:58:26 +0000 (14:58 +0000)]
Give device_printf format string to pacify clang warning.

11 years agoCast VIA_READ to (void) where the result is unused.
Roman Divacky [Fri, 9 Nov 2012 14:46:23 +0000 (14:46 +0000)]
Cast VIA_READ to (void) where the result is unused.

11 years agoRemove unused static mn_timeout.
Roman Divacky [Fri, 9 Nov 2012 14:14:51 +0000 (14:14 +0000)]
Remove unused static mn_timeout.

11 years agoGive panic format string to pacify clang warning.
Roman Divacky [Fri, 9 Nov 2012 13:58:52 +0000 (13:58 +0000)]
Give panic format string to pacify clang warning.

11 years agoBelatedly (by many years!) bump the version numbers in the src/-side
Glen Barber [Fri, 9 Nov 2012 04:52:15 +0000 (04:52 +0000)]
Belatedly (by many years!) bump the version numbers in the src/-side
documentation pages.

Saddened by: this file

11 years ago - Correct rev 242734, segments can sometimes get stuck. Be a bit more
Jeff Roberson [Fri, 9 Nov 2012 04:04:25 +0000 (04:04 +0000)]
 - Correct rev 242734, segments can sometimes get stuck.  Be a bit more
   defensive with segment state.

Reported by:  b. f. <bf1783@googlemail.com>

11 years agoAdd support for Advantech PCI-1602 RS-485/RS-422 serial card
Eitan Adler [Fri, 9 Nov 2012 01:51:06 +0000 (01:51 +0000)]
Add support for Advantech PCI-1602 RS-485/RS-422 serial card

PR: kern/169726
Submitted by: Jan Mikkelsen <janm@transactionware.com>
Approved by: cperciva (implicit)
MFC after: 5 days

11 years agoFix a very incorrect description.
Adrian Chadd [Fri, 9 Nov 2012 01:28:11 +0000 (01:28 +0000)]
Fix a very incorrect description.

11 years agoReplace spaces by tabs where appropriate.
Greg Lehey [Thu, 8 Nov 2012 23:46:15 +0000 (23:46 +0000)]
Replace spaces by tabs where appropriate.

Reminded by: jh@

11 years agoReplace spaces with tabs where appropriate.
Greg Lehey [Thu, 8 Nov 2012 23:45:19 +0000 (23:45 +0000)]
Replace spaces with tabs where appropriate.

Reminded by: jh@

11 years agoShrink boot2 by 8 bytes, by eliminating some unneeded instructions in
Dimitry Andric [Thu, 8 Nov 2012 23:21:02 +0000 (23:21 +0000)]
Shrink boot2 by 8 bytes, by eliminating some unneeded instructions in
sio.S.  This is not particularly needed for head right now, but it is
intended to merge to stable/9, to fix boot2 build with clang there.

Reviewed by: avg
MFC after: 3 days

11 years agoFix the build - fix up the ath_alq code to not compile by default.
Adrian Chadd [Thu, 8 Nov 2012 23:11:59 +0000 (23:11 +0000)]
Fix the build - fix up the ath_alq code to not compile by default.

11 years agoNot only load pccard_ether settings, also load network settings. This
Nick Hibma [Thu, 8 Nov 2012 20:34:12 +0000 (20:34 +0000)]
Not only load pccard_ether settings, also load network settings. This
is only a problem when a /etc/rc.conf.d/network file is being used.

PR: conf/160373
Submitted by: n_hibma
MFC after: 1 week

11 years agoAdd some hooks into the driver to attach, detach and record EDMA descriptor
Adrian Chadd [Thu, 8 Nov 2012 18:11:31 +0000 (18:11 +0000)]
Add some hooks into the driver to attach, detach and record EDMA descriptor
events.

This is primarily for the TX EDMA and TX EDMA completion. I haven't yet
tied it into the EDMA RX path or the legacy TX/RX path.

Things that I don't quite like:

* Make the pointer type 'void' in ath_softc and have if_ath_alq*()
  return a malloc'ed buffer.  That would remove the need to include
  if_ath_alq.h in if_athvar.h.
* The sysctl setup needs to be cleaned up.

11 years agoAdd my initial cut at driver-layer ALQ support.
Adrian Chadd [Thu, 8 Nov 2012 18:07:29 +0000 (18:07 +0000)]
Add my initial cut at driver-layer ALQ support.

I'm using this to debug EDMA TX and RX descriptors and it's really helpful
to have a non-printf() way to decode frames.

I won't link this into the build until I've tidied it up a little more.

This will eventually be behind ATH_DEBUG_ALQ.

11 years agoOops, fix bogus spacing.
Adrian Chadd [Thu, 8 Nov 2012 17:46:27 +0000 (17:46 +0000)]
Oops, fix bogus spacing.

11 years agoImplement the ATH_RESET_NOLOSS path for TX stop and start; this is needed
Adrian Chadd [Thu, 8 Nov 2012 17:43:58 +0000 (17:43 +0000)]
Implement the ATH_RESET_NOLOSS path for TX stop and start; this is needed
for 802.11n TX device restarting.

Remove the debug printf()s; they're no longer needed here.

11 years agoConvert this to a debug printf; it's working fine now.
Adrian Chadd [Thu, 8 Nov 2012 17:32:55 +0000 (17:32 +0000)]
Convert this to a debug printf; it's working fine now.

11 years agoMake the USB ethernet methods constant again in if_udav.c,
Hans Petter Selasky [Thu, 8 Nov 2012 16:31:13 +0000 (16:31 +0000)]
Make the USB ethernet methods constant again in if_udav.c,
so that both adapter variants can be attached at the same
time.

MFC after: 0 days

11 years agosh: Add tests for modifying an alias (r242766).
Jilles Tjoelker [Thu, 8 Nov 2012 13:36:19 +0000 (13:36 +0000)]
sh: Add tests for modifying an alias (r242766).

Note: parser/alias10.0 will eat a lot of memory/cpu time when it fails (with
the old sh).

11 years agosh: Fix two issues when an alias is redefined:
Jilles Tjoelker [Thu, 8 Nov 2012 13:33:48 +0000 (13:33 +0000)]
sh: Fix two issues when an alias is redefined:

 * The last character is not displayed.
 * If the alias ends with itself (as a word), an infinite memory-eating loop
   occurs.

If an alias is defined initially, a space is appended to avoid recursion but
this did not happen when an alias was later modified.

PR: bin/173418
Submitted by: Daniel F.
MFC after: 1 week

11 years agoAdd new USB device ID.
Kevin Lo [Thu, 8 Nov 2012 09:29:05 +0000 (09:29 +0000)]
Add new USB device ID.

11 years agoMake r242655 build on sparc64. While at it, make vm_{max,min}_kernel_address
Marius Strobl [Thu, 8 Nov 2012 08:10:32 +0000 (08:10 +0000)]
Make r242655 build on sparc64. While at it, make vm_{max,min}_kernel_address
vm_offset_t as they should be.

11 years agoReduce differences between these two initarms a bit more.
Warner Losh [Thu, 8 Nov 2012 04:02:36 +0000 (04:02 +0000)]
Reduce differences between these two initarms a bit more.

11 years agoCleanup some whitspace in this file to get it out of an upcoming patch.
Bjoern A. Zeeb [Thu, 8 Nov 2012 03:29:55 +0000 (03:29 +0000)]
Cleanup some whitspace in this file to get it out of an upcoming patch.

MFC after: 10 days

11 years agoZero the newly allocated md(4) swap-backed page to prevent random
Konstantin Belousov [Thu, 8 Nov 2012 03:17:41 +0000 (03:17 +0000)]
Zero the newly allocated md(4) swap-backed page to prevent random
kernel memory leakage to userspace. For the typical use, when a
filesystem put on the md disk, the change only results in CPU and
memory bandwidth spent to zero the page, since filsystems make sure
that user never see unwritten content.  But if md disk is used as raw
device by userspace, the garbage is exposed.

Reported by: Paul Schenkeveld <freebsd@psconsult.nl>
MFC after: 2 weeks

11 years agoMake parameters to -c and -k options optional. If no parameters are
Greg Lehey [Thu, 8 Nov 2012 02:55:30 +0000 (02:55 +0000)]
Make parameters to -c and -k options optional.  If no parameters are
supplied, print information for all keywords.

Improve output of -c option, in particular in conjunction with -k
option.

MFC after: 14 days

11 years agoClarify that the ' flag is an apostrophe.
Greg Lehey [Thu, 8 Nov 2012 02:01:04 +0000 (02:01 +0000)]
Clarify that the ' flag is an apostrophe.

MFC after: 2 weeks

11 years agoAdd DTrace to 32-bit PowerPC GENERIC now.
Justin Hibbits [Thu, 8 Nov 2012 01:51:23 +0000 (01:51 +0000)]
Add DTrace to 32-bit PowerPC GENERIC now.

MFC after: 1 month

11 years ago - Change ULE to use dynamic slice sizes for the timeshare queue in order
Jeff Roberson [Thu, 8 Nov 2012 01:46:47 +0000 (01:46 +0000)]
 - Change ULE to use dynamic slice sizes for the timeshare queue in order
   to further reduce latency for threads in this queue.  This should help
   as threads transition from realtime to timeshare.  The latency is
   bound to a max of sched_slice until we have more than sched_slice / 6
   threads runnable.  Then the min slice is allotted to all threads and
   latency becomes (nthreads - 1) * min_slice.

Discussed with: mav

11 years ago - Implement BIO_FLUSH support around journal entries. This will not 100%
Jeff Roberson [Thu, 8 Nov 2012 01:41:04 +0000 (01:41 +0000)]
 - Implement BIO_FLUSH support around journal entries.  This will not 100%
   solve power loss problems with dishonest write caches.  However, it
   should improve the situation and force a full fsck when it is unable
   to resolve with the journal.
 - Resolve a case where the journal could wrap in an unsafe way causing
   us to prematurely lose journal entries in very specific scenarios.

Discussed with: mckusick
MFC after: 1 month

11 years agoUse .PATH instead of VPATH.
Kevin Lo [Thu, 8 Nov 2012 01:35:02 +0000 (01:35 +0000)]
Use .PATH instead of VPATH.

Reviewed by: gnn,rwatson

11 years agoUpdate vendor/illumos/dist and vendor-sys/illumos/dist
Martin Matuska [Thu, 8 Nov 2012 01:24:06 +0000 (01:24 +0000)]
Update vendor/illumos/dist and vendor-sys/illumos/dist
to illumos-gate 13849:3468a95b27cd
(illumos ZFS issues #3145, #3212, #3258)

11 years agoICMPV6_FILTER should read ICMP6_FILTER.
Xin LI [Thu, 8 Nov 2012 00:54:43 +0000 (00:54 +0000)]
ICMPV6_FILTER should read ICMP6_FILTER.

Submitted by: Frédéric Perrin <frederic.perrin resel.fr>
MFC after: 2 weeks

11 years ago- Current caching mode is completely broken because it simply relies
Attilio Rao [Thu, 8 Nov 2012 00:32:49 +0000 (00:32 +0000)]
- Current caching mode is completely broken because it simply relies
  on timing of the operations and not real lookup, bringing too many
  false positives. Remove the whole mechanism. If it needs to be
  implemented, next time it should really be done in the proper way.
- Fix VOP_GETATTR() in order to cope with userland bugs that would
  change the type of file and not panic. Instead it gets the entry as
  if it is not existing.

Reported and tested by: flo
MFC after: 2 months
X-MFC: 241519, 242536,242616

11 years agoAdd support for SCSI pass through devices to be attached and
Doug Ambrisko [Thu, 8 Nov 2012 00:32:36 +0000 (00:32 +0000)]
Add support for SCSI pass through devices to be attached and
detached.

PR: 172864
Submitted by: rstone@

11 years agoAdd y flag and environment variable LS_SAMESORT to specify the same
Greg Lehey [Thu, 8 Nov 2012 00:24:26 +0000 (00:24 +0000)]
Add y flag and environment variable LS_SAMESORT to specify the same
sorting order for time and name with the -t option.  IEEE Std 1003.2
(POSIX.2) mandates that the -t option sort in descending order, and
that if two files have the same timestamp, they should be sorted in
ascending order of their names.  The -r flag reverses both of these
sort orders, so they're never the same.  This creates significant
problems for sequentially named files stored on FAT file systems,
where it can be impossible to list them in the order in which they
were created.

Add , (comma) option to print file sizes grouped and separated by
thousands using the non-monetary separator returned by localeconv(3),
typically a comma or period.

MFC after:  14 days

11 years agoUpdate AP96 to directly attach an arswitch.
Adrian Chadd [Wed, 7 Nov 2012 23:50:28 +0000 (23:50 +0000)]
Update AP96 to directly attach an arswitch.

11 years agoImplement DTrace for PowerPC. This includes both 32-bit and 64-bit.
Justin Hibbits [Wed, 7 Nov 2012 23:45:09 +0000 (23:45 +0000)]
Implement DTrace for PowerPC.  This includes both 32-bit and 64-bit.

There is one known issue:  Some probes will display an error message along the
lines of:  "Invalid address (0)"

I tested this with both a simple dtrace probe and dtruss on a few different
binaries on 32-bit.  I only compiled 64-bit, did not run it, but I don't expect
problems without the modules loaded.  Volunteers are welcome.

MFC after: 1 month