]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
11 years agoAdd a new (skeleton) spectral mode manager module.
Adrian Chadd [Wed, 2 Jan 2013 03:59:02 +0000 (03:59 +0000)]
Add a new (skeleton) spectral mode manager module.

11 years agoFix the short repeat option code to not flip the option to 0 when
Adrian Chadd [Wed, 2 Jan 2013 03:56:20 +0000 (03:56 +0000)]
Fix the short repeat option code to not flip the option to 0 when
we call this w/ NOVAL set.

11 years agoAdd spectral HAL accessor methods.
Adrian Chadd [Wed, 2 Jan 2013 01:40:23 +0000 (01:40 +0000)]
Add spectral HAL accessor methods.

11 years agoAdd a method to explicitly disable radar reporting if required.
Adrian Chadd [Wed, 2 Jan 2013 01:36:10 +0000 (01:36 +0000)]
Add a method to explicitly disable radar reporting if required.

11 years agoAdd missing update to link ar5416_spectral.c into the build.
Adrian Chadd [Wed, 2 Jan 2013 01:24:36 +0000 (01:24 +0000)]
Add missing update to link ar5416_spectral.c into the build.

11 years agoBring over the basic spectral scan framework code from Qualcomm Atheros.
Adrian Chadd [Wed, 2 Jan 2013 00:38:01 +0000 (00:38 +0000)]
Bring over the basic spectral scan framework code from Qualcomm Atheros.

This includes the HAL routines to setup, enable/activate/disable spectral
scan and configure the relevant registers.

This still requires driver interaction to enable spectral scan reporting.
Specifically:

* call ah_spectralConfigure() to configure and enable spectral scan;
* .. there's currently no way to disable spectral scan... that will have
  to follow.
* call ah_spectralStart() to force start a spectral report;
* call ah_spectralStop() to force stop an active spectral report.

The spectral scan results appear as PHY errors (type 0x5 on the AR9280,
same as radar) but with the spectral scan bit set (0x10 in the last byte
of the frame) identifying it as a spectral report rather than a radar
FFT report.

Caveats:

* It's likely quite difficult to run spectral _and_ radar at the same
  time.  Enabling spectral scan disables the radar thresholds but
  leaves radar enabled.  Thus, the driver (for now) needs to ensure
  that only one or the other is enabled.

* .. it needs testing on HT40 mode.

Tested:

* AR9280 in STA mode, HT/20 only

TODO:

* Test on AR9285, AR9287;
* Test in both HT20 and HT40 modes;
* .. all the driver glue.

Obtained from: Qualcomm Atheros

11 years agoMerge @219932 from Perforce:
Robert Watson [Tue, 1 Jan 2013 19:42:06 +0000 (19:42 +0000)]
Merge @219932 from Perforce:

  FDT headers can't be included if the kernel is compiled without
  FDT support, due to dependence on generated kobj headers.  BERI
  supports both FDT and non-FDT kernels.

  Spotted by: bz

11 years agolibedit: bind the correct command when using "bind -k".
Pedro F. Giffuni [Tue, 1 Jan 2013 19:20:51 +0000 (19:20 +0000)]
libedit: bind the correct command when using "bind -k".

"ed-argument-digit" (i. e. command 0) was incorrectly used
instead.
This bug comes from the original sources imported in 1994
and has been confirmed in upstream NetBSD.

Reported by: Yamagi Burmeister
Submitted by: Christoph Mallon
MFC after: 3 days

11 years agoOverhauled CPSW driver for TI CPSW Ethernet module
Tim Kientzle [Tue, 1 Jan 2013 18:55:04 +0000 (18:55 +0000)]
Overhauled CPSW driver for TI CPSW Ethernet module
(as used in AM335x SoC for BeagleBone).

Among other things:
 * Watchdog reset doesn't hang the driver.
 * Disconnecting cable doesn't hang the driver.
 * ifconfig up/down doesn't hang the driver
 * Out-of-memory no longer panics the driver.

Known issues:
 * Doesn't have good support for fragmented packets
   (calls m_defrag() on TX, assumes RX packets are never fragmented)
 * Promisc and allmulti still unimplimented
 * addmulti and delmulti still unimplemented
 * TX queue still stalls (but watchdog now consistently recovers in ~5s)
 * No sysctl monitoring
 * Only supports port0
 * No switch configuration support
 * Not tested on anything but BeagleBone

Committed from: BeagleBone

11 years agoFix a dependency
Antoine Brodin [Tue, 1 Jan 2013 18:31:21 +0000 (18:31 +0000)]
Fix a dependency

11 years agoAdd missing DPADD
Antoine Brodin [Tue, 1 Jan 2013 18:29:25 +0000 (18:29 +0000)]
Add missing DPADD

11 years agopflogd does not depend on libutil
Antoine Brodin [Tue, 1 Jan 2013 18:28:21 +0000 (18:28 +0000)]
pflogd does not depend on libutil

11 years agoAdd missing DPADD.
Antoine Brodin [Tue, 1 Jan 2013 18:25:14 +0000 (18:25 +0000)]
Add missing DPADD.

11 years agoRemove unneeded semicolons.
Antoine Brodin [Tue, 1 Jan 2013 18:16:49 +0000 (18:16 +0000)]
Remove unneeded semicolons.

Reviewed by: md5 of the object files

11 years agoThe process_deferred_inactive() function locks the vnodes of the ufs
Konstantin Belousov [Tue, 1 Jan 2013 16:14:48 +0000 (16:14 +0000)]
The process_deferred_inactive() function locks the vnodes of the ufs
mount, which means that is must not be called while the snaplock is
owned.  The vfs_write_resume(9) does call the function as the
VFS_SUSP_CLEAN() method, which is too early and falls into the region
still protected by snaplock.

Add yet another flag for the vfs_write_resume_flags() to avoid calling
suspension cleanup handler after the suspend is lifted, and use it in
the ffs_snapshot() call to vfs_write_resume.

Reported and tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks

11 years agosh: Fix a Subversion property.
Jilles Tjoelker [Tue, 1 Jan 2013 12:48:24 +0000 (12:48 +0000)]
sh: Fix a Subversion property.

Test scripts should have keywords; only pure output that must match exactly
should have keywords disabled.

11 years agoDocument the known values of the RTL release field in the cache is register
Andrew Turner [Tue, 1 Jan 2013 03:48:39 +0000 (03:48 +0000)]
Document the known values of the RTL release field in the cache is register

11 years agoAdd interrupt for PL310 controller
Oleksandr Tymoshenko [Mon, 31 Dec 2012 22:48:54 +0000 (22:48 +0000)]
Add interrupt for PL310 controller

11 years agoExplicitly specify that the beforelinking target depends on the
Mark Johnston [Mon, 31 Dec 2012 21:54:43 +0000 (21:54 +0000)]
Explicitly specify that the beforelinking target depends on the
generated object files, ensuring that the beforelinking recipe won't be
executed until compilation has finished.

Also define SHLIB_NAME_FULL to denote ${SHLIB_NAME}.debug if
DEBUG_FILES is set and ${SHLIB_NAME} otherwise, which helps avoid
obfuscating the compilation and linking rules.

Reviewed by: emaste
Approved by: emaste (co-mentor)

11 years agoPL310 driver update:
Oleksandr Tymoshenko [Mon, 31 Dec 2012 21:19:44 +0000 (21:19 +0000)]
PL310 driver update:

- Add pl310.disable tunable to disable L2 cache altogether. In
    order to make sure that it's 100% disabled we use cache event
    counters for cache line eviction and read allocate events
    and panic if any of these counters increased. This is purely
    for debugging purpose
- Direct access DEBUG_CTRL and CTRL might be unavailable in
    unsecure mode, so use platform-specific functions for
    these registers
- Replace #if 1 with proper erratum numbers
- Add erratum 753970 workaround
- Remove wait function for atomic operations
- Protect cache operations with spin mutex in order to prevent race condition
- Disable instruction cache prefetch and make sure data cache
    prefetch is enabled in OMAP4-specific intialization

11 years agoMerge r234561 from busdma_machdep.c to ARMv6 version of busdma:
Oleksandr Tymoshenko [Mon, 31 Dec 2012 21:00:38 +0000 (21:00 +0000)]
Merge r234561 from busdma_machdep.c to ARMv6 version of busdma:

Interrupts must be disabled while handling a partial cache line flush,
as otherwise the interrupt handling code may modify data in the non-DMA
part of the cache line while we have it stashed away in the temporary
stack buffer, then we end up restoring a stale value.

PR:             160431
Submitted by:   Ian Lepore

11 years agoHappy New Year 2013!
Bjoern A. Zeeb [Mon, 31 Dec 2012 11:22:55 +0000 (11:22 +0000)]
Happy New Year 2013!

11 years agoIf FDT is compiled into a FreeBSD/beri kernel, initialise OpenFirmware.
Robert Watson [Mon, 31 Dec 2012 11:06:37 +0000 (11:06 +0000)]
If FDT is compiled into a FreeBSD/beri kernel, initialise OpenFirmware.

Sponsored by: DARPA, AFRL

11 years agoWhen a partially initialised FDT-based device instance can't attach,
Robert Watson [Sun, 30 Dec 2012 21:10:48 +0000 (21:10 +0000)]
When a partially initialised FDT-based device instance can't attach,
perhaps due to an interrupt configuration problem, do not try to free
device ivars that have not yet have been allocated.

MFC after: 1 week
Reviewed by: gonzo
Sponsored by: DARPA, AFRL

11 years agoWith the old sade removed, libdisk is no longer used by anything in HEAD
Nathan Whitehorn [Sun, 30 Dec 2012 16:28:06 +0000 (16:28 +0000)]
With the old sade removed, libdisk is no longer used by anything in HEAD
and uses a number of problematic pre-gpart interfaces. Since it has been
entirely obsoleted by interfaces in geom, remove it.

11 years ago- Use BN_set_negative() and BN_is_negative() instead of subtracting or
Kevin Lo [Sun, 30 Dec 2012 15:20:27 +0000 (15:20 +0000)]
- Use BN_set_negative() and BN_is_negative() instead of subtracting or
  comparing to zero.
- Fix fractional number exponentiation, especially for negative exponents.

Obtained from: OpenBSD

11 years agoAdd a man page for the new sade(8), which is mostly the same uninformative
Nathan Whitehorn [Sun, 30 Dec 2012 14:44:14 +0000 (14:44 +0000)]
Add a man page for the new sade(8), which is mostly the same uninformative
manpage as the old sade(8).

11 years agoReplace sade the extracted piece of sysinstall with sade the extracted
Nathan Whitehorn [Sun, 30 Dec 2012 14:35:00 +0000 (14:35 +0000)]
Replace sade the extracted piece of sysinstall with sade the extracted
piece of bsdinstall (although this time with a symlink instead of duplicated
source code).

Discussed on: freebsd-geom
MFC after: 3 months

11 years agoIf invoked as sade, don't complain about having a filesystem on / already
Nathan Whitehorn [Sun, 30 Dec 2012 14:33:16 +0000 (14:33 +0000)]
If invoked as sade, don't complain about having a filesystem on / already
-- it's something you expect (and indeed hope for).

MFC after: 1 week

11 years agoAllow to create pool even if mount point directory is not empty if -f is given.
Pawel Jakub Dawidek [Sun, 30 Dec 2012 13:47:40 +0000 (13:47 +0000)]
Allow to create pool even if mount point directory is not empty if -f is given.

Obtained from: WHEEL Systems

11 years agoAdd directories which were wrongly created during auditdistd addition.
Hajimu UMEMOTO [Sun, 30 Dec 2012 08:00:40 +0000 (08:00 +0000)]
Add directories which were wrongly created during auditdistd addition.

Requested by: netchild

11 years agoAdd the initial HAL glue for the spectral analysis support.
Adrian Chadd [Sun, 30 Dec 2012 06:48:12 +0000 (06:48 +0000)]
Add the initial HAL glue for the spectral analysis support.

* Finish adding the HAL capability to announce whether a NIC supports
  spectral scan or not;
* Add spectral scan methods to the HAL structure;
* Add HAL_SPECTRAL_PARAM for configuration of the spectral scan logic.

The capability ID and HAL_SPECTRAL_PARAM struct are from Qualcomm
Atheros.

11 years agoAdd spectral scan capability.
Adrian Chadd [Sun, 30 Dec 2012 06:46:11 +0000 (06:46 +0000)]
Add spectral scan capability.

11 years agoRevert r241818 that updated dialog to 20120706. This turns out to horribly
Nathan Whitehorn [Sun, 30 Dec 2012 04:22:34 +0000 (04:22 +0000)]
Revert r241818 that updated dialog to 20120706. This turns out to horribly
break mixed form dialogs in conjunction with the FreeBSD termcap, making
the bsdinstall partition editor Add dialog, among other things, completely
nonfunctional. This restores dialog 20110707.

11 years agoAdd new USB ID.
Hans Petter Selasky [Sat, 29 Dec 2012 21:22:38 +0000 (21:22 +0000)]
Add new USB ID.

PR: usb/174814
Submitted by: Andy Balholm
MFC after: 1 week

11 years agoMinor changes to force commit these files so new freebsd*.cf files are
Gregory Neil Shapiro [Sat, 29 Dec 2012 20:42:28 +0000 (20:42 +0000)]
Minor changes to force commit these files so new freebsd*.cf files are
built to use the new sendmail-8.14.6/cf tree.

While here, update DNSBL link once again.

MFC after: 4 days

11 years agoUpdate import info based on recent experiences and sendmail 8.14.6 import.
Gregory Neil Shapiro [Sat, 29 Dec 2012 20:38:57 +0000 (20:38 +0000)]
Update import info based on recent experiences and sendmail 8.14.6 import.

MFC after: 4 days

11 years agoMerge sendmail 8.14.6 to HEAD
Gregory Neil Shapiro [Sat, 29 Dec 2012 20:30:46 +0000 (20:30 +0000)]
Merge sendmail 8.14.6 to HEAD

MFC after: 4 days

11 years agoImport sendmail 8.14.6
Gregory Neil Shapiro [Sat, 29 Dec 2012 20:13:40 +0000 (20:13 +0000)]
Import sendmail 8.14.6

11 years agoAdd missing closing quote on commented out example
Gregory Neil Shapiro [Sat, 29 Dec 2012 19:57:52 +0000 (19:57 +0000)]
Add missing closing quote on commented out example

PR: bin/174108
Obtained from: Julian H. Stacey
MFC after: 1 day

11 years agoTest commit to make sure commit mail works after moving the server.
Peter Wemm [Sat, 29 Dec 2012 16:03:23 +0000 (16:03 +0000)]
Test commit to make sure commit mail works after moving the server.

Approved by: core (implicit)

11 years agoComments.
Devin Teske [Sat, 29 Dec 2012 00:30:30 +0000 (00:30 +0000)]
Comments.

11 years agoAdd missing enforcement of height restriction after minor adjustment.
Devin Teske [Sat, 29 Dec 2012 00:06:26 +0000 (00:06 +0000)]
Add missing enforcement of height restriction after minor adjustment.

11 years agoAdd support for running without a controlling terminal (for example, when
Devin Teske [Sat, 29 Dec 2012 00:01:24 +0000 (00:01 +0000)]
Add support for running without a controlling terminal (for example, when
running as an rvalue to a pipe).

11 years agoAllow debug output to be logged to a file (set $debugFile to target pathname)
Devin Teske [Fri, 28 Dec 2012 23:49:17 +0000 (23:49 +0000)]
Allow debug output to be logged to a file (set $debugFile to target pathname)
or both stdout and a file (precede $debugFile pathname with a plus-sign, `+').

11 years agoWhitespace.
Devin Teske [Fri, 28 Dec 2012 23:40:13 +0000 (23:40 +0000)]
Whitespace.

11 years agoFix incorrect function name. s/[[:<:]]dialog_clear/f_&/
Devin Teske [Fri, 28 Dec 2012 23:27:17 +0000 (23:27 +0000)]
Fix incorrect function name. s/[[:<:]]dialog_clear/f_&/

11 years agoIt's no longer necessary to explicitly specify isa bus.
Xin LI [Fri, 28 Dec 2012 23:21:54 +0000 (23:21 +0000)]
It's no longer necessary to explicitly specify isa bus.

MFC after: 2 weeks

11 years agoMake it possible to atomically resume writes on the mount and account
Konstantin Belousov [Fri, 28 Dec 2012 23:08:30 +0000 (23:08 +0000)]
Make it possible to atomically resume writes on the mount and account
the write start, by adding a variation of the vfs_write_resume(9)
which accepts flags.

Use the new function to prevent a deadlock between parallel suspension
and snapshotting a UFS mount.  The ffs_snapshot() code performed
vfs_write_resume() followed by vn_start_write() while owning the
snaplock.  If the suspension intervene between resume and
vn_start_write(), the deadlock occured after the suspending thread
tried to lock the snaplock, most typically during the write in the
ffs_copyonwrite().

Reported and tested by: Andreas Longwitz <longwitz@incore.de>
Reviewed by: mckusick
MFC after: 2 weeks
X-MFC-note: make the vfs_write_resume(9) function a macro after the MFC,
in HEAD

11 years agogcc: avoid generating negative values to DW_AT_byte_size.
Pedro F. Giffuni [Fri, 28 Dec 2012 22:06:50 +0000 (22:06 +0000)]
gcc: avoid generating negative values to DW_AT_byte_size.

The previous change was actually a NOP because size is
unsigned and compilers are dumb.

Submitted by: Christoph Mallon
MFC after: 1 week

11 years agoFix wording nit.
Glen Barber [Fri, 28 Dec 2012 22:06:33 +0000 (22:06 +0000)]
Fix wording nit.

PR: 174787
Submitted by: Cody Rank
MFC after: 3 days

11 years agoFix typo in comment.
Baptiste Daroussin [Fri, 28 Dec 2012 21:59:47 +0000 (21:59 +0000)]
Fix typo in comment.

Submitted by: Christoph Mallon <christoph.mallon@gmx.de>

11 years agoAdd DEVICE_IDENTIFY method for wbwd(4), required on most of recent
Xin LI [Fri, 28 Dec 2012 21:35:47 +0000 (21:35 +0000)]
Add DEVICE_IDENTIFY method for wbwd(4), required on most of recent
Supermicro motherboards.

Tested on X8STi and X8DTH boards.

Sponsored by: iXsystems, Inc.
MFC after: 2 weeks

11 years agoUse DEVMETHOD_END.
Xin LI [Fri, 28 Dec 2012 20:44:23 +0000 (20:44 +0000)]
Use DEVMETHOD_END.

MFC after: 2 weeks

11 years agoSimplify pointing dst after the end of all the gr_mem pointers in newgr
Baptiste Daroussin [Fri, 28 Dec 2012 20:44:10 +0000 (20:44 +0000)]
Simplify pointing dst after the end of all the gr_mem pointers in newgr

Submitted by: pjd
Reviewed by: db

11 years agoerrno = ENOMEM was supposed to be removed not return (NULL);
Baptiste Daroussin [Fri, 28 Dec 2012 20:30:04 +0000 (20:30 +0000)]
errno = ENOMEM was supposed to be removed not return (NULL);

Submitted by: gcooper

11 years agomalloc() sets errno to ENOMEM already.
Baptiste Daroussin [Fri, 28 Dec 2012 20:21:14 +0000 (20:21 +0000)]
malloc() sets errno to ENOMEM already.

Submitted by: Christoph Mallon <christoph.mallon@gmx.de>

11 years agoDo not leave parts of the new group uninitialized in gr_dup().
Baptiste Daroussin [Fri, 28 Dec 2012 20:19:54 +0000 (20:19 +0000)]
Do not leave parts of the new group uninitialized in gr_dup().

Submitted by: Christoph Mallon <christoph.mallon@gmx.de>
Reported by: pjd

11 years agogcc: avoid generating negative values to DW_AT_byte_size.
Pedro F. Giffuni [Fri, 28 Dec 2012 19:02:28 +0000 (19:02 +0000)]
gcc: avoid generating negative values to DW_AT_byte_size.

There is a bug in gcc (GCC/35998) where dwarf reports
sizes of unsigned -1 (0xffffffff).

On NetBSD this generated a faulty CTF entry which then
caused a segfault in ctfmerge. The issue was worked
around in NetBSD's Dtrace but since the issue originated
in gcc, it seems reasonable to fix it here.

Upstream gcc has been slow to react to this issue and
the author that submitted the patch is not interested
in licensing the change to us, so I did an independent
workaround for the issue.

MFC after: 1 week

11 years agoImprove bufring impl:
Attilio Rao [Fri, 28 Dec 2012 17:41:36 +0000 (17:41 +0000)]
Improve bufring impl:
- Remove unused br_prod_bufs member
- Fixup r241037: buf_ring pads br_prod_* and br_cons_* members at 128
  bytes, assuming a fixed cache line size for all the architectures.
  However, the above mentioned revision broke the padding.
  Use explicit padding to the CACHE_LINE_SIZE on the members that
  mark the initial new padded sections. Of course, the padding is not
  important for performance reasons in the DEBUG_BUFRING case, leaving
  br_cons members to share the cache line with br_lock.
- Fixup r244732: by removing incorrectly added membar in
  buf_ring_dequeue_sc() where surrounding locking shoud be enough.
- Drastically reduce the number of membar used (pratically reverting
  r244732) by switching rmb() in buf_ring_dequeue_mc() and wmb() in
  buf_ring_enqueue() to be complete barriers.  This, along with
  br_prod_bufs departure, should fix ordering issues as explained in
  the provided comments.

This patch is not targeted for MFC.

Sponsored by: EMC / Isilon storage division
Reviewed by: glebius

11 years agoFix location of /var/audit/dist and /var/audit/remote.
Hajimu UMEMOTO [Fri, 28 Dec 2012 10:42:01 +0000 (10:42 +0000)]
Fix location of /var/audit/dist and /var/audit/remote.
Note that those who did installworld after r243752 should
remove wrongly created /var/dist and /var/remote.

Reviewed by: pjd

11 years agoIn netpfil/pf:
Gleb Smirnoff [Fri, 28 Dec 2012 09:19:49 +0000 (09:19 +0000)]
In netpfil/pf:
  - Add my copyright to files I've touched a lot this year.
  - Add dash in front of all copyright notices according to style(9).
  - Move $OpenBSD$ down below copyright notices.
  - Remove extra line between cdefs.h and __FBSDID.

11 years agoAdd the AR9280 and later spectral scan register definitions.
Adrian Chadd [Fri, 28 Dec 2012 08:00:31 +0000 (08:00 +0000)]
Add the AR9280 and later spectral scan register definitions.

Obtained from: Linux ath9k, Qualcomm Atheros (datasheet)

11 years agoAdd radar_bin_thresh_sel (bit 24:26), which defines when
Adrian Chadd [Fri, 28 Dec 2012 07:49:45 +0000 (07:49 +0000)]
Add radar_bin_thresh_sel (bit 24:26), which defines when
to consider the radar FFT report bins as "strong".

11 years agoFix build on ARM (and probably other platforms)
Oleksandr Tymoshenko [Fri, 28 Dec 2012 06:52:53 +0000 (06:52 +0000)]
Fix build on ARM (and probably other platforms)

11 years agoAdd makeshift implementation for framebuffer console's cursor
Oleksandr Tymoshenko [Fri, 28 Dec 2012 03:18:05 +0000 (03:18 +0000)]
Add makeshift implementation for framebuffer console's cursor

Basically it's replica of VersatilePB code which is replica of XBox FB
code. All of them are linear framebuffers and should have common bits
moved to reusable framework.

11 years agoUse strlcpy to NULL-terminate error message even if user provided a short
David Xu [Fri, 28 Dec 2012 02:43:33 +0000 (02:43 +0000)]
Use strlcpy to NULL-terminate error message even if user provided a short
buffer.

11 years agoFix event timer on Raspberry Pi
Oleksandr Tymoshenko [Fri, 28 Dec 2012 01:38:43 +0000 (01:38 +0000)]
Fix event timer on Raspberry Pi

- Disable interrupt when updating compare value in order to
   make this operation atomical

- Increase minimum period for event timer. Systimer on BCM2835
    is compare timer, so if minimum period is too small it might
    be less then fraction of time between "read current value" and
    "set compare timer" operations. It means that when timer is armed
    actual counter value is more then compare value and it will take
    whole cycle (~32sec for 1MHz timer) to fire interrupt.

Submitted by: Daisuke Aoyama <aoyama at peach.ne.jp>

11 years agoAdd derived versions for common Simplified Chinese encodings.
Xin LI [Fri, 28 Dec 2012 01:23:12 +0000 (01:23 +0000)]
Add derived versions for common Simplified Chinese encodings.

MFC after: 2 weeks

11 years agoAdd zh_CN.UTF-8 catalog.
Xin LI [Fri, 28 Dec 2012 01:09:30 +0000 (01:09 +0000)]
Add zh_CN.UTF-8 catalog.

Reviewed by: alphachi <alphachi mediaspirit.org>
MFC after: 2 weeks

11 years agoAdd custom renderer for poor man's cursor support for framebuffer console
Oleksandr Tymoshenko [Fri, 28 Dec 2012 00:55:43 +0000 (00:55 +0000)]
Add custom renderer for poor man's cursor support for framebuffer console

11 years agoAdd an ability to set net.link.stf.permit_rfc1918 from the loader.
Andrey V. Elsukov [Thu, 27 Dec 2012 21:26:08 +0000 (21:26 +0000)]
Add an ability to set net.link.stf.permit_rfc1918 from the loader.

MFC after: 2 weeks

11 years agoAdd net.link.stf.permit_rfc1918 sysctl variable. It can be used to allow
Andrey V. Elsukov [Thu, 27 Dec 2012 20:59:22 +0000 (20:59 +0000)]
Add net.link.stf.permit_rfc1918 sysctl variable. It can be used to allow
the use of private IPv4 addresses with stf(4).

MFC after: 2 weeks

11 years agoavoid arithmetic on uintptr_t
Baptiste Daroussin [Thu, 27 Dec 2012 20:47:34 +0000 (20:47 +0000)]
avoid arithmetic on uintptr_t

Submitted by: pjd
Reviewed by: jilles

11 years agocast to uintptr_t to properly calculate offset
Baptiste Daroussin [Thu, 27 Dec 2012 20:31:12 +0000 (20:31 +0000)]
cast to uintptr_t to properly calculate offset

Reported by: mdf
Submitted by: db

11 years agoAdd O_CLOEXEC to flopen
Baptiste Daroussin [Thu, 27 Dec 2012 20:24:44 +0000 (20:24 +0000)]
Add O_CLOEXEC to flopen

Requested by: jilles

11 years agoSimplify vnextgrent and vnextpwent reusing pw_scan and gr_scan from libutil.
Baptiste Daroussin [Thu, 27 Dec 2012 19:43:29 +0000 (19:43 +0000)]
Simplify vnextgrent and vnextpwent reusing pw_scan and gr_scan from libutil.

11 years agogr_dup: simplify duplication of group
Baptiste Daroussin [Thu, 27 Dec 2012 19:33:43 +0000 (19:33 +0000)]
gr_dup: simplify duplication of group

Submitted by: db

11 years ago- Clean up previous gr_add use malloc instead of calloc
Baptiste Daroussin [Thu, 27 Dec 2012 16:51:29 +0000 (16:51 +0000)]
- Clean up previous gr_add use malloc instead of calloc
- Fix tinderbox error

Submitted by: db

11 years agoFix a regression in "pw group show" introduced r242349:
Baptiste Daroussin [Thu, 27 Dec 2012 14:44:13 +0000 (14:44 +0000)]
Fix a regression in "pw group show" introduced r242349:
print a newline after printing each group line.

PR: bin/174731
Submitted by: Jan Beich <jbeich@tormail.org>

11 years agoSimplify the code by using the new gr_add function
Baptiste Daroussin [Thu, 27 Dec 2012 14:35:06 +0000 (14:35 +0000)]
Simplify the code by using the new gr_add function

11 years agoNew gr_add function to provide a clean and safe method to append a new member
Baptiste Daroussin [Thu, 27 Dec 2012 14:30:19 +0000 (14:30 +0000)]
New gr_add function to provide a clean and safe method to append a new member
into an existing group.

Submitted by: db

11 years agoUse flopen(3) instead of open(2) + flock(2)
Baptiste Daroussin [Thu, 27 Dec 2012 14:09:50 +0000 (14:09 +0000)]
Use flopen(3) instead of open(2) + flock(2)

11 years agotest(1): Document == alias for =.
Jilles Tjoelker [Thu, 27 Dec 2012 13:21:37 +0000 (13:21 +0000)]
test(1): Document == alias for =.

Reviewed by: gjb
Requested by: gjb
MFC after: 1 week

11 years agoRemove an unused var.
Attilio Rao [Thu, 27 Dec 2012 12:53:46 +0000 (12:53 +0000)]
Remove an unused var.

Sponsored by: EMC / Isilon storage division
MFC after: 3 days

11 years agobr_prod_tail and br_cons_tail members are used as barrier to
Attilio Rao [Thu, 27 Dec 2012 12:36:58 +0000 (12:36 +0000)]
br_prod_tail and br_cons_tail members are used as barrier to
signal bug_ring ownership. However, instructions can be reordered
around members write leading to stale values for ie. br_prod_bufs.

Use correct memory barriers to ensure proper ordering of the
ownership tokens updates.

Sponsored by: EMC / Isilon storage division
MFC after: 2 weeks

11 years agouse nanosecond resolution, make sure gettimeofday
Luigi Rizzo [Thu, 27 Dec 2012 09:15:21 +0000 (09:15 +0000)]
use nanosecond resolution, make sure gettimeofday
is called at most every 100 packets.

11 years agoSome cleanups.
Michael Tuexen [Thu, 27 Dec 2012 08:10:58 +0000 (08:10 +0000)]
Some cleanups.

MFC after: 3 days

11 years agoMinor cleanups of debug messages.
Michael Tuexen [Thu, 27 Dec 2012 08:06:58 +0000 (08:06 +0000)]
Minor cleanups of debug messages.

MFC after: 3 days

11 years agoFix a copy and paste error.
Michael Tuexen [Thu, 27 Dec 2012 08:02:58 +0000 (08:02 +0000)]
Fix a copy and paste error.

MFC after: 3 days

11 years agoComments and fit to 80-columns.
Devin Teske [Thu, 27 Dec 2012 07:52:50 +0000 (07:52 +0000)]
Comments and fit to 80-columns.

11 years agoSimplify copying of group members by using memcpy
Baptiste Daroussin [Wed, 26 Dec 2012 23:16:24 +0000 (23:16 +0000)]
Simplify copying of group members by using memcpy

Submitted by: Christoph Mallon <christoph.mallon@gmx.de>

11 years agoFix off-by-one error in memory allocation: j entries, one new and a null
Baptiste Daroussin [Wed, 26 Dec 2012 23:14:33 +0000 (23:14 +0000)]
Fix off-by-one error in memory allocation: j entries, one new and a null
terminator is j + 2.

Submitted by: Christoph Mallon <christoph.mallon@gmx.de>

11 years agoAdd the NO_SYNC_CACHE quirk for all Apple USB MSC devices, as they
Mark Johnston [Wed, 26 Dec 2012 22:21:27 +0000 (22:21 +0000)]
Add the NO_SYNC_CACHE quirk for all Apple USB MSC devices, as they
typically do not handle the SYNCHRONIZE_CACHE command - they either
return an error or the firmware enters a reset loop.

Reviewed by: hselasky
Approved by: rstone (co-mentor)
MFC after: 2 weeks

11 years agoReset provider-specific fields when resending I/O request in low memory
Pawel Jakub Dawidek [Wed, 26 Dec 2012 20:07:47 +0000 (20:07 +0000)]
Reset provider-specific fields when resending I/O request in low memory
conditions. This fixes assertion which checks those fields when kernel is
compiled with DIAGNOSTIC.

Reported by: kib, pho
MFC after: 1 week

11 years agoUse DEVMETHOD_END, as suggested by hselasky@.
Raphael Kubo da Costa [Wed, 26 Dec 2012 19:14:21 +0000 (19:14 +0000)]
Use DEVMETHOD_END, as suggested by hselasky@.

Approved by: glebius

11 years agoUse the correct USB interface macros instead of USB_IF_CSI.
Raphael Kubo da Costa [Wed, 26 Dec 2012 19:12:31 +0000 (19:12 +0000)]
Use the correct USB interface macros instead of USB_IF_CSI.

As pointed out by hselasky@, USB_IF_CSI is the wrong macro here since we want
to declare the device's interface class, subclass and protocol, not class,
subclass and driver info.

Follow-up to r244704.

PR: kern/174707
Approved by: glebius
MFC after: 1 week

11 years agoIn case of the deletion of a user those whole database has to be regenerated,
Baptiste Daroussin [Wed, 26 Dec 2012 18:28:17 +0000 (18:28 +0000)]
In case of the deletion of a user those whole database has to be regenerated,
otherwise the user planned to be deleted remain in the pwd.db while removed from
the plain text password file.

11 years agoFix creating a user and adding it to a group
Baptiste Daroussin [Wed, 26 Dec 2012 18:14:45 +0000 (18:14 +0000)]
Fix creating a user and adding it to a group

Reported by: "Sam Fourman Jr." <sfourman@gmail.com>, dim

11 years agoFixup r244240: mp_ncpus will be 1 also in the !SMP and smp_disabled=1
Attilio Rao [Wed, 26 Dec 2012 15:20:32 +0000 (15:20 +0000)]
Fixup r244240: mp_ncpus will be 1 also in the !SMP and smp_disabled=1
case. There is no point in optimizing further the code and use a TRUE
litteral for a path that does heavyweight stuff anyway (like lock acq),
at the price of obfuscated code.

Use the appropriate check where necessary and remove a macro.

Sponsored by: EMC / Isilon storage division
MFC after: 3 days