]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
12 years agoIn preparation for multi-FIB IPv6 support, factor the code for joining
Bjoern A. Zeeb [Fri, 3 Feb 2012 08:50:19 +0000 (08:50 +0000)]
In preparation for multi-FIB IPv6 support, factor the code for joining
and leaving multicast groups out from in6_update_ifa() and in6_purgeaddr().

Sponsored by: Cisco Systems, Inc.

12 years agoWhitespace removal.
Rui Paulo [Fri, 3 Feb 2012 02:06:14 +0000 (02:06 +0000)]
Whitespace removal.

12 years agoAdd some missing radar config parameters to athradar.
Adrian Chadd [Thu, 2 Feb 2012 21:04:24 +0000 (21:04 +0000)]
Add some missing radar config parameters to athradar.

12 years agoInsert ordered command every 1/4 of the current command timeout, not 1/4
Alexander Motin [Thu, 2 Feb 2012 19:02:15 +0000 (19:02 +0000)]
Insert ordered command every 1/4 of the current command timeout, not 1/4
of the default one.

Without this change setting kern.cam.ada.default_timeout to 1 instead of 30
allowed me to trigger several false positive command timeouts under heavy
ZFS load on a SiI3132 siis(4) controller with 5 HDDs on a port multiplier.

MFC after: 1 week

12 years agoFix the netback driver build for i386.
Kenneth D. Merry [Thu, 2 Feb 2012 17:54:35 +0000 (17:54 +0000)]
Fix the netback driver build for i386.

netback.c: Add missing VM includes.

xen/xenvar.h,
xen/xenpmap.h: Move some XENHVM macros from <machine/xen/xenpmap.h> to
<machine/xen/xenvar.h> on i386 to match the amd64 headers.

conf/files: Add netback to the build.

Submitted by: jhb
MFC after: 3 days

12 years agoRemove leftovers from previous commits.
Joel Dahl [Thu, 2 Feb 2012 17:25:39 +0000 (17:25 +0000)]
Remove leftovers from previous commits.

Approved by: jhb

12 years agoAdd copyright information on last commits to comply with CDDL.
Martin Matuska [Thu, 2 Feb 2012 16:33:58 +0000 (16:33 +0000)]
Add copyright information on last commits to comply with CDDL.

Discussed with: pluknet@
MFC after: 3 days

12 years agoFix out of bounds write causing random panics,
Martin Matuska [Thu, 2 Feb 2012 16:18:40 +0000 (16:18 +0000)]
Fix out of bounds write causing random panics,
uncovered by the change in r230256

Reviewed by: pluknet@
MFC after: 3 days

12 years agoMake CAM ATA honor old hw.ata.ata_dma and hw.ata.atapi_dma tunables.
Alexander Motin [Thu, 2 Feb 2012 14:17:58 +0000 (14:17 +0000)]
Make CAM ATA honor old hw.ata.ata_dma and hw.ata.atapi_dma tunables.

Even having more specific hint.ata.X.mode controls, global ones are
still could be useful from some points, including compatibility.

PR: kern/164651
MFC after: 1 week

12 years agoUnbreak the i386 full PV (XEN kernel) build by reverting rev 230878 and
Kenneth D. Merry [Thu, 2 Feb 2012 07:38:40 +0000 (07:38 +0000)]
Unbreak the i386 full PV (XEN kernel) build by reverting rev 230878 and
taking the netback driver out of the build temporarily.

12 years agoCleanup some sound headers that don't need to be in the modules makefiles.
Pedro F. Giffuni [Thu, 2 Feb 2012 00:10:20 +0000 (00:10 +0000)]
Cleanup some sound headers that don't need to be in the modules makefiles.

Approved by: jhb (mentor)
MFC after: 1 week

12 years agoUse new OSS-based BSD-licensed header for cs sound driver.
Pedro F. Giffuni [Wed, 1 Feb 2012 21:38:01 +0000 (21:38 +0000)]
Use new OSS-based BSD-licensed header for cs sound driver.

The cs driver requires a table with firmware values. An
alternative firmware is available in a similar Open Sound
System driver. This is actually a partial revert of
Revision 77504.

Special thanks to joel@ for patiently testing several
replacement attempts.

The csa driver and the complete sound system are now free
of the GPL.

Tested by: joel
Approved by: jhb (mentor)
MFC after: 3 weeks

12 years agoAdd the GSO prefix descriptor define.
Kenneth D. Merry [Wed, 1 Feb 2012 20:19:33 +0000 (20:19 +0000)]
Add the GSO prefix descriptor define.

MFC after: 3 days

12 years agoAdd the netback driver to the build.
Kenneth D. Merry [Wed, 1 Feb 2012 20:18:57 +0000 (20:18 +0000)]
Add the netback driver to the build.

MFC after: 3 days

12 years agoFix NULL dereference panic on attempt to turn off (on system shutdown)
Alexander Motin [Wed, 1 Feb 2012 20:12:44 +0000 (20:12 +0000)]
Fix NULL dereference panic on attempt to turn off (on system shutdown)
disconnected swap device.

This is quick and imperfect solution, as swap device will still be opened
and GEOM will not be able to destroy it. Proper solution would be to
automatically turn off and close disconnected swap device, but with existing
code it will cause panic if there is at least one page on device, even if
it is unimportant page of the user-level process. It needs some work.

Reviewed by: kib@
MFC after: 1 week

12 years agoAdd myself to the relevant docs.
Davide Italiano [Wed, 1 Feb 2012 19:46:12 +0000 (19:46 +0000)]
Add myself to the relevant docs.

Approved by: alc, gnn (mentors)

12 years agoTry to avoid ambiguity when sysctl returns ENOMEM additionally
Mikolaj Golub [Wed, 1 Feb 2012 18:03:21 +0000 (18:03 +0000)]
Try to avoid ambiguity when sysctl returns ENOMEM additionally
checking the returned oldlen: when ENOMEM is due to the supplied
buffer being too short the return oldlen is equal to buffer size.

Without this additional check sockstat gets stuck in loop leaking the
memory if the returned ENOMEM was due the exceeded memorylocked
limit. This is easily can be observed running `limits -l 1k sockstat'.

Submitted by: Andrey Zonov <andrey zonov org>
MFC after: 1 week

12 years agoTry to avoid ambiguity when sysctl returns ENOMEM additionally
Mikolaj Golub [Wed, 1 Feb 2012 18:02:13 +0000 (18:02 +0000)]
Try to avoid ambiguity when sysctl returns ENOMEM additionally
checking the returned oldlen: when ENOMEM is due to the supplied
buffer being too short the return oldlen is equal to buffer size.

Without this additional check kvm_getprocs() gets stuck in loop if the
returned ENOMEM was due the exceeded memorylocked limit. This is
easily can be observed running `limits -l 1k top'.

Submitted by: Andrey Zonov <andrey zonov org>
MFC after: 1 week

12 years agoChange the notes about the pidfile to include Doug's preference
Guy Helmer [Wed, 1 Feb 2012 16:40:59 +0000 (16:40 +0000)]
Change the notes about the pidfile to include Doug's preference
for pre-creating the pidfile with appropriate owner and permissions.

Requested by dougb

12 years agoAdd kqueue support to /dev/klog.
Konstantin Belousov [Wed, 1 Feb 2012 14:34:52 +0000 (14:34 +0000)]
Add kqueue support to /dev/klog.

Submitted by: Mateusz Guzik <mjguzik gmail com>
PR:   kern/156423
MFC after: 1 weeks

12 years agoAdjust mvs(4) to handle interrupt cause reg depending on the actual number of
Rafal Jaworowski [Wed, 1 Feb 2012 13:39:52 +0000 (13:39 +0000)]
Adjust mvs(4) to handle interrupt cause reg depending on the actual number of
channels available

- current code treats bits 4:7 in 'SATAHC interrupt mask' and 'SATAHC
  interrupt cause' as flags for SATA channels 2 and 3

- for embedded SATA controllers (SoC) these bits have been marked as reserved
  in datasheets so far, but for some new and upcoming chips they are used for
  purposes other than SATA

Submitted by: Lukasz Plachno
Reviewed by: mav
Obtained from: Semihalf
MFC after: 2 weeks

12 years agoMake the sys/ucontext.h self-contained by changing the return type
Konstantin Belousov [Wed, 1 Feb 2012 13:33:53 +0000 (13:33 +0000)]
Make the sys/ucontext.h self-contained by changing the return type
of __getcontextx_size(3) from size_t to int.

PR: ports/164654
MFC after: 1 month

12 years agoo Provide functions carp_ifa_addroute()/carp_ifa_delroute()
Gleb Smirnoff [Wed, 1 Feb 2012 11:07:41 +0000 (11:07 +0000)]
o Provide functions carp_ifa_addroute()/carp_ifa_delroute()
  to cleanup routes from a single ifa.
o Implement carp_addroute()/carp_delroute() via above functions.
o Call carp_ifa_delroute() in the carp_detach() to avoid
  junk routes left in routing table, in case if user
  removes an address in a MASTER state. [1]

Reported by: az [1]

12 years ago- remove non-existent input directories from the configs
Alexander Leidinger [Wed, 1 Feb 2012 10:59:19 +0000 (10:59 +0000)]
- remove non-existent input directories from the configs
- add the target-arch to the project name where applicable (arch dependend code)
- fix comment for __FreeBSD__ [1]

Noticed by: bz [1]

12 years agoThe scheme code may not know about some inconsistency in the metadata.
Andrey V. Elsukov [Wed, 1 Feb 2012 09:28:16 +0000 (09:28 +0000)]
The scheme code may not know about some inconsistency in the metadata.
So, add an integrity check after recovery attempt.

MFC after: 1 week

12 years agoEnable the pccard/cardbus modules for powerpc.
Justin Hibbits [Wed, 1 Feb 2012 03:42:14 +0000 (03:42 +0000)]
Enable the pccard/cardbus modules for powerpc.

Approved by: nwhitehorn (mentor)

12 years agoIf multiple threads call kevent() to get AIO events on same kqueue fd,
David Xu [Wed, 1 Feb 2012 02:53:06 +0000 (02:53 +0000)]
If multiple threads call kevent() to get AIO events on same kqueue fd,
it is possible that a single AIO event will be reported to multiple
threads, it is not threading friendly, and the existing API can not
control this behavior.
Allocate a kevent flags field sigev_notify_kevent_flags for AIO event
notification in sigevent, and allow user to pass EV_CLEAR, EV_DISPATCH
or EV_ONESHOT to AIO kernel code, user can control whether the event
should be cleared once it is retrieved by a thread. This change should
be comptaible with existing application, because the field should have
already been zero-filled, and no additional action will be taken by
kernel.

PR: kern/156567

12 years agoSupport AR9281/AR5B91 - a 1x2 stream device based on the AR9280.
Adrian Chadd [Tue, 31 Jan 2012 22:31:16 +0000 (22:31 +0000)]
Support AR9281/AR5B91 - a 1x2 stream device based on the AR9280.

* Override the TX/RX stream count if the EEPROM reports a single RX or
  TX stream, rather than assuming the device will always be a 2x2 strea
  device.

* For AR9280 devices, don't hard-code 2x2 stream.  Instead, allow the
  ar5416FillCapabilityInfo() routine to correctly determine things.

The latter should be done for all 11n chips now that
ar5416FillCapabilityInfo() will set the TX/RX stream count based on the
active TX/RX chainmask in the EEPROM.

Thanks to Maciej Milewski for donating some AR9281 NICs to me for
testing.

12 years agoCorrectly fetch the TX/RX stream count from the HAL.
Adrian Chadd [Tue, 31 Jan 2012 22:27:35 +0000 (22:27 +0000)]
Correctly fetch the TX/RX stream count from the HAL.

Pointy hat to: me

12 years agoMake sound(4) more flexible in setting soft buffer and block sizes when
Alexander Motin [Tue, 31 Jan 2012 21:46:28 +0000 (21:46 +0000)]
Make sound(4) more flexible in setting soft buffer and block sizes when
hardware imposes strict limitations on hard buffer and block sizes.

Previous code set soft buffer to be no smaller then hard buffer. On some
cards with fixed 64K physical buffer that caused up to 800ms play latency.
New code allows to set soft buffer size down to just two blocks of the hard
buffer and to not write more then that size ahead to the hardware buffer.
As result of that change I was able to reduce full practically measured
record-playback loop delay in those conditions down to only about 115ms
with theoretical playback latency of only about 50ms.

New code works fine for both vchans and direct cases. In both cases sound(4)
tries to follow hw.snd.latency_profile and hw.snd.latency values and
application-requested buffer and block sizes as much as limitation of two
hardware blocks allows.

Reviewed by: silence on multimedia@

12 years agoAdd isci(4) driver for amd64 and i386 targets.
Jim Harris [Tue, 31 Jan 2012 19:38:18 +0000 (19:38 +0000)]
Add isci(4) driver for amd64 and i386 targets.

The isci driver is for the integrated SAS controller in the Intel C600
(Patsburg) chipset.  Source files in sys/dev/isci directory are
FreeBSD-specific, and sys/dev/isci/scil subdirectory contains
an OS-agnostic library (SCIL) published by Intel to control the SAS
controller.  This library is used primarily as-is in this driver, with
some post-processing to better integrate into the kernel build
environment.

isci.4 and a README in the sys/dev/isci directory contain a few
additional details.

This driver is only built for amd64 and i386 targets.

Sponsored by: Intel
Reviewed by: scottl
Approved by: scottl

12 years ago- Restore XCR0 before restoring extended FPU states.
Jung-uk Kim [Tue, 31 Jan 2012 17:51:30 +0000 (17:51 +0000)]
- Restore XCR0 before restoring extended FPU states.
- Update my copyright dates.

Reviewed by: kib

12 years agoFix a spelling mistake in the surprise link down error constant.
John Baldwin [Tue, 31 Jan 2012 15:48:40 +0000 (15:48 +0000)]
Fix a spelling mistake in the surprise link down error constant.

Submitted by: glebius

12 years agoAdd -e to set arbitrary kernel environment variables.
Ed Maste [Tue, 31 Jan 2012 15:32:05 +0000 (15:32 +0000)]
Add -e to set arbitrary kernel environment variables.

Nextboot(8) can now set any combination of kernel name (-k), kernel
options (-o), and environment strings (-e).  As a result of this change
-k also becomes optional.

Reviewed by: freebsd-current (Ian Lepore, pluknet@, jhb@)

12 years agoIsolate v_caddr_t in the ie driver.
Sergey Kandaurov [Tue, 31 Jan 2012 13:00:40 +0000 (13:00 +0000)]
Isolate v_caddr_t in the ie driver.

Submitted by: Bruce Evans on net@

12 years agoJust in case, clear stream interrupts before enabling them.
Alexander Motin [Tue, 31 Jan 2012 12:57:21 +0000 (12:57 +0000)]
Just in case, clear stream interrupts before enabling them.

12 years agoFix a whitespace nit
Kevin Lo [Tue, 31 Jan 2012 05:49:49 +0000 (05:49 +0000)]
Fix a whitespace nit

Reviewed by: bz

12 years agoWhen a "mount -u" switches an NFS mount point from TCP to UDP,
Rick Macklem [Tue, 31 Jan 2012 03:58:26 +0000 (03:58 +0000)]
When a "mount -u" switches an NFS mount point from TCP to UDP,
any thread doing an I/O RPC with a transfer size greater than
NFS_UDPMAXDATA will be hung indefinitely, retrying the RPC.
After a discussion on freebsd-fs@, I decided to add a warning
message for this case, as suggested by Jeremy Chadwick.

Suggested by: freebsd at jdc.parodius.com (Jeremy Chadwick)
MFC after: 2 weeks

12 years agojwd@ reported a problem via email to freebsd-fs@ on Aug 25, 2011
Rick Macklem [Tue, 31 Jan 2012 02:11:05 +0000 (02:11 +0000)]
jwd@ reported a problem via email to freebsd-fs@ on Aug 25, 2011
under the subject "F_RDLCK lock to FreeBSD NFS fails to R/O target file".
This occurred because the server side NLM always checked for VWRITE
access, irrespective of the type of lock request. This patch
replaces VOP_ACCESS(..VWRITE..) with one appropriate to
the lock operation. It allows unlock and lock cancellation
to be done without a check of VOP_ACCESS(), so that files
can't be left locked indefinitely after the file permissions
have been changed.

Discussed with: zack
Submitted by: jwd (earlier version)
Reviewed by: dfr
MFC after: 2 weeks

12 years agoAllow contents of multiple directories to be merged to the current image.
Jung-uk Kim [Tue, 31 Jan 2012 00:32:37 +0000 (00:32 +0000)]
Allow contents of multiple directories to be merged to the current image.
Note this patch was submitted to NetBSD and they already adopted it.

http://mail-index.netbsd.org/source-changes/2012/01/28/msg031078.html

MFC after: 1 week

12 years agoAdd a DFS debugging mode which is useful when doing automated DFS
Adrian Chadd [Tue, 31 Jan 2012 00:03:49 +0000 (00:03 +0000)]
Add a DFS debugging mode which is useful when doing automated DFS
compliance testing.

In order to allow for radar pattern matching to occur, the DFS CAC/NOL
handling needs to be made configurable.  This commit introduces a new
sysctl, "net.wlan.dfs_debug", which controls which DFS debug mode
net80211 is in.

* 0 = default, CSA/NOL handling as per normal.
* 1 = announce a CSA, but don't add the channel to the non-occupy list
  (NOL.)
* 2 = disable both CSA and NOL - only print that a radar event occured.

This code is not compiled/enabled by default as it breaks regulatory
handling.  A user must enable IEEE80211_DFS_DEBUG in their kernel
configuration file for this option to become available.

Obtained from: Atheros

12 years agoRadar API related fixes.
Adrian Chadd [Mon, 30 Jan 2012 23:07:27 +0000 (23:07 +0000)]
Radar API related fixes.

* For legacy NICs, the combined RSSI should be used.
  For earlier AR5416 NICs, use control chain 0 RSSI rather than combined
  RSSI.
  For AR5416 > version 2.1, use the combined RSSI again.

* Add in a missing AR5212 HAL method (get11nextbusy) which may be called
  by radar code.

This serves no functional change for what's currently in FreeBSD.

12 years agoWrap the bool typedef
Jack F Vogel [Mon, 30 Jan 2012 23:03:21 +0000 (23:03 +0000)]
Wrap the bool typedef

12 years agoImprove media status reporting, when the driver knows...
Warner Losh [Mon, 30 Jan 2012 22:47:02 +0000 (22:47 +0000)]
Improve media status reporting, when the driver knows...

12 years agoManpages for the buf_ring and drbr interfaces.
Christian Brueffer [Mon, 30 Jan 2012 21:02:25 +0000 (21:02 +0000)]
Manpages for the buf_ring and drbr interfaces.

Submitted by: kmacy (aeons ago)
MFC after: 1 week

12 years agoAllow specification of build shell for the buildenv target.
Warner Losh [Mon, 30 Jan 2012 20:19:28 +0000 (20:19 +0000)]
Allow specification of build shell for the buildenv target.

Submitted by: ian lepore

12 years agoA debugger which requested PT_FOLLOW_FORK should get the notification
Konstantin Belousov [Mon, 30 Jan 2012 20:00:29 +0000 (20:00 +0000)]
A debugger which requested PT_FOLLOW_FORK should get the notification
about new child not only when doing PT_TO_SCX, but also for PT_CONTINUE.
If TDB_FORK flag is set, always issue a stop, the same as is done for
TDB_EXEC.

Reported by: Dmitry Mikulin <dmitrym juniper net>
MFC after: 1 week

12 years agoAdd support for GNU RELRO.
Konstantin Belousov [Mon, 30 Jan 2012 19:52:17 +0000 (19:52 +0000)]
Add support for GNU RELRO.

Submitted by: John Marino <draco marino st>
MFC after: 2 weeks

12 years agoAdd definition for PT_GNU_RELRO.
Konstantin Belousov [Mon, 30 Jan 2012 19:50:13 +0000 (19:50 +0000)]
Add definition for PT_GNU_RELRO.

MFC after: 3 days

12 years agoRefine the implementation of POSIX_FADV_NOREUSE for the read(2) case such
John Baldwin [Mon, 30 Jan 2012 19:35:15 +0000 (19:35 +0000)]
Refine the implementation of POSIX_FADV_NOREUSE for the read(2) case such
that instead of using direct I/O it allows read-ahead similar to
POSIX_FADV_NORMAL, but invokes VOP_ADVISE(POSIX_FADV_DONTNEED) after the
read(2) has completed to purge just-read data.  The write(2) path continues
to use direct I/O for POSIX_FADV_NOREUSE for now.  Note that NOREUSE works
optimally if an application reads and writes full fs blocks.

12 years agoFix build for the case of powerpc64 kernel without COMPAT_FREEBSD32.
Konstantin Belousov [Mon, 30 Jan 2012 19:31:17 +0000 (19:31 +0000)]
Fix build for the case of powerpc64 kernel without COMPAT_FREEBSD32.

MFC after: 2 months

12 years agoWhen detaching an AIO or LIO requests grab the lock and tell knlist_remove
Doug Ambrisko [Mon, 30 Jan 2012 19:19:22 +0000 (19:19 +0000)]
When detaching an AIO or LIO requests grab the lock and tell knlist_remove
that we have the lock now.  This cleans up a locking panic ASSERT when
knlist_empty is called without a lock when INVARIANTS etc. are turned.

Reviewed by: kib jhb
MFC after: 1 week

12 years agoNaturally align a newly added wakeup_fpusave.
Jung-uk Kim [Mon, 30 Jan 2012 18:28:56 +0000 (18:28 +0000)]
Naturally align a newly added wakeup_fpusave.

12 years agoNew hardware support: Intel X540 adapter support added.
Jack F Vogel [Mon, 30 Jan 2012 16:42:02 +0000 (16:42 +0000)]
New hardware support:  Intel X540 adapter support added.
Some shared code reorganization along with the new adapter.
Sync changes to OACTIVE in igb into this driver.
Misc small fixes.

12 years agoAdd a constant for the PCI-e surprise link down uncorrectable error.
John Baldwin [Mon, 30 Jan 2012 15:09:03 +0000 (15:09 +0000)]
Add a constant for the PCI-e surprise link down uncorrectable error.

12 years agoFix typo in comments.
Alexander Motin [Mon, 30 Jan 2012 09:59:33 +0000 (09:59 +0000)]
Fix typo in comments.

12 years agoFinally, try to enable the nxstacks on amd64 and powerpc64 for both 64bit
Konstantin Belousov [Mon, 30 Jan 2012 07:56:00 +0000 (07:56 +0000)]
Finally, try to enable the nxstacks on amd64 and powerpc64 for both 64bit
and 32bit ABIs. Also try to enable nxstacks for PAE/i386 when supported,
and some variants of powerpc32.

MFC after: 2 months (if ever)

12 years agoMove xrstor/xsave/xsetbv into fpu.c and reorder them.
Konstantin Belousov [Mon, 30 Jan 2012 07:53:33 +0000 (07:53 +0000)]
Move xrstor/xsave/xsetbv into fpu.c and reorder them.

Requested by: bde
MFC after: 1 month

12 years agoSynchronize the struct sigcontext definitions on x86 with mcontext_t.
Konstantin Belousov [Mon, 30 Jan 2012 07:51:52 +0000 (07:51 +0000)]
Synchronize the struct sigcontext definitions on x86 with mcontext_t.

Pointed out by: bde
MFC after: 1 month

12 years agoUupdate code to vendor rev. 4183 (release/2.8)
Martin Matuska [Sun, 29 Jan 2012 22:20:28 +0000 (22:20 +0000)]
Uupdate code to vendor rev. 4183 (release/2.8)

Fixes vendor issue 224:
"Mishandling CD9660 images with RockRidge extensions from FreeBSD makefs"

References:
http://code.google.com/p/libarchive/issues/detail?id=224

MFC after: 1 week

12 years agoUpdate to vendor revision 4183
Martin Matuska [Sun, 29 Jan 2012 22:12:48 +0000 (22:12 +0000)]
Update to vendor revision 4183

Obtained from: http://libarchive.googlecode.com/svn/release/2.8

12 years agoAlways return 0 if the sysctl failed.
Mikolaj Golub [Sun, 29 Jan 2012 20:39:42 +0000 (20:39 +0000)]
Always return 0 if the sysctl failed.

This fixes the bug: when procstat -xa was run and the sysctl for a
process returned ESRCH or EPERM, for this process procstat output the
result collected for the previous successful process.

12 years agoMissing code for receiving MESH PREP that should be part of r230409.
Adrian Chadd [Sun, 29 Jan 2012 19:35:40 +0000 (19:35 +0000)]
Missing code for receiving MESH PREP that should be part of r230409.

Submitted by: <monthadar@gmail.com>

12 years agoGS105v3 exhibit the same behavior
Eitan Adler [Sun, 29 Jan 2012 14:52:42 +0000 (14:52 +0000)]
GS105v3 exhibit the same behavior

PR: docs/135999
Submitted by: Boris Kochergin <spawky@acm.poly.edu>
No objection from: jfv
Approved by: cperciva
MFC after: 3 days

12 years agoFix build when RELEASE_CRUNCH is defined.
Yoshihiro Takahashi [Sun, 29 Jan 2012 09:12:34 +0000 (09:12 +0000)]
Fix build when RELEASE_CRUNCH is defined.

12 years agoPerform IPv6 DAD only in ifn_start.
Hiroki Sato [Sun, 29 Jan 2012 08:33:34 +0000 (08:33 +0000)]
Perform IPv6 DAD only in ifn_start.

12 years agoalways exclude data bufs regardless of debug settings
Kip Macy [Sun, 29 Jan 2012 00:19:19 +0000 (00:19 +0000)]
always exclude data bufs regardless of debug settings

12 years agoupdate .Dd
Kip Macy [Sat, 28 Jan 2012 23:30:39 +0000 (23:30 +0000)]
update .Dd

12 years agoAs it turns out r227960 may still be insufficient with PREEMPTION
Marius Strobl [Sat, 28 Jan 2012 22:42:33 +0000 (22:42 +0000)]
As it turns out r227960 may still be insufficient with PREEMPTION
so try harder to get the CDMA sync interrupt delivered and also in
a more efficient way:
- wrap the whole process of sending and receiving the CDMA sync
  interrupt in a critical section so we don't get preempted,
- send the CDMA sync interrupt to the CPU that is actually waiting
  for it to happen so we don't take a detour via another CPU,
- instead of waiting for up to 15 seconds for the interrupt to
  trigger try the whole process for up to 15 times using a one
  second timeout (the code was also changed to just ignore belated
  interrupts of a previous tries should they appear).

According to testing done by Peter Jeremy with the debugging also
added as part of this commit the first two changes apparently are
sufficient to now properly get the CDMA sync interrupts delivered
at the first try though.

12 years agoOops, commit a missing implementation change.
Adrian Chadd [Sat, 28 Jan 2012 22:24:59 +0000 (22:24 +0000)]
Oops, commit a missing implementation change.

Whilst I'm here, add a comment about what would happen in this function
if hypothetically you had a radar pattern matching detector written.

12 years agoFully disable interrupts while we fiddle with the FP context in the
Marius Strobl [Sat, 28 Jan 2012 22:22:05 +0000 (22:22 +0000)]
Fully disable interrupts while we fiddle with the FP context in the
VIS-based block copy/zero implementations. While with 4BSD it's
sufficient to just disable the tick interrupts, with ULE+PREEMPTION
it's otherwise also possible that these are preempted via IPIs.

12 years agoChange the prototype so the radar enable can fail.
Adrian Chadd [Sat, 28 Jan 2012 21:44:42 +0000 (21:44 +0000)]
Change the prototype so the radar enable can fail.

12 years agoTwo changes from my DFS work:
Adrian Chadd [Sat, 28 Jan 2012 21:37:33 +0000 (21:37 +0000)]
Two changes from my DFS work:

* Grab the net80211com lock when calling ieee80211_dfs_notify_radar().
* Use the tsf extend function to turn the 64 bit base TSF into a per-
  frame 64 bit TSF.  This will improve radiotap logging (which will
  now have a (more) correct per-frame TSF, rather then the single TSF64
  value read at the beginning of ath_rx_proc().

12 years agomsdos was renamed to msdosfs in 2001 by r77577.
Sean Farley [Sat, 28 Jan 2012 21:21:05 +0000 (21:21 +0000)]
msdos was renamed to msdosfs in 2001 by r77577.

MFC after: 5 days

12 years agoSince April 2, 2006, Indiana has observed DST.
Sean Farley [Sat, 28 Jan 2012 21:06:45 +0000 (21:06 +0000)]
Since April 2, 2006, Indiana has observed DST.

MFC after: 5 days

12 years agoMake tip exit if the device disappears, for instance when unplugging
Poul-Henning Kamp [Sat, 28 Jan 2012 20:45:47 +0000 (20:45 +0000)]
Make tip exit if the device disappears, for instance when unplugging
or resetting USB serial devices.

Somebody[tm] should rewrite tip(1) to use two thread instead of two
processes or maybe even use that new-fangled "select(2)" or positively
futuristic "poll(2)" system call.

12 years agoFix decoding of escape sequences in format strings:
Tijl Coosemans [Sat, 28 Jan 2012 18:49:04 +0000 (18:49 +0000)]
Fix decoding of escape sequences in format strings:
- Zero-terminate the resulting string by letting the for-loop copy the
  terminating zero.
- Exit the for-loop after handling a backslash at the end of the format
  string to fix a buffer overrun.
- Remove some unnecessary comments and blank lines. [1]

Requested by: bde [1]
PR: bin/144722
Approved by: kib (mentor)

12 years agoMove descriptions of file caching commands out of the file locking section.
Tijl Coosemans [Sat, 28 Jan 2012 18:35:10 +0000 (18:35 +0000)]
Move descriptions of file caching commands out of the file locking section.

Approved by: kib (mentor)

12 years agoadd tunable for developers working on areas outside of ZFS
Kip Macy [Sat, 28 Jan 2012 17:41:42 +0000 (17:41 +0000)]
add tunable for developers working on areas outside of ZFS
to further reduce core size by excluding ARC metadata buffers
from core dumps

12 years agoalphabetically sort optional flags and add documentation for VM_ALLOC_NODUMP
Kip Macy [Sat, 28 Jan 2012 17:27:58 +0000 (17:27 +0000)]
alphabetically sort optional flags and add documentation for VM_ALLOC_NODUMP

12 years agodocument M_NODUMP flag
Kip Macy [Sat, 28 Jan 2012 17:21:12 +0000 (17:21 +0000)]
document M_NODUMP flag

12 years agoo Preserve argv[0] to use it later in usage().
Maxim Konovalov [Sat, 28 Jan 2012 15:29:43 +0000 (15:29 +0000)]
o Preserve argv[0] to use it later in usage().

PR: bin/164570
Submitted by: Klaus Aehlig
MFC after: 1 week

12 years agoAvoid to check the same cache line/variable from all the locking
Attilio Rao [Sat, 28 Jan 2012 14:00:21 +0000 (14:00 +0000)]
Avoid to check the same cache line/variable from all the locking
primitives by breaking stop_scheduler into a per-thread variable.
Also, store the new td_stopsched very close to td_*locks members as
they will be accessed mostly in the same codepaths as td_stopsched and
this results in avoiding a further cache-line pollution, possibly.

STOP_SCHEDULER() was pondered to use a new 'thread' argument, in order to
take advantage of already cached curthread, but in the end there should
not really be a performance benefit, while introducing a KPI breakage.

In collabouration with: flo
Reviewed by: avg
MFC after: 3 months (or never)
X-MFC: r228424

12 years agoRemove trailing whitespace.
Jaakko Heinonen [Sat, 28 Jan 2012 13:41:34 +0000 (13:41 +0000)]
Remove trailing whitespace.

12 years agoFix HBR enabling condition. cchs is from 0 to 7, not from 1 to 8.
Alexander Motin [Sat, 28 Jan 2012 09:24:57 +0000 (09:24 +0000)]
Fix HBR enabling condition. cchs is from 0 to 7, not from 1 to 8.

12 years agoRemove the notify match from a couple devd apple events, the events don't
Justin Hibbits [Sat, 28 Jan 2012 02:52:22 +0000 (02:52 +0000)]
Remove the notify match from a couple devd apple events, the events don't
include notify tags.

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

12 years agopmc_*_initialize may return NULL if the CPU is not supported, so check
Ed Maste [Sat, 28 Jan 2012 01:38:48 +0000 (01:38 +0000)]
pmc_*_initialize may return NULL if the CPU is not supported, so check
that md is not null before dereferencing it.

PR: kern/156540

12 years agoSet SVN text/plain property for some shell scripts that
Pedro F. Giffuni [Sat, 28 Jan 2012 00:17:17 +0000 (00:17 +0000)]
Set SVN text/plain property for some shell scripts that
happen to have a .exe extension.

While here fix the shebang of a shell script that
was looking for /bin/bash.

Reviewed by: gnn
Approved by: jhb (mentor)
MFC after: 2 weeks

12 years agoCommit file missed in r230633.
Marius Strobl [Fri, 27 Jan 2012 23:25:24 +0000 (23:25 +0000)]
Commit file missed in r230633.

12 years agoNow that we have a working OF_printf() since r230631 and a OF_panic()
Marius Strobl [Fri, 27 Jan 2012 23:21:54 +0000 (23:21 +0000)]
Now that we have a working OF_printf() since r230631 and a OF_panic()
helper since r230632, use these for output and panicing during the
early cycles and move cninit() until after the static per-CPU data
has been set up. This solves a couple of issue regarding the non-
availability of the static per-CPU data:
- panic() not working and only making things worse when called,
- having to supply a special DELAY() implementation to the low-level
  console drivers,
- curthread accesses of mutex(9) usage in low-level console drivers
  that aren't conditional due to compiler optimizations (basically,
  this is the problem described in r227537 but in this case for
  keyboards attached via uart(4)). [1]

PR: 164123 [1]

12 years ago- Now that we have a working OF_printf() since r230631, use it for
Marius Strobl [Fri, 27 Jan 2012 22:35:53 +0000 (22:35 +0000)]
- Now that we have a working OF_printf() since r230631, use it for
  implementing a simple OF_panic() that may be used during the early
  cycles when panic() isn't available, yet.
- Mark cpu_{exit,shutdown}() as __dead2 as appropriate.

12 years agoImplement OF_printf() using kvprintf() directly, avoiding to use a
Marius Strobl [Fri, 27 Jan 2012 22:29:29 +0000 (22:29 +0000)]
Implement OF_printf() using kvprintf() directly, avoiding to use a
buffer and allowing to handle newlines properly

12 years agoFor machines where the kernel address space is unrestricted increase
Marius Strobl [Fri, 27 Jan 2012 22:25:46 +0000 (22:25 +0000)]
For machines where the kernel address space is unrestricted increase
VM_KMEM_SIZE_SCALE to 2, awaiting more insight from alc@. As it turns
out, the VM apparently has problems with machines that have large holes
in the physical address space, causing the kmem_suballoc() call in
kmeminit() to fail with a VM_KMEM_SIZE_SCALE of 1. Using a value of 2
allows these, namely Blade 1500 with 2GB of RAM, to boot.

PR: 164227

12 years agoadd KM_NODEBUG needed by ARC buffer core dump exclusion change
Kip Macy [Fri, 27 Jan 2012 22:24:03 +0000 (22:24 +0000)]
add KM_NODEBUG needed by ARC buffer core dump exclusion change

12 years agoMark cpu_{halt,reset}() as __dead2 as appropriate.
Marius Strobl [Fri, 27 Jan 2012 22:04:43 +0000 (22:04 +0000)]
Mark cpu_{halt,reset}() as __dead2 as appropriate.

12 years agoUsing ATA_CAM along with ATAPI DMA causes data corruption with ALI_NEW
Marius Strobl [Fri, 27 Jan 2012 21:52:59 +0000 (21:52 +0000)]
Using ATA_CAM along with ATAPI DMA causes data corruption with ALI_NEW
and CMD controllers for reasons unknown so disable it.

PR: 164226

12 years agoSome laptops have weak power controllers that cannot tolerate multiple
Warner Losh [Fri, 27 Jan 2012 21:49:02 +0000 (21:49 +0000)]
Some laptops have weak power controllers that cannot tolerate multiple
cards powering up at once.  Work around the easy case (multiple cards
inserted on boot) with a short sleep and a long comment.  This
improves reliability on those laptops with power hungry cards.

12 years agoexclude kmem_alloc'ed ARC data buffers from kernel minidumps on amd64
Kip Macy [Fri, 27 Jan 2012 20:18:31 +0000 (20:18 +0000)]
exclude kmem_alloc'ed ARC data buffers from kernel minidumps on amd64
excluding other allocations including UMA now entails the addition of
a single flag to kmem_alloc or uma zone create

Reviewed by: alc, avg
MFC after: 2 weeks

12 years agoWhen the buildkernel stage 2.3 (build tools) runs, the PATH is still set
Dimitry Andric [Fri, 27 Jan 2012 18:29:03 +0000 (18:29 +0000)]
When the buildkernel stage 2.3 (build tools) runs, the PATH is still set
to the default from the top-level Makefile.  Therefore, invocations of
lex and yacc (used during building of aicasm) will use the executables
in /usr/bin, not those optionally built during the previous buildworld
or kernel-toolchain.  This makes kernel builds from older FreeBSD
releases more difficult than necessary.

Fix this by setting PATH to ${BPATH}:${PATH} in stage 2.3, so the
bootstrap tools directories are searched before the regular ones.

Silence from: svn-src-{all,head}
MFC after: 1 week

12 years agoOn state changes from RUN to anything else the AGGR sessions are
Bernhard Schmidt [Fri, 27 Jan 2012 17:39:02 +0000 (17:39 +0000)]
On state changes from RUN to anything else the AGGR sessions are
cleared/dropped leading to qid2tap[n] being NULL as there no longer
is a tap. Now, if there have been lots of frames queued the firmware
processes and returns those after the tap is gone.

Tested by: osa
MFC after: 1 week