]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
14 years agoFixed error checking of pthread(3) functions.
ru [Fri, 12 Feb 2010 16:33:03 +0000 (16:33 +0000)]
Fixed error checking of pthread(3) functions.

PR: 143807
Submitted by: pluknet (partly)

14 years agoMake manpage's SYNOPSIS match program's usage().
ru [Fri, 12 Feb 2010 15:07:24 +0000 (15:07 +0000)]
Make manpage's SYNOPSIS match program's usage().

Submitted by: Alexander Best (manpage)

14 years agoRemove the PCI_IOSPACE_SIZE and PCI_IOSPACE_ADDR hack from nexus.c. Implement
neel [Fri, 12 Feb 2010 02:59:49 +0000 (02:59 +0000)]
Remove the PCI_IOSPACE_SIZE and PCI_IOSPACE_ADDR hack from nexus.c. Implement
this in the Sibyte PCI hostbridge driver instead.

The nexus driver sees resource allocation requests for memory and irq
resources only. These are legitimate resources on all MIPS platforms.

Suggested by: imp

14 years ago- Reorganize code in 'plugin' to share log processing.
fabient [Thu, 11 Feb 2010 22:51:44 +0000 (22:51 +0000)]
- Reorganize code in 'plugin' to share log processing.
- Kcachegrind (calltree) support with assembly/source
code mapping and call count estimator (-F).
- Top mode for calltree and callgraph plugin (-T).

MFC after: 1 month

14 years agoInitialize pve_fsid and pve_fileid to VNOVAL.
marcel [Thu, 11 Feb 2010 21:10:56 +0000 (21:10 +0000)]
Initialize pve_fsid and pve_fileid to VNOVAL.

14 years agos/APIS/APIs - not part of the original submission.
trhodes [Thu, 11 Feb 2010 19:20:06 +0000 (19:20 +0000)]
s/APIS/APIs - not part of the original submission.

14 years agoacpi: drop the second bus_generic_attach pass
avg [Thu, 11 Feb 2010 18:24:00 +0000 (18:24 +0000)]
acpi: drop the second bus_generic_attach pass

It is belived that that pass s not needed anymore.
Specifically it is not required now for the reasons that were given
in the removed comment.

Discussed with: jhb
MFC after: 4 weeks

14 years agoOne last pass to get all the unsigned comparisons correct.
mckusick [Thu, 11 Feb 2010 18:14:53 +0000 (18:14 +0000)]
One last pass to get all the unsigned comparisons correct.

14 years agoo Add support for COMPAT_IA32.
marcel [Thu, 11 Feb 2010 18:00:53 +0000 (18:00 +0000)]
o  Add support for COMPAT_IA32.
o  Incorporate review comments:
   -  Properly reference and lock the map
   -  Take into account that the VM map can change inbetween requests
   -  Add the fileid and fsid attributes

Credits: kib@
Reviewed by: kib@

14 years agoi doesn't need to be signed here, make it unsigned.
imp [Thu, 11 Feb 2010 17:30:30 +0000 (17:30 +0000)]
i doesn't need to be signed here, make it unsigned.

14 years agoFix a installation issue with $DESTDIR.
nork [Thu, 11 Feb 2010 15:34:54 +0000 (15:34 +0000)]
Fix a installation issue with $DESTDIR.

Approved by: imp (mentor)

14 years agoCorrect two typoes.
trhodes [Thu, 11 Feb 2010 14:45:00 +0000 (14:45 +0000)]
Correct two typoes.

Submitted by: Matthew Seaman <m.seaman@infracaninophile.co.uk>

14 years agoacpi cpu: probe+attach before all other enumerated children on acpi bus
avg [Thu, 11 Feb 2010 08:50:21 +0000 (08:50 +0000)]
acpi cpu: probe+attach before all other enumerated children on acpi bus

Some current systems dynamically load SSDT(s) when _PDC/_OSC method
of Processor is evaluated.  Other devices in ACPI namespace may access
objects defined in the dynamic SSDT.  Drivers for such devices might
have to have a rather high priority, because of other dependencies.
Good example is acpi_ec driver for EC.
Thus we attach to Processors as early as possible to load the SSDTs
before any other drivers may try to evaluate control methods.
It also seems to be a natural order for a processor in a device
hierarchy.

On the other hand, some child devices on acpi cpu bus need to access
other system resources like PCI configuration space of chipset devices,
so they need to be probed and attached rather late.
For this reason we probe and attach the cpu bus at
SI_SUB_CONFIGURE:SI_ORDER_MIDDLE SYSINIT level.
In the future this could be done more elegantly via multipass.

Please note that acpi drivers that might access ACPI namespace from
device_identify will do that before _PDC/_OSC of Processors are evaluated.

Legacy cpu driver is not affected by this change.

PR: kern/142561 (in part)
Reviewed by: jhb
Silence from: acpi@
MFC after: 5 weeks

14 years agoRemove redundand headers and use more standard ones where necessary.
wkoszek [Thu, 11 Feb 2010 08:34:41 +0000 (08:34 +0000)]
Remove redundand headers and use more standard ones where necessary.

Reviewed by: hps@

14 years agoUse more standard way for setting nonblocking flag for a filedescriptor.
wkoszek [Thu, 11 Feb 2010 08:30:43 +0000 (08:30 +0000)]
Use more standard way for setting nonblocking flag for a filedescriptor.
This makes libusb porting a bit easier.

There shouldn't by any negative change in behaviour after this commit.

Remove redundant headers.

Reviewed by: hps@

14 years agoWithin libusb 0.1 API, bus number is always faked to 0. Device numbers,
wkoszek [Thu, 11 Feb 2010 08:24:35 +0000 (08:24 +0000)]
Within libusb 0.1 API, bus number is always faked to 0. Device numbers,
however, are possitive and seem to be reverse sorted in the list.

Conform device numbering and bring a result that is consistent with the
libusb 0.1 API.  It is now possible to distinguish a device based on its
(bus, dev) numbers.

There shouldn't be any negative change in behavior after this commit.

Tested with:    scanimage (sane)
http://freebsd.czest.pl/~wkoszek/qemu/l.c
Reviewed by:    hps@

14 years agoBack to WARNS=3. The breakage wasn't what I thought it was :(
imp [Thu, 11 Feb 2010 07:09:04 +0000 (07:09 +0000)]
Back to WARNS=3.  The breakage wasn't what I thought it was :(

14 years agoQuiet spurious warnings.
mckusick [Thu, 11 Feb 2010 06:42:08 +0000 (06:42 +0000)]
Quiet spurious warnings.

14 years agoQuiet spurious warnings.
mckusick [Thu, 11 Feb 2010 06:33:35 +0000 (06:33 +0000)]
Quiet spurious warnings.

14 years agoKeep Tinderbox happy.
mckusick [Thu, 11 Feb 2010 06:22:38 +0000 (06:22 +0000)]
Keep Tinderbox happy.

14 years agoIncreased warnings weren't tested on ARM. Bump warnings back down to
imp [Thu, 11 Feb 2010 04:45:08 +0000 (04:45 +0000)]
Increased warnings weren't tested on ARM.  Bump warnings back down to
0 until it can be properly tested by those raising the warnings.
Remember: make universe is required when changing the WARNS level.

14 years agoEnsure that newfs will never create a filesystem with more than 2^32
mckusick [Wed, 10 Feb 2010 20:17:46 +0000 (20:17 +0000)]
Ensure that newfs will never create a filesystem with more than 2^32
inodes by cutting back on the number of inodes per cylinder group if
necessary to stay under the limit. For a default (16K block) file
system, this limit begins to take effect for file systems above 32Tb.

This fix is in addition to -r203763 which corrected a problem in the
kernel that treated large inode numbers as negative rather than unsigned.
For a default (16K block) file system, this bug began to show up at a
file system size above about 16Tb.

Reported by: Scott Burns, John Kilburg, Bruce Evans
Followup by: Jeff Roberson
PR:          133980
MFC after:   2 weeks

14 years agoThis fix corrects a problem in the file system that treats large
mckusick [Wed, 10 Feb 2010 20:10:35 +0000 (20:10 +0000)]
This fix corrects a problem in the file system that treats large
inode numbers as negative rather than unsigned. For a default
(16K block) file system, this bug began to show up at a file system
size above about 16Tb.

To fully handle this problem, newfs must be updated to ensure that
it will never create a filesystem with more than 2^32 inodes. That
patch will be forthcoming soon.

Reported by: Scott Burns, John Kilburg, Bruce Evans
Followup by: Jeff Roberson
PR:          133980
MFC after:   2 weeks

14 years agoStart sentences with a new line.
trasz [Wed, 10 Feb 2010 19:03:48 +0000 (19:03 +0000)]
Start sentences with a new line.

Submitted by: brueffer

14 years agoRemove unused variable.
trasz [Wed, 10 Feb 2010 18:56:49 +0000 (18:56 +0000)]
Remove unused variable.

14 years agoImprove time precision for grdc(6):
delphij [Wed, 10 Feb 2010 18:20:20 +0000 (18:20 +0000)]
Improve time precision for grdc(6):

Traditionally, grdc would obtain time through time(3) which in turn gets
only the second part of clock (CLOCK_SECOND), and sleep for 1 second after
each screen refresh.

This approach would have two problems.  First, we are not guaranteed to
be waken up at the beginning of a whole second, which will typically
exhibit as a "lag" on second number.  Second, because we sleep for whole
second, and the refresh process would take some time, the error would
accumulate from time to time, making the lag variable.

Make grdc(6) to use time(3) to get time only at the beginning, and sample
time in CLOCK_REALTIME_FAST granularity after refreshing, and use the
nanosecond part to caculate how much time we want to sleep.

PR: bin/120813
MFC after: 1 month

14 years agoImprove description for Giant and mention blocking inside interrupt threads.
trasz [Wed, 10 Feb 2010 17:02:06 +0000 (17:02 +0000)]
Improve description for Giant and mention blocking inside interrupt threads.

14 years agoAdd the options DEADLKRES (introducing the deadlock resolver thread) in
attilio [Wed, 10 Feb 2010 16:30:04 +0000 (16:30 +0000)]
Add the options DEADLKRES (introducing the deadlock resolver thread) in
the 'debugging' section of any HEAD kernel and enable for the mainstream
ones, excluding the embedded architectures.
It may, of course, enabled on a case-by-case basis.

Sponsored by: Sandvine Incorporated
Requested by: emaste
Discussed with: kib

14 years ago- Remove reference to nfs4. mount_nfs4(8) was removed in r192578.
jh [Wed, 10 Feb 2010 16:18:28 +0000 (16:18 +0000)]
- Remove reference to nfs4. mount_nfs4(8) was removed in r192578.
- Add newnfs.

14 years agoIf a mbuf is split across two pages, we
rrs [Wed, 10 Feb 2010 13:48:34 +0000 (13:48 +0000)]
If a mbuf is split across two pages, we
have code that detects this and makes two
transmit descriptors. However its possible
that the algorithm detects when the second
page is not used (when the data aligns perfectly
to the bottom of the page). This caused a 0
len descriptor to be added which locks up the
rge device. Skip such things with a continue.

JC provided this patch... Thanks JC :-)
Obtained from: JC (c.jayachandran@gmail.com)

14 years agoTurn on the front LED at boot time like we do with the Avila.
rpaulo [Wed, 10 Feb 2010 11:40:18 +0000 (11:40 +0000)]
Turn on the front LED at boot time like we do with the Avila.

14 years agoFix typo in comment.
rpaulo [Wed, 10 Feb 2010 11:12:39 +0000 (11:12 +0000)]
Fix typo in comment.

Pointed out by: danfe

14 years ago't' stands for Turbo and is a valid mode, so fix previous commit.
rpaulo [Wed, 10 Feb 2010 11:11:37 +0000 (11:11 +0000)]
't' stands for Turbo and is a valid mode, so fix previous commit.

Pointed out by: sam

14 years agoRemove unneeded CFLAGS.
ed [Wed, 10 Feb 2010 07:15:21 +0000 (07:15 +0000)]
Remove unneeded CFLAGS.

14 years agoCode cleanup:
neel [Wed, 10 Feb 2010 06:57:05 +0000 (06:57 +0000)]
Code cleanup:
- make some variables static
- remove unused variables.

14 years agoCall profclock() and statclock() explicitly on all cpus. Prior to this
neel [Wed, 10 Feb 2010 06:29:43 +0000 (06:29 +0000)]
Call profclock() and statclock() explicitly on all cpus. Prior to this
change these functions were called only on the BSP indirectly via hardclock().

top -P now shows usage statistics of all cpus.

14 years agoIn function umtxq_insert_queue, use parameter q (shared/exclusive queue)
davidxu [Wed, 10 Feb 2010 05:47:34 +0000 (05:47 +0000)]
In function umtxq_insert_queue, use parameter q (shared/exclusive queue)
instead of hard coded constant. This does not affect RELENG_8 and previous,
because the code only exists in the HEAD.

14 years agoEnable interrupts before doing AST processing to avoid a deadlock.
neel [Wed, 10 Feb 2010 05:43:31 +0000 (05:43 +0000)]
Enable interrupts before doing AST processing to avoid a deadlock.

Specifically on an SMP kernel it was observed that if both the
processors are doing an exit1() via ast()->postsig()->sigexit()
then we will deadlock.

This happens because exit1() calls vmspace_exit() that in turn
calls pmap_invalidate_all(). This function tries to do a
smp_rendezvous() which blocks because the other processor is not
responding to IPIs - because it too is doing AST processing with
interrupts disabled.

14 years agoAdd new -o option (r203042) to manpage synopsis and usage().
emaste [Wed, 10 Feb 2010 00:18:15 +0000 (00:18 +0000)]
Add new -o option (r203042) to manpage synopsis and usage().

Submitted by: ru

14 years agoFix PR
cracauer [Wed, 10 Feb 2010 00:02:09 +0000 (00:02 +0000)]
Fix PR
http://www.freebsd.org/cgi/query-pr.cgi?pr=bin/143350
Empty string test gone wrong.

Testing this requires that you have a locale that has the sign string
unset but has int_n_sign_posn set (the default locale falls through to
use "()" around negative numbers which is probably another bug).

I created that setup by hand and indeed without this fix negative
numbers are put out as positive numbers (doesn't fall through to use
"-" as default indicator).

Unfixed example in nl_NL.ISO8859-1 with lc->negative_sign set to empty
string:
  strfmon(buf, sizeof(buf), "%-8i", -42.0);
==>
example2: 'EUR  42,00' 'Eu 42,00'

Fixed:
example2: 'EUR  42,00-' 'Eu 42,00-'

This file and suggested fix are identical in at least freebsd-8.
Backport might be appropriate but some expert on locales should
probably have a look at us defaulting to negative numbers in
parenthesis when LC_* is default.  That doesn't look right and is not
what other OSes are doing.

PR: 143350
Submitted by: Corinna Vinschen
Reviewed by: bug reporter submitted, tested by me

14 years agoSymbolic link to mailwrapper should only be removed if both MK_MAILWRAPPER
delphij [Wed, 10 Feb 2010 00:01:35 +0000 (00:01 +0000)]
Symbolic link to mailwrapper should only be removed if both MK_MAILWRAPPER
and MK_SENDMAIL is "no".

Reported by: Rob Farmer <rfarmer at predatorlabs.net>

14 years ago- Move nfs_realign() from the NFS client to the shared NFS code and
marius [Tue, 9 Feb 2010 23:45:14 +0000 (23:45 +0000)]
- Move nfs_realign() from the NFS client to the shared NFS code and
  remove the NFS server version in order to reduce code duplication.
  The shared version now uses a second parameter how, which is passed
  on to m_get(9) and m_getcl(9) as the server used M_WAIT while the
  client requires M_DONTWAIT, and replaces the the previously unused
  parameter hsiz.
- Change nfs_realign() to use nfsm_aligned() so as with other NFS code
  the alignment check isn't actually performed on platforms without
  strict alignment requirements for performance reasons because as the
  comment suggests unaligned data only occasionally occurs with TCP.
- Change fha_extract_info() to use nfs_realign() with M_DONTWAIT rather
  than M_WAIT because it's called with the RPC sp_lock held.

Reviewed by: jhb, rmacklem
MFC after: 1 week

14 years agoSome style(9) fixes
marius [Tue, 9 Feb 2010 23:40:07 +0000 (23:40 +0000)]
Some style(9) fixes

14 years agoAdd DDB support for printing vnet_sysinit and vnet_sysuninit
bz [Tue, 9 Feb 2010 22:39:34 +0000 (22:39 +0000)]
Add DDB support for printing vnet_sysinit and vnet_sysuninit
ordered call lists. Try to lookup function/symbol names and print
those in addition to the pointers, along with the constants for
subsystem and order.
This is useful for debugging vnet teardown ordering issues.

Make it possible to call the actual printing frunction from normal
code at runtime, ie. from vnet_sysuninit(), if DDB support is there.

Sponsored by: ISPsystem
MFC After: 8 days

14 years ago - Return EAFNOSUPPORT instead of EINVAL for unsupported address family,
delphij [Tue, 9 Feb 2010 22:30:51 +0000 (22:30 +0000)]
 - Return EAFNOSUPPORT instead of EINVAL for unsupported address family,
   this matches the Linux behavior.
 - Check if we have sufficient space allocated for socket structure, which
   fixes a buffer overflow when wrong length is being passed into the
   emulation layer. [1]

PR: kern/138860
Submitted by: Mateusz Guzik <mjguzik gmail com>
Reported by: Alexander Best [1]
MFC after: 2 weeks

14 years agoAdd an SDT provider for "vnet"s along with probes for vnet_alloc
bz [Tue, 9 Feb 2010 22:15:59 +0000 (22:15 +0000)]
Add an SDT provider for "vnet"s along with probes for vnet_alloc
and vnet_destroy.
Use the line number rather than NULL as dummy argument.

Note: the fbt provider does not reliably provide :return probes
(depending on optimization levels used at compile time) making
it unusable for scripts to generate complete call-traces with
well defined boundaries over allocations or destructions of
virtual network stacks.

Sponsored by: ISPsystem
MFC After: 8 days

14 years agoProperly free resources when destroying the TCP hostcache while
bz [Tue, 9 Feb 2010 21:31:53 +0000 (21:31 +0000)]
Properly free resources when destroying the TCP hostcache while
tearing down a network stack (in the VIMAGE jail+vnet case).

For that break out the logic from tcp_hc_purge() into an internal
function we can call from both, the sysctl handler and the
tcp_hc_destroy().

Sponsored by: ISPsystem
Reviewed by: silby, lstewart
MFC After: 8 days

14 years agoDon't let find(1) depend on struct timeb.
ed [Tue, 9 Feb 2010 21:24:41 +0000 (21:24 +0000)]
Don't let find(1) depend on struct timeb.

This structure is deprecated and only used by ftime(2), which is part of
libcompat. The second argument of get_date() is unused, which means we
can just remove it entirely.

14 years agoRemove VOP_VPTOFH(9) from references in VFS(9), as it's a VOP,
trasz [Tue, 9 Feb 2010 21:01:18 +0000 (21:01 +0000)]
Remove VOP_VPTOFH(9) from references in VFS(9), as it's a VOP,
not VFS operation.

14 years agoAdd references to VOP_* man pages to vnode(9).
trasz [Tue, 9 Feb 2010 20:58:39 +0000 (20:58 +0000)]
Add references to VOP_* man pages to vnode(9).

14 years ago- Deal with some special cases [1]
gabor [Tue, 9 Feb 2010 19:37:12 +0000 (19:37 +0000)]
- Deal with some special cases [1]
- style(9) nits

Pointed out by: jilles [1]
Approved by: delphij (mentor)

14 years agoAdd -i to usage()
gavin [Tue, 9 Feb 2010 19:13:45 +0000 (19:13 +0000)]
Add -i to usage()

Noticed by: ru
MFC after: 5 days

14 years agoMove device specific flag configuration to attach routine.
yongari [Tue, 9 Feb 2010 19:12:06 +0000 (19:12 +0000)]
Move device specific flag configuration to attach routine.
The softc obtained in device probe wouldn't be the same one used in
device attach. Drivers should not assume any values stored in softc
structure in probe routine will be available for its attach routine.

14 years agoVarious fixes like spelling, style and syntax.
marcel [Tue, 9 Feb 2010 18:43:20 +0000 (18:43 +0000)]
Various fixes like spelling, style and syntax.

Submitted by: ru (thanks!)

14 years agoAdd description for libefi.
marcel [Tue, 9 Feb 2010 18:40:40 +0000 (18:40 +0000)]
Add description for libefi.

Pointed out by: ru

14 years agoAs it turns out, fmt(1) is being used by the base system as well as the
delphij [Tue, 9 Feb 2010 18:22:26 +0000 (18:22 +0000)]
As it turns out, fmt(1) is being used by the base system as well as the
ports tree extensively and it is probably a good idea to keep it
regardless of NO_MAIL setting.

Reported by: Alexander Best
Reviewed by: antoine
X-MFC-With: r203584

14 years agoWhen you have multiple addresses on the same network on different
imp [Tue, 9 Feb 2010 18:10:56 +0000 (18:10 +0000)]
When you have multiple addresses on the same network on different
interfaces (such as when you are part of a carp pool), and you run
rpcbind -h to restrict which interfaces have rpc services, rpcbind can
none-the-less return addresses that aren't in the -h list.  This patch
enforces the rule that when you specify -h on the command line, then
services returned from rpcbind must be to one of the addresses listed
in -h, or be a loopback address (since localhost is implicit when
running -h).

The root cause of this is the assumption in addrmerge that there can
be only one interface that matches a given network IP address.  This
turns out not to be the case.  To retain historical behavior, I didn't
try to fix the routine to prefer the address that the request came
into, since I didn't know the side effects that might cause in the
normal case.  My quick analysis suggests that it wouldn't be a
problem, but since this code is tricky I opted for the more
conservative patch of only restricting the reply when -h is in effect.

Hence, this change will have no effect when you are running rpcbind
without -h.

Reviewed by: alfred@
Sponsored by: iX Systems
MFC after: 2 weeks

14 years agoUnbreak building kernels with COMPAT_32 enabled. The actual support
marcel [Tue, 9 Feb 2010 17:20:00 +0000 (17:20 +0000)]
Unbreak building kernels with COMPAT_32 enabled. The actual support
for the PT_VM_ENTRY request from 32-bit processes will follow.

Pointy hat: marcel

14 years agoAnything that casts struct sockaddr * to struct sockaddr_foo is safe
imp [Tue, 9 Feb 2010 17:16:13 +0000 (17:16 +0000)]
Anything that casts struct sockaddr * to struct sockaddr_foo is safe
due to careful design.  We've not yet figured out how to properly
annotate the sockaddr structs to communicate this to the compiler and
there's a number of constructs in the tree that make this annotation
challenging.

As such, reduce warns to 3 here because this code really isn't warns 6
safe, even if it kinda sorta appears to be on intel (which has no such
alignment restrictions).  Warns 4 adds the -Wcast-align warning.

# fixes the mips tinderbox build

14 years agoAdd the definition of Mistakeholder
gnn [Tue, 9 Feb 2010 14:51:39 +0000 (14:51 +0000)]
Add the definition of Mistakeholder

14 years ago- pt_BR.ISO8859-1 catalog
gabor [Tue, 9 Feb 2010 10:47:44 +0000 (10:47 +0000)]
- pt_BR.ISO8859-1 catalog
- Add link to pt_PT.ISO8859-1

Submitted by: sylvio
Approved by: delphij (mentor)

14 years agoSet ut_line to "ftpd" for ftpd.
ed [Tue, 9 Feb 2010 07:35:12 +0000 (07:35 +0000)]
Set ut_line to "ftpd" for ftpd.

This makes it a little easier to figure out which application was
responsible for this log entry. Ideally we should add an ut_process or
something similar.

Suggested by: Vincent Poy <vincepoy gmail com>

14 years agoSMP support for the mips port.
neel [Tue, 9 Feb 2010 06:24:43 +0000 (06:24 +0000)]
SMP support for the mips port.

The platform that supports SMP currently is a SWARM with a dual-core Sibyte
processor. The kernel config file to use is SWARM_SMP.

Reviewed by: imp, rrs

14 years agoAdd PT_VM_TIMESTAMP and PT_VM_ENTRY so that the tracing process can
marcel [Tue, 9 Feb 2010 05:52:35 +0000 (05:52 +0000)]
Add PT_VM_TIMESTAMP and PT_VM_ENTRY so that the tracing process can
obtain the memory map of the traced process. PT_VM_TIMESTAMP can be
used to check if the memory map changed since the last time to avoid
iterating over all the VM entries unnecesarily.

MFC after: 1 month

14 years agoFixing compilation bustage by removing a stray comment fragment.
avatar [Tue, 9 Feb 2010 04:07:39 +0000 (04:07 +0000)]
Fixing compilation bustage by removing a stray comment fragment.

14 years agoDisable the use of the IAAD usb doorbell on NVidia controllers as it can cause
thompsa [Tue, 9 Feb 2010 00:38:40 +0000 (00:38 +0000)]
Disable the use of the IAAD usb doorbell on NVidia controllers as it can cause
the hardware to stall.

Submitted by: Hans Petter Selasky

14 years agoKernel modules for these drivers are installed on all platforms, so
gavin [Mon, 8 Feb 2010 23:30:28 +0000 (23:30 +0000)]
Kernel modules for these drivers are installed on all platforms, so
install the man pages on all platforms too.

14 years agoUpdate documentation for the iwn and iwnfw drivers: they support the 1000, 5150,...
brucec [Mon, 8 Feb 2010 21:38:42 +0000 (21:38 +0000)]
Update documentation for the iwn and iwnfw drivers: they support the 1000, 5150, 6000 and 6050 devices too, with firmware modules for the 4965, 1000, 5000, 5150 and 6000.

Add documentation for mwl and all the wireless firmware drivers.

Approved by: rrs (mentor)

14 years agoXorg isn't treated as a distribution, so /usr/X11R6/lib shouldn't be configured when...
brucec [Mon, 8 Feb 2010 21:29:34 +0000 (21:29 +0000)]
Xorg isn't treated as a distribution, so /usr/X11R6/lib shouldn't be configured when running ldconfig.

PR: bin/138945
Approved by: rrs (mentor)
MFC after: 3 days

14 years agoInstall the padlock(4) man page on amd64 as well as i386, to match the
gavin [Mon, 8 Feb 2010 21:24:12 +0000 (21:24 +0000)]
Install the padlock(4) man page on amd64 as well as i386, to match the
platforms where the driver itself is compiled and installed.

PR: docs/130895
Reported by: George Hartzell <hartzell alerce.com>
MFC after: 1 week

14 years agoInitialize the execfile argument to NULL instead of _PATH_DEVNULL. This allows the...
brucec [Mon, 8 Feb 2010 21:23:48 +0000 (21:23 +0000)]
Initialize the execfile argument to NULL instead of _PATH_DEVNULL. This allows the -M option to be used without specifying -N.

PR: bin/138146
Approved by: rrs (mentor)
MFC after: 3 days

14 years agoUpdate .Dt to reflect the fact that these drivers and man pages are
gavin [Mon, 8 Feb 2010 21:01:41 +0000 (21:01 +0000)]
Update .Dt to reflect the fact that these drivers and man pages are
installed on more than just i386.

14 years agoIf there is only one NIC in the system that is up and running, the
mbr [Mon, 8 Feb 2010 20:57:49 +0000 (20:57 +0000)]
If there is only one NIC in the system that is up and running, the
interface specifier on the command line can be ommited.

Besides of this, the bpf is being reused for each machine
that has to be woken up.

Submitted by: Marc Balmer <marc@msys.ch>

14 years agoDocument the usfs driver and the NO_SYSCTL_DESCR option, and update the comment for...
brucec [Mon, 8 Feb 2010 20:57:42 +0000 (20:57 +0000)]
Document the usfs driver and the NO_SYSCTL_DESCR option, and update the comment for umass.
Don't include the sysctl description variables in aic7xxx when NO_SYSCTL_DESCR is used.

Approved by: rrs (mentor)

14 years agoUpdate .Dt on these man pages: the kernel modules and corresponding man
gavin [Mon, 8 Feb 2010 20:53:29 +0000 (20:53 +0000)]
Update .Dt on these man pages: the kernel modules and corresponding man
pages are installed on more platforms than just i386.

14 years agoAdd multicast key search support. This fixes corrupted mcast packets
rpaulo [Mon, 8 Feb 2010 20:23:20 +0000 (20:23 +0000)]
Add multicast key search support. This fixes corrupted mcast packets
when we have more than one hostap vap.

Submitted by: Russell Yount <russell.yount at gmail.com>
MFC after: 2 weeks

14 years agoFix TX power problems with AR9285.
rpaulo [Mon, 8 Feb 2010 20:12:01 +0000 (20:12 +0000)]
Fix TX power problems with AR9285.

14 years agoFix typo and remove extra spaces.
brucec [Mon, 8 Feb 2010 20:04:16 +0000 (20:04 +0000)]
Fix typo and remove extra spaces.

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

14 years agoFix typo in comment.
rpaulo [Mon, 8 Feb 2010 20:04:02 +0000 (20:04 +0000)]
Fix typo in comment.

14 years agoRemove the usb2_input_kbd directive that was missed during the renaming of the driver...
brucec [Mon, 8 Feb 2010 19:48:33 +0000 (19:48 +0000)]
Remove the usb2_input_kbd directive that was missed during the renaming of the drivers in the usb2 stack.

Approved by: rrs (mentor)

14 years agoRename usb2_ structures and variables to usb_.
brucec [Mon, 8 Feb 2010 19:44:09 +0000 (19:44 +0000)]
Rename usb2_ structures and variables to usb_.

Approved by: rrs (mentor)
Discussed with: hps

14 years agosh: Make sure the mail2.0 test can actually fail if $MAIL is not touched.
jilles [Mon, 8 Feb 2010 18:57:18 +0000 (18:57 +0000)]
sh: Make sure the mail2.0 test can actually fail if $MAIL is not touched.

14 years agoIntroduce new rc.conf variable firewall_coscripts. It can be used to
emax [Mon, 8 Feb 2010 18:51:24 +0000 (18:51 +0000)]
Introduce new rc.conf variable firewall_coscripts. It can be used to
specify list of executables and/or rc scripts that should be executed
after firewall starts/stops.

Submitted by: Yuri Kurenkov <y dot kurenkov at init dot ru>
Reviewed by: rhodes, rc@
MFC after: 1 week

14 years agoEnsure that tkip_mixing_phase1() is called after a rekeying event when
bschmidt [Mon, 8 Feb 2010 18:16:59 +0000 (18:16 +0000)]
Ensure that tkip_mixing_phase1() is called after a rekeying event when
useing plain s/w crypto.

Approved by: rpaulo (mentor)
Reviewed by: sam

14 years agoMake sure that FTS_COMFOLLOW is not set when the -P option is in effect.
jh [Mon, 8 Feb 2010 15:42:55 +0000 (15:42 +0000)]
Make sure that FTS_COMFOLLOW is not set when the -P option is in effect.
Otherwise the -i option will show the inode number of the referenced file
for symbolic links given on the command line. Similarly, the file color
was printed according to the link target in colorized output.

PR: bin/102394
Reviewed by: jilles
MFC after: 2 weeks

14 years agoRemove unused LIBCOMPAT keyword from syscalls.master.
ed [Mon, 8 Feb 2010 10:02:01 +0000 (10:02 +0000)]
Remove unused LIBCOMPAT keyword from syscalls.master.

14 years agoFall back to ASCII codepoints for box drawing.
ed [Mon, 8 Feb 2010 09:16:59 +0000 (09:16 +0000)]
Fall back to ASCII codepoints for box drawing.

Even though the default VGA font provides box drawing fonts, there is no
guarantee any font will provide these as well (i.e. ISO-8859-*, KOI8-R).
Just use ASCII characters for box drawing.

PR: kern/141633

14 years agoSet waiters flag before checking semaphore's counter,
davidxu [Mon, 8 Feb 2010 07:31:05 +0000 (07:31 +0000)]
Set waiters flag before checking semaphore's counter,
otherwise we might lose a wakeup. Tested on postgresql database server.

14 years agoImprove checking whether an ARM VA has a valid mapping before performing cache
raj [Sun, 7 Feb 2010 20:48:57 +0000 (20:48 +0000)]
Improve checking whether an ARM VA has a valid mapping before performing cache
sync.

VIPT/PIPT caches need valid VA-PA mapping in PTE for a cache operation to
succeed (unlike VIVT). Prior to this fix pmap was using l2pte_valid() for that
check, but this is not sufficient as the function merely checks if a PTE
exists (there can be existing but _invalid_ entries in the table).

A new pmap_has_valid_mapping() routine is introduced to do this job right by
checking proper PTE flags.

Among other potential problems this cures coherency issues with L2 caches on
MV-78100.

Submitted by: Grzegorz Bernacki, Piotr Ziecik
Reviewed, tested by: marcel
Obtained from: Semihalf
MFC after: 1 week

14 years agoCorrect arguments to free_unr(), "item" was missing.
gavin [Sun, 7 Feb 2010 20:36:23 +0000 (20:36 +0000)]
Correct arguments to free_unr(), "item" was missing.

MFC after: 1 week

14 years agomerge from my tbemd branch: cmpdi2 and ucmpdi2 are genereated when the
imp [Sun, 7 Feb 2010 19:49:01 +0000 (19:49 +0000)]
merge from my tbemd branch: cmpdi2 and ucmpdi2 are genereated when the
inline limit is pushed to a very low level.  Include them here until
we can find a better way to optionally include them.  They are small...

14 years agoAdd support for a few more Sony-specific ACPI features (default display
gavin [Sun, 7 Feb 2010 18:36:30 +0000 (18:36 +0000)]
Add support for a few more Sony-specific ACPI features (default display
brightness, wired LAN power and bass gain), and update the description of
one previously unknown feature (display contrast).  While here, expand on
a comment and remove two defines left over from an old version of the code.

Also update man page to document the above changes, and correct grammar.

PR: kern/127581

14 years agoBump .Dd for r203620
gavin [Sun, 7 Feb 2010 18:06:55 +0000 (18:06 +0000)]
Bump .Dd for r203620

MFC after: 1 week

14 years agoDocument support for the D-Link DFE520-TX card (supported with the vr(4)
gavin [Sun, 7 Feb 2010 18:05:12 +0000 (18:05 +0000)]
Document support for the D-Link DFE520-TX card (supported with the vr(4)
driver)

PR: kern/135989
Submitted by: "Rashid N. Achilov"  citycat4 ngs.ru

14 years agoSpelling nit
gavin [Sun, 7 Feb 2010 18:00:13 +0000 (18:00 +0000)]
Spelling nit

14 years agoTell the compiler these structures are aligned to a byte boundary.
imp [Sun, 7 Feb 2010 17:05:22 +0000 (17:05 +0000)]
Tell the compiler these structures are aligned to a byte boundary.
All the elements of these structs are char anyway, so it won't hurt
performance.

Bump warns back up to the default.

# we likely should have CTASSERTS to make sure they are the right size.
# but with libarchive based tar maybe we shouldn't bother.

14 years agoRemove statistics from the TTY queues.
ed [Sun, 7 Feb 2010 15:42:15 +0000 (15:42 +0000)]
Remove statistics from the TTY queues.

I added counters to see how often fast copying to userspace was actually
performed, which was only useful during development. Remove these
statistics now we know it to be effective.

14 years agoFix whitespace in pho's entry, results in nicer output files.
brueffer [Sun, 7 Feb 2010 13:59:03 +0000 (13:59 +0000)]
Fix whitespace in pho's entry, results in nicer output files.

PR: 143613
Submitted by: pluknet <pluknet@gmail.com>
Committed from: Debian/kFreeBSD talk at FOSDEM

14 years agoInitialize fromlen before calling recvfrom to avoid passing in random
imp [Sun, 7 Feb 2010 07:50:41 +0000 (07:50 +0000)]
Initialize fromlen before calling recvfrom to avoid passing in random
stack garbage.

Obtained from: NetBSD 1.13

14 years agoSimplify, remove unnecessary code.
kientzle [Sun, 7 Feb 2010 02:04:58 +0000 (02:04 +0000)]
Simplify, remove unnecessary code.