]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoImprove markup.
Ruslan Ermilov [Thu, 21 Dec 2006 22:44:41 +0000 (22:44 +0000)]
Improve markup.

17 years agoDocument MTX_NOPROFILE flag.
Pawel Jakub Dawidek [Thu, 21 Dec 2006 22:42:18 +0000 (22:42 +0000)]
Document MTX_NOPROFILE flag.

17 years agoAllow this module to get its options from the kernel build directory
Yaroslav Tykhiy [Thu, 21 Dec 2006 21:35:49 +0000 (21:35 +0000)]
Allow this module to get its options from the kernel build directory
instead of always hard coding them in CFLAGS.  POLA is kept here:
The module file built with GENERIC stays the same.

Tested with: md5(1)

17 years agoPay attention to return value from xpt_bus_register in xpt_init.
Matt Jacob [Thu, 21 Dec 2006 20:06:30 +0000 (20:06 +0000)]
Pay attention to return value from xpt_bus_register in xpt_init.

Obtained from: Xin Li (Coverity)
MFC after: 3 days

17 years agoThe prepend function did not handle non-pkthdr's correctly.
Randall Stewart [Thu, 21 Dec 2006 19:58:04 +0000 (19:58 +0000)]
The prepend function did not handle non-pkthdr's correctly.
It always called MH_ALIGN for small lengths being
prepended (less than MHLEN). This meant that if you did
a prepend on a non M_PKTHDR the system would panic with
the KASSERT in MH_ALIGN. Instead we are not aware of
this and do a MH_ALIGN or M_ALIGN as appropriate.

Reviewed by: andre
Approved by: gnn

17 years agoCorrect the description of minpoll and maxpoll.
Ceri Davies [Thu, 21 Dec 2006 19:08:25 +0000 (19:08 +0000)]
Correct the description of minpoll and maxpoll.

Note that while later versions of the ntpd documentation use the term
"dual logarithm", the text added here is consistent with the remainder
of the current document.

PR: docs/106926
Submitted by: Jeremy Chadwick

17 years agoRemove duplicate "clear" subcommand.
Ceri Davies [Thu, 21 Dec 2006 18:30:23 +0000 (18:30 +0000)]
Remove duplicate "clear" subcommand.

PR: docs/106947
Submitted by: TAOKA Fumiyoshi

17 years agoMFP4 (110956):
Alexander Leidinger [Thu, 21 Dec 2006 13:11:06 +0000 (13:11 +0000)]
MFP4 (110956):
Add definition for LINUX_MSG_INFO.

This fixes the tinderbox errors.

Submitted by: rdivacky

17 years agoImprove markup and do some minor wordsmithing.
Ruslan Ermilov [Thu, 21 Dec 2006 10:59:48 +0000 (10:59 +0000)]
Improve markup and do some minor wordsmithing.

17 years agoReturn value PKT_ALIAS_FOUND_HEADER_FRAGMENT isn't an error case. The
Gleb Smirnoff [Thu, 21 Dec 2006 10:26:01 +0000 (10:26 +0000)]
Return value PKT_ALIAS_FOUND_HEADER_FRAGMENT isn't an error case. The
packet shouldn't be dropped.

Submitted by: Alexander Motin <mav alkar.net>

17 years agoMinor style fixes.
Robert Watson [Thu, 21 Dec 2006 09:58:02 +0000 (09:58 +0000)]
Minor style fixes.

17 years agoRemove mac_enforce_subsystem debugging sysctls. Enforcement on
Robert Watson [Thu, 21 Dec 2006 09:51:34 +0000 (09:51 +0000)]
Remove mac_enforce_subsystem debugging sysctls.  Enforcement on
subsystems will be a property of policy modules, which may require
access control check entry points to be invoked even when not actively
enforcing (i.e., to track information flow without providing
protection).

Obtained from: TrustedBSD Project
Suggested by: Christopher dot Vance at sparta dot com

17 years agoUnbreak 64-bit little-endian systems that do require alignment.
Marcel Moolenaar [Thu, 21 Dec 2006 05:40:46 +0000 (05:40 +0000)]
Unbreak 64-bit little-endian systems that do require alignment.
The fix involves using le16dec(), le32dec(), le16enc() and
le32enc(). This eliminates invalid casts and duplicated logic.

17 years agoia64 is nit-picking according to tinderbox so temporary disconnect
Bjoern A. Zeeb [Thu, 21 Dec 2006 01:58:22 +0000 (01:58 +0000)]
ia64 is nit-picking according to tinderbox so temporary disconnect
the bridge module from the build until this is fixed.

17 years agoComment LABEL_TO_SLOT() macro, including observing that we'd like to improve
Robert Watson [Wed, 20 Dec 2006 23:41:59 +0000 (23:41 +0000)]
Comment LABEL_TO_SLOT() macro, including observing that we'd like to improve
this policy API to avoid encoding struct label binary layout in policy
modules.

Obtained from: TrustedBSD Project

17 years agoTrim trailing white space, clean up comment line wrapping and formatting.
Robert Watson [Wed, 20 Dec 2006 23:18:17 +0000 (23:18 +0000)]
Trim trailing white space, clean up comment line wrapping and formatting.

Document mac_associate_nfsd_label().

Obtained from: TrustedBSD Project

17 years agoTrim trailing white space.
Robert Watson [Wed, 20 Dec 2006 23:17:34 +0000 (23:17 +0000)]
Trim trailing white space.

17 years agoDocument socket labeling model.
Robert Watson [Wed, 20 Dec 2006 23:16:41 +0000 (23:16 +0000)]
Document socket labeling model.

Clean up comment white space and wrapping.

Obtained from: TrustedBSD Project

17 years agoClean up comment white space and line wrapping.
Robert Watson [Wed, 20 Dec 2006 23:16:01 +0000 (23:16 +0000)]
Clean up comment white space and line wrapping.

17 years agoAdditional comments regarding the interaction between the kernel privilege
Robert Watson [Wed, 20 Dec 2006 23:15:27 +0000 (23:15 +0000)]
Additional comments regarding the interaction between the kernel privilege
model and the MAC Framework.

Obtained from: TrustedBSD Project

17 years agoDocument that we could allocate the mbuf label as part of the tag rather
Robert Watson [Wed, 20 Dec 2006 23:14:33 +0000 (23:14 +0000)]
Document that we could allocate the mbuf label as part of the tag rather
than from the slab, but don't.

Document mac_mbuf_to_label(), mac_copy_mbuf_tag().

Clean up white space/wrapping for other comments.

Obtained from: TrustedBSD Project

17 years agoStaticize and comment zone_label.
Robert Watson [Wed, 20 Dec 2006 23:13:04 +0000 (23:13 +0000)]
Staticize and comment zone_label.

Obtained from: TrustedBSD Project

17 years agoClean up comments, trailing white space.
Robert Watson [Wed, 20 Dec 2006 23:12:36 +0000 (23:12 +0000)]
Clean up comments, trailing white space.

Provide a comment describing MAC_EXTERNALIZE().

Obtained from: TrustedBSD Project

17 years agoRe-wrap comment at 77 character columns.
Robert Watson [Wed, 20 Dec 2006 23:11:01 +0000 (23:11 +0000)]
Re-wrap comment at 77 character columns.

17 years agoAdd a (disabled) configuration line to enable snmp_bridge(3) module.
Shteryana Shopova [Wed, 20 Dec 2006 22:10:34 +0000 (22:10 +0000)]
Add a (disabled) configuration line to enable snmp_bridge(3) module.

Approved by: bz (mentor)

17 years agoConnect the snmp_bridge(3) module to the build.
Shteryana Shopova [Wed, 20 Dec 2006 22:08:35 +0000 (22:08 +0000)]
Connect the snmp_bridge(3) module to the build.

Approved by: bz (mentor)

17 years agoKeep in sync with the if_bridge(4) module (rev. 1.20 if_bridgevar.h,
Shteryana Shopova [Wed, 20 Dec 2006 22:02:45 +0000 (22:02 +0000)]
Keep in sync with the if_bridge(4) module (rev. 1.20 if_bridgevar.h,
1.12 bridgestp.h) and rename all PointToPoint related variables
from P2P to PTP (s/P2P/PTP/g s/p2p/ptp/g).

Approved by: bz (mentor)

17 years agoRemove an unused variable.
Shteryana Shopova [Wed, 20 Dec 2006 21:31:28 +0000 (21:31 +0000)]
Remove an unused variable.

Approved by: bz (mentor)

17 years agoComment and white space cleanup.
Robert Watson [Wed, 20 Dec 2006 20:43:19 +0000 (20:43 +0000)]
Comment and white space cleanup.

Exapnd comments on System V IPC labeling methods, which could use improved
consistency with respect to other object types.

Obtained from: TrustedBSD Project

17 years agoRegen (just to fix 'generated from' line from the previous commit).
Jung-uk Kim [Wed, 20 Dec 2006 20:42:58 +0000 (20:42 +0000)]
Regen (just to fix 'generated from' line from the previous commit).

17 years agoExternalize local stack copy of the ifnet label, rather than the copy on
Robert Watson [Wed, 20 Dec 2006 20:40:29 +0000 (20:40 +0000)]
Externalize local stack copy of the ifnet label, rather than the copy on
the ifnet itself.  The stack copy has been made while holding the mutex
protecting ifnet labels, so copying from the ifnet copy could result in
an inconsistent version being copied out.

Reported by: Todd.Miller@sparta.com
Obtained from: TrustedBSD Project
MFC after: 3 weeks

17 years agoExpand commenting on label slots, justification for the MAC Framework locking
Robert Watson [Wed, 20 Dec 2006 20:38:44 +0000 (20:38 +0000)]
Expand commenting on label slots, justification for the MAC Framework locking
model, interactions between locking and policy init/destroy methods.

Rewrap some comments to 77 character line wrap.

Obtained from: TrustedBSD Project

17 years agoAdd linux_nanosleep() and regen.
Jung-uk Kim [Wed, 20 Dec 2006 20:21:48 +0000 (20:21 +0000)]
Add linux_nanosleep() and regen.

17 years agoreduce padding to compensate for recent change to sys/pcpu.h (tinderbox fix)
Kip Macy [Wed, 20 Dec 2006 20:18:07 +0000 (20:18 +0000)]
reduce padding to compensate for recent change to sys/pcpu.h (tinderbox fix)

17 years agoMFP4: 109655
Jung-uk Kim [Wed, 20 Dec 2006 20:17:35 +0000 (20:17 +0000)]
MFP4: 109655

- Move linux_nanosleep() from src/sys/amd64/linux32/linux32_machdep.c to
src/sys/compat/linux/linux_time.c.
- Validate timespec ranges before use as Linux kernel does.
- Fix l_timespec structure.
- Clean up style(9) nits.

17 years agoMFP4: 110179
Jung-uk Kim [Wed, 20 Dec 2006 20:08:45 +0000 (20:08 +0000)]
MFP4: 110179

Add rudimentary IPC_INFO/MSG_INFO command support for linux_msgctl()
to pacify Linux ipcs(1).  While I am here, add more bound checks
for linux_msgsnd() and linux_msgrcv().

17 years agoRegen.
Jung-uk Kim [Wed, 20 Dec 2006 19:39:10 +0000 (19:39 +0000)]
Regen.

17 years agoMFP4: (part of) 110058
Jung-uk Kim [Wed, 20 Dec 2006 19:36:03 +0000 (19:36 +0000)]
MFP4: (part of) 110058

Fix 32-bit msgsnd(3) and msgrcv(3) emulations for amd64.

17 years agoMFP4: (part of) 110058
Jung-uk Kim [Wed, 20 Dec 2006 19:30:52 +0000 (19:30 +0000)]
MFP4: (part of) 110058

Use new kern_msgsnd()/kern_msgrcv() to fix linux32 emulation on amd64.

17 years agoMFP4: (part of) 110058
Jung-uk Kim [Wed, 20 Dec 2006 19:26:30 +0000 (19:26 +0000)]
MFP4: (part of) 110058

copyin()/copyout() for message type is separated from msgsnd()/msgrcv() and
it is done from its wrapper functions to support 32-bit emulations.  After I
implemented this, I have briefly referenced NetBSD and Darwin.  NetBSD passes
copyin()/copyout() function pointers from wrappers.  Darwin passes size of
message type as an argument, which is actually similar to my first
implementation (P4 109706).  We may revisit these implementations later.

17 years agoMFP4: (part of) 109714
Jung-uk Kim [Wed, 20 Dec 2006 19:00:52 +0000 (19:00 +0000)]
MFP4: (part of) 109714

Add SYSCALL_MODULE_PRESENT() macro.  The idea was borrowed from
syscall_register().

17 years agoMFp4: differences for bwct ethernet attachment
Warner Losh [Wed, 20 Dec 2006 18:26:37 +0000 (18:26 +0000)]
MFp4: differences for bwct ethernet attachment

17 years agoMFp4: Differences in flash part for bwct. need a more generic way to cope.
Warner Losh [Wed, 20 Dec 2006 18:25:16 +0000 (18:25 +0000)]
MFp4: Differences in flash part for bwct.  need a more generic way to cope.

17 years agoMFp4: Add timeout to eeprom access for lame eeprom that go awol
Warner Losh [Wed, 20 Dec 2006 18:19:52 +0000 (18:19 +0000)]
MFp4: Add timeout to eeprom access for lame eeprom that go awol

17 years agoMFp4: bwct memory size and PLL parameters
Warner Losh [Wed, 20 Dec 2006 18:18:24 +0000 (18:18 +0000)]
MFp4: bwct memory size and PLL parameters

17 years agoMFp4: bwct boot rom is different. need a more generic way to cope long term.
Warner Losh [Wed, 20 Dec 2006 18:16:49 +0000 (18:16 +0000)]
MFp4: bwct boot rom is different.  need a more generic way to cope long term.

17 years agoMFp4: Delay a second or two after the upload before printing Done.
Warner Losh [Wed, 20 Dec 2006 17:50:02 +0000 (17:50 +0000)]
MFp4: Delay a second or two after the upload before printing Done.
Add an automatic reset for remote operational luvin' goodness.

17 years agoMFp4: bwct is a new board choice.
Warner Losh [Wed, 20 Dec 2006 17:47:54 +0000 (17:47 +0000)]
MFp4: bwct is a new board choice.

17 years agoOn amd64 platform, use linux32 headers so 32-bit Linux applications
Xin LI [Wed, 20 Dec 2006 17:10:53 +0000 (17:10 +0000)]
On amd64 platform, use linux32 headers so 32-bit Linux applications
would be able to work with aac(4).

This approach is used by some other drivers as well.  However, we
need a more generic way to do this in order to avoid having to
special case headers in individual drivers for each platform.

Obtained from: Adaptec (version b11518)
Approved by: scottl

17 years agoMarkup cosmetics.
Ruslan Ermilov [Wed, 20 Dec 2006 16:57:13 +0000 (16:57 +0000)]
Markup cosmetics.

17 years agoDocument some details better, making it easier to translate.
Ruslan Ermilov [Wed, 20 Dec 2006 15:34:01 +0000 (15:34 +0000)]
Document some details better, making it easier to translate.

17 years agoSyscons cannot be stopped, so provide a no-op stop method.
Yaroslav Tykhiy [Wed, 20 Dec 2006 12:59:50 +0000 (12:59 +0000)]
Syscons cannot be stopped, so provide a no-op stop method.
The default stop method from rc.subr isn't suited for this
case and produces a bogus warning: "syscons not running".

Suggested by: matteo

17 years agoIn bge_txeof(), cancel the watchdog timeout if all descriptors have
Bruce Evans [Wed, 20 Dec 2006 12:03:21 +0000 (12:03 +0000)]
In bge_txeof(), cancel the watchdog timeout if all descriptors have
been handled instead of when at least one descriptor was just handled.
For bge, it is normal to get a txeof when only a small fraction of the
queued tx descriptors have been handled, so the bug broke the watchdog
in a usual case.

17 years agoBe more accurate in the description of the -I option:
Ruslan Ermilov [Wed, 20 Dec 2006 11:57:22 +0000 (11:57 +0000)]
Be more accurate in the description of the -I option:
signaling to a process doesn't necessarily kill it.

17 years agoImprove rc.d conformance:
Yaroslav Tykhiy [Wed, 20 Dec 2006 11:37:15 +0000 (11:37 +0000)]
Improve rc.d conformance:
- don't play a needless trick with prestart, just use start method;
- provide no-op stop method so that we don't get bogus "abi not running" error.

17 years agoAvoid a race and a pessimization in bge_intr():
Bruce Evans [Wed, 20 Dec 2006 11:14:45 +0000 (11:14 +0000)]
Avoid a race and a pessimization in bge_intr():

- moved the synchronizing bus read to after the bus write for the first
  interrupt ack so that it actually synchronizes everything necessary.

  We were acking not only the status update that triggered the interrupt
  together with any status updates that occurred before we got around
  to the bus write for the ack, but also any status updates that occur
  after we do the bus write but before the write reaches the device.
  The corresponding race for the second interrupt ack resulted in
  sometimes returning from the interrupt handler with acked but
  unserviced interrupt events.  Such events then remain unserviced
  until further events cause another interrupt or the watchdog times
  out.

  The race was often lost on my 5705, apparently since my 5705 has broken
  event coalescing which causes a status update for almost every packet,
  so another status update is quite likely to occur while the interrupt
  handler is running.  Watchdog timeouts weren't very noticeable,
  apparently because bge_txeof() has one of the usual bugs resetting the
  watchdog.

- don't disable device interrupts while bge_intr() is running.  Doing this
  just had the side effects of:
  - entering a device mode in which different coalescing parameters apply.
    Different coalescing parameters can be used to either inhibit or
    enhance the chance of getting another status update while in the
    interrupt handler.  This feature is useless with the current
    organization of the interrupt handler but might be useful with a
    taskqueue handler.
  - giving a race for ack+reenable/return.  This cannot be handled
    by simply rearranging the order of bus accesses like the race for
    ack+keepenable/entry.  It is necessary to sync the ack and then
    check for new events.
  - taking longer, especially with the extra code to avoid the race on
    ack+reenable/return.

Reviewed by: ru, gleb, scottl

17 years agoIn rev. 1.514, iodone on async buffer may happen before code checks the
Konstantin Belousov [Wed, 20 Dec 2006 09:22:31 +0000 (09:22 +0000)]
In rev. 1.514, iodone on async buffer may happen before code checks the
vnode v_flag. For cluster buffers this would result in dereferencing NULL
b_vp. To prevent the panic, cache relevant vnode flag before calling
bstrategy.

Reported by: Peter Holm, kris
Tested by: Peter Holm
Reviewed by: tegge
Pointy hat to: kib

17 years agoRemove -F option from getopts string -- this option has never done
Colin Percival [Wed, 20 Dec 2006 06:56:25 +0000 (06:56 +0000)]
Remove -F option from getopts string -- this option has never done
anything apart from invoking usage(), and apparently slipped in by
accident.

Approved by: kientzle
MFC after: 3 days

17 years agoMake grammar a bit more consistent in this document.
Yaroslav Tykhiy [Wed, 20 Dec 2006 06:21:51 +0000 (06:21 +0000)]
Make grammar a bit more consistent in this document.

17 years agoAllow for module-path being a semicolon-separated list of dirs.
Yaroslav Tykhiy [Wed, 20 Dec 2006 06:20:04 +0000 (06:20 +0000)]
Allow for module-path being a semicolon-separated list of dirs.
This is consistent with kern.module_path sysctl and also compensates
for the unconventional syntax of asf(8) where the last of multiple
arguments is the output file, which prevents us from using the
traditional Unix syntax "foo file ..." to specify multiple module
dirs.

Submitted by: emaste
MFC after: 1 week

17 years agoget LIBPTHREAD_ADAPTIVE_SPIN early, so it can be used for some global
David Xu [Wed, 20 Dec 2006 05:05:44 +0000 (05:05 +0000)]
get LIBPTHREAD_ADAPTIVE_SPIN early, so it can be used for some global
mutexes.

17 years agoCheck environment variable PTHREAD_ADAPTIVE_SPIN, if it is set, use
David Xu [Wed, 20 Dec 2006 04:43:34 +0000 (04:43 +0000)]
Check environment variable PTHREAD_ADAPTIVE_SPIN, if it is set, use
it as a default spin cycle count.

17 years agoAdd a lwpid field into per-cpu structure, the lwpid represents current
David Xu [Wed, 20 Dec 2006 04:40:39 +0000 (04:40 +0000)]
Add a lwpid field into per-cpu structure, the lwpid represents current
running thread's id on each cpu. This allow us to add in-kernel adaptive
spin for user level mutex. While spinning in user space is possible,
without correct thread running state exported from kernel, it hardly
can be implemented efficiently without wasting cpu cycles, however
exporting thread running state unlikely will be implemented soon as
it has to design and stablize interfaces. This implementation is
transparent to user space, it can be disabled dynamically. With this
change, mutex ping-pong program's performance is improved massively on
SMP machine. performance of mysql super-smack select benchmark is increased
about 7% on Intel dual dual-core2 Xeon machine, it indicates on systems
which have bunch of cpus and system-call overhead is low (athlon64, opteron,
and core-2 are known to be fast), the adaptive spin does help performance.

Added sysctls:
    kern.threads.umtx_dflt_spins
        if the sysctl value is non-zero, a zero umutex.m_spincount will
        cause the sysctl value to be used a spin cycle count.
    kern.threads.umtx_max_spins
        the sysctl sets upper limit of spin cycle count.

Tested on: Athlon64 X2 3800+, Dual Xeon 5130

17 years agoBack out rev. 1.266. The real cause for the recent panics has been fixed
Martin Blapp [Wed, 20 Dec 2006 02:49:59 +0000 (02:49 +0000)]
Back out rev. 1.266. The real cause for the recent panics has been fixed
in rev. 1.267 and there is no need to keep this test.

17 years ago- Use the re_tick() callout instead of if_slowtimo() for driving
Marius Strobl [Wed, 20 Dec 2006 02:13:59 +0000 (02:13 +0000)]
- Use the re_tick() callout instead of if_slowtimo() for driving
  re_watchdog() in order to avoid races accessing if_timer.
- Use bus_get_dma_tag() so re(4) works on platforms requiring it.
- Remove invalid BUS_DMA_ALLOCNOW when creating the parent DMA tag
  and the tags that are used for static memory allocations.
- Don't bother to set if_mtu to ETHERMTU, ether_ifattach() does that.
- Remove an unused variable in re_intr().

17 years agoFix a bug originally introduced in rev. 1.74; don't reloaded the
Marius Strobl [Wed, 20 Dec 2006 01:49:56 +0000 (01:49 +0000)]
Fix a bug originally introduced in rev. 1.74; don't reloaded the
watchdog timer in dc_txeof() in case there are still unhandled
descriptors as dc_poll() invokes dc_poll() unconditionally.
Otherwise this would result in the watchdog timer constantly being
being reloaded and thus circumvent that the watchdog ever fires in
the DEVICE_POLLING case.

Pointed out by: bde

17 years agoPartially back out rev. 1.148 and use new ETHER_BPF_MTAP() macro for VLAN.
Jung-uk Kim [Wed, 20 Dec 2006 01:12:07 +0000 (01:12 +0000)]
Partially back out rev. 1.148 and use new ETHER_BPF_MTAP() macro for VLAN.
Remaining changes are cosmetic.

17 years agoRemove bogus increment of re-hashed PTEG index. This snuck in with r1.12 of
Peter Grehan [Wed, 20 Dec 2006 01:10:21 +0000 (01:10 +0000)]
Remove bogus increment of re-hashed PTEG index. This snuck in with r1.12 of
pmap.c, and is potentially the cause of hangs reported on machines with a
small amount of memory. On machines with sufficient RAM, and without a lot
of processes running, this situation would probably never occur.

Testing is still incomplete, but it is obviously wrong so remove the
offending code now.

The issue of what to do when both the primary and secondary hash overflow
is still open.

Reported by: Dan Kresja at windriver dot com, via alc

17 years agoUse BMSR for link status in one more place and clean up more.
Jung-uk Kim [Wed, 20 Dec 2006 00:34:12 +0000 (00:34 +0000)]
Use BMSR for link status in one more place and clean up more.

17 years ago- Do not depend on auto negotiation for link speed/duplex status.
Jung-uk Kim [Wed, 20 Dec 2006 00:08:47 +0000 (00:08 +0000)]
- Do not depend on auto negotiation for link speed/duplex status.
- Read link status from BMSR instead of auxilary status register.
- Clean up style(9) nits.

17 years agoClear full-duplex when half-duplex flag is set. This actually makes
Jung-uk Kim [Tue, 19 Dec 2006 22:50:49 +0000 (22:50 +0000)]
Clear full-duplex when half-duplex flag is set.  This actually makes
'mediaopt half-duplex' working as it should.  It is now equivalent of
'-mediaopt full-duplex'.

17 years agoGiant might have been temporarily dropped while waiting for proctree_lock, allowing...
Martin Blapp [Tue, 19 Dec 2006 22:34:32 +0000 (22:34 +0000)]
Giant might have been temporarily dropped while waiting for proctree_lock, allowing for an
intervening tty_close() that cleared tp->t_session.

Submitted by: tegge
MFC: 1 day

17 years agoadd a missing the...
John-Mark Gurney [Tue, 19 Dec 2006 20:20:39 +0000 (20:20 +0000)]
add a missing the...

MFC after: 3 days

17 years agoRemove dependency on deprecated if_watchdog ABI.
Bruce M Simpson [Tue, 19 Dec 2006 17:37:41 +0000 (17:37 +0000)]
Remove dependency on deprecated if_watchdog ABI.
Tested with a Sitecom RT2661 based card.

17 years agoFix distorted sound on Fujitsu Siemens AMILO Pa 1510.
Ariff Abdullah [Tue, 19 Dec 2006 16:52:09 +0000 (16:52 +0000)]
Fix distorted sound on Fujitsu Siemens AMILO Pa 1510.

Reported/Tested by: infofarmer

17 years ago- Add geography and spanish ports categories
Erwin Lansing [Tue, 19 Dec 2006 16:46:44 +0000 (16:46 +0000)]
- Add geography and spanish ports categories
- Use "related to" instead of "relating to" consistently [1]

Submitted by:   Rostislav Krasny <rosti.bsd@gmail.com> [1]
Reviewed by:    simon
MFC after:      3 days

17 years agoAdd the tp->t_refcnt validity check back. There are still some race
Martin Blapp [Tue, 19 Dec 2006 16:46:13 +0000 (16:46 +0000)]
Add the tp->t_refcnt validity check back. There are still some race
conditions where tp->t_refcnt can go to zero.

17 years agoRemove unused sysctls.
David Xu [Tue, 19 Dec 2006 13:06:01 +0000 (13:06 +0000)]
Remove unused sysctls.

17 years agoUse pipe_direct_write() optimization only if the data is in process' memory.
Pawel Jakub Dawidek [Tue, 19 Dec 2006 12:52:22 +0000 (12:52 +0000)]
Use pipe_direct_write() optimization only if the data is in process' memory.
This fixes sending data through pipe from the kernel.

Fix suggested by: rwatson

17 years agoThe -r option can also cause the files to be extended (not
Ruslan Ermilov [Tue, 19 Dec 2006 11:20:34 +0000 (11:20 +0000)]
The -r option can also cause the files to be extended (not
necessarily truncated).

17 years agoRefine markup and add the EXIT STATUS section where appropriate.
Ruslan Ermilov [Tue, 19 Dec 2006 10:34:14 +0000 (10:34 +0000)]
Refine markup and add the EXIT STATUS section where appropriate.

17 years ago- Add missing callout_drain() call.
Oleg Bulyzhin [Tue, 19 Dec 2006 08:57:46 +0000 (08:57 +0000)]
- Add missing callout_drain() call.
- Synchronize bge_tick() with callout_reset/callout_stop() calls.
- Avoid using bge_tick() inside bge_link_upd(), use mii_pollstat() instead.

MFC after: 2 month

17 years ago- Fix autonegotiation timer.
Oleg Bulyzhin [Tue, 19 Dec 2006 08:41:48 +0000 (08:41 +0000)]
- Fix autonegotiation timer.
- Use MII_ANEGTICKS/MII_ANEGTICKS_GIGE defines instead of hardcoded value.
- Fix some comments.
- style(9)

MFC after: 2 month

17 years agoFor big-endian version of getulong() macro, cast result to u_int32_t.
Craig Rodrigues [Tue, 19 Dec 2006 02:31:58 +0000 (02:31 +0000)]
For big-endian version of getulong() macro, cast result to u_int32_t.
This macro was written expecting a 32-bit unsigned long, and
doesn't work properly on 64-bit systems.  This bug caused vn_stat()
to return incorrect values for files larger than 2gb on msdosfs filesystems
on 64-bit systems.

PR: 106703
Submitted by: Axel Gonzalez <loox e-shell net>
MFC after: 3 days

17 years agoFix get_ulong() macro on AMD64 (or any little-endian 64-bit platform).
Craig Rodrigues [Tue, 19 Dec 2006 01:55:45 +0000 (01:55 +0000)]
Fix get_ulong() macro on AMD64 (or any little-endian 64-bit platform).
This bug caused vn_stat() to fail on files larger than 2gb on msdosfs
filesystems on AMD64.

PR: 106703
Tested by: Axel Gonzalez <loox e-shell net>
MFC after: 3 days

17 years agoRestore revision 1.126 that got accidentally nuked.
Matt Jacob [Mon, 18 Dec 2006 23:53:39 +0000 (23:53 +0000)]
Restore revision 1.126 that got accidentally nuked.

17 years agoAdd back some CAM_NEW_TRAN code to make backporting to RELENG_6 easier.
Matt Jacob [Mon, 18 Dec 2006 23:50:30 +0000 (23:50 +0000)]
Add back some CAM_NEW_TRAN code to make backporting to RELENG_6 easier.

17 years agoRemove stall comment.
Takanori Watanabe [Mon, 18 Dec 2006 18:57:41 +0000 (18:57 +0000)]
Remove stall comment.

Pointed out by:Ed Maste.

17 years agoSpell "Kensington Thinking Mouse" correctly.
Giorgos Keramidas [Mon, 18 Dec 2006 18:48:28 +0000 (18:48 +0000)]
Spell "Kensington Thinking Mouse" correctly.

17 years agoClean bound and non-bound pthread structures consistently before
Peter Edwards [Mon, 18 Dec 2006 17:08:07 +0000 (17:08 +0000)]
Clean bound and non-bound pthread structures consistently before
they become candidates for reuse.  Without this fix, some of the
state from a thread structure's previous incarnation could interfere
with its new one. Specifically, a non-bound thread started as
"suspended" (see pthread_attr_setcreatesuspend_np()) might not get
scheduled at all when resumed, as the "active" flag would be set
spuriously.

Reviewed by: deischen@, davidxu@
MFC after: 1 week

17 years ago- Remove stale VPD support and its comment and get device name from VPD API.
Jung-uk Kim [Mon, 18 Dec 2006 16:40:04 +0000 (16:40 +0000)]
- Remove stale VPD support and its comment and get device name from VPD API.
- Do not repeatedly read vendor/device IDs while probing.
- Remove redundant bzero(3) for softc.  device_get_softc(9) does it for free[1].

Reviewed by: glebius
Suggested by: glebius[1]

17 years agoUse a standard section name.
Ruslan Ermilov [Mon, 18 Dec 2006 15:45:01 +0000 (15:45 +0000)]
Use a standard section name.

17 years agoImprove markup.
Ruslan Ermilov [Mon, 18 Dec 2006 15:36:31 +0000 (15:36 +0000)]
Improve markup.

17 years agoOops, I forget to add amd64 as ACPI arch.
Takanori Watanabe [Mon, 18 Dec 2006 14:38:31 +0000 (14:38 +0000)]
Oops, I forget to add amd64 as ACPI arch.

17 years agoFor consistency use 'unsigned' instead of 'u_int'.
Pawel Jakub Dawidek [Mon, 18 Dec 2006 11:12:00 +0000 (11:12 +0000)]
For consistency use 'unsigned' instead of 'u_int'.

17 years agoremove unneeded operations in tsb_set_tte_real - the function is
Kip Macy [Mon, 18 Dec 2006 07:46:59 +0000 (07:46 +0000)]
remove unneeded operations in tsb_set_tte_real - the function is
only used early in initialization so SMP safeness isn't really an
issue

17 years agoadd an interface for passing the entire kernel size up front to the
Kip Macy [Mon, 18 Dec 2006 07:35:14 +0000 (07:35 +0000)]
add an interface for passing the entire kernel size up front to the
loader so that it can memory can be allocated aligned at the beginning of
the desired large page

17 years agoadd new large page sizes for use by shared loader
Kip Macy [Mon, 18 Dec 2006 07:28:59 +0000 (07:28 +0000)]
add new large page sizes for use by shared loader

17 years agobuild new mdesc file
Kip Macy [Mon, 18 Dec 2006 07:26:35 +0000 (07:26 +0000)]
build new mdesc file

17 years agoadd declaration for new helper function
Kip Macy [Mon, 18 Dec 2006 07:25:26 +0000 (07:25 +0000)]
add declaration for new helper function