]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoInclude systat.h for cmdtab, to make gcc4 happy.
Xin LI [Mon, 2 May 2005 10:50:10 +0000 (10:50 +0000)]
Include systat.h for cmdtab, to make gcc4 happy.

19 years agoCleanups for gcc4:
Xin LI [Mon, 2 May 2005 10:31:27 +0000 (10:31 +0000)]
Cleanups for gcc4:

 - Use socklen_t for socket functions
 - Remove an unnecessary extern
 - Signed/unsigned cleanups

Reviewed by: harti

19 years agoConsistently use signed char so gcc4 won't complain
Xin LI [Mon, 2 May 2005 10:23:03 +0000 (10:23 +0000)]
Consistently use signed char so gcc4 won't complain

19 years agoWARNS=6 and gcc4 cleanup:
Xin LI [Mon, 2 May 2005 10:13:38 +0000 (10:13 +0000)]
WARNS=6 and gcc4 cleanup:
- Use const where necessary
- Use __unused where applicable
- Rename variables that is conflicit with global definations

19 years agoNeaten usage message.
Joseph Koshy [Mon, 2 May 2005 10:07:11 +0000 (10:07 +0000)]
Neaten usage message.

19 years agoinclude stdarg.h for va_list
Xin LI [Mon, 2 May 2005 10:04:16 +0000 (10:04 +0000)]
include stdarg.h for va_list

19 years agoMake gcc4 happy by making consistent signedness.
Xin LI [Mon, 2 May 2005 10:00:39 +0000 (10:00 +0000)]
Make gcc4 happy by making consistent signedness.

19 years agofix typo in previos commit.
Hajimu UMEMOTO [Mon, 2 May 2005 09:42:59 +0000 (09:42 +0000)]
fix typo in previos commit.

19 years agoteach IPv6 to async DNS resolver in ypserv:
Hajimu UMEMOTO [Mon, 2 May 2005 09:34:07 +0000 (09:34 +0000)]
teach IPv6 to async DNS resolver in ypserv:
- query AAAA RR on ipnodes.byname query.
- query .ip6.arpa. on ipnodes.byaddr query.

19 years agofixed a possible reference to a broken memory
SUZUKI Shinsuke [Mon, 2 May 2005 07:44:24 +0000 (07:44 +0000)]
fixed a possible reference to a broken memory

MFC after: 1 day

19 years agoAlways attach a subdisk even if no valid metadata found.
Søren Schmidt [Mon, 2 May 2005 07:06:50 +0000 (07:06 +0000)]
Always attach a subdisk even if no valid metadata found.
This allows the disks to be used later in a raid create.

19 years agoRemove GIANT_REQUIRED from vmspace_exec().
Alan Cox [Mon, 2 May 2005 07:05:20 +0000 (07:05 +0000)]
Remove GIANT_REQUIRED from vmspace_exec().

Prodded by: jeff

19 years agoNew release notes: musycc gone, new ndis(4) driver generation/load
Bruce A. Mah [Mon, 2 May 2005 05:56:30 +0000 (05:56 +0000)]
New release notes:  musycc gone, new ndis(4) driver generation/load
mechanism, bsnmp 1.9, libz 1.2.2.

19 years agothe 3rd argument of getsockname() should be socklen_t*.
Hajimu UMEMOTO [Mon, 2 May 2005 04:43:32 +0000 (04:43 +0000)]
the 3rd argument of getsockname() should be socklen_t*.

Submitted by: stefanf

19 years agoSince it is not possible for curthread to be NULL in this context,
Christian S.J. Peron [Mon, 2 May 2005 02:07:55 +0000 (02:07 +0000)]
Since it is not possible for curthread to be NULL in this context,
drop the check+initialization for a straight initialization. Also
assert that curthread will never be NULL just to be sure.

Discussed with: rwatson, peter
MFC after: 1 week

19 years agoNew release notes: IPFilter 4.1.8, tcsh 6.14.00.
Bruce A. Mah [Mon, 2 May 2005 01:14:36 +0000 (01:14 +0000)]
New release notes:  IPFilter 4.1.8, tcsh 6.14.00.

19 years agoFix a few grammos and style nits here and there.
Bruce A. Mah [Mon, 2 May 2005 01:09:32 +0000 (01:09 +0000)]
Fix a few grammos and style nits here and there.

19 years agoMake caesar and pom WARNS 6 clean. While there, use prototypes everywhere.
Stefan Farfeleder [Sun, 1 May 2005 19:34:22 +0000 (19:34 +0000)]
Make caesar and pom WARNS 6 clean.  While there, use prototypes everywhere.

PR: 62456

19 years agoRemove lame and ineffectual attempt to fix ipf breakage of tenderbox.
Warner Losh [Sun, 1 May 2005 17:36:09 +0000 (17:36 +0000)]
Remove lame and ineffectual attempt to fix ipf breakage of tenderbox.

19 years agoMFen 1.845-1.849
Xin LI [Sun, 1 May 2005 15:07:58 +0000 (15:07 +0000)]
MFen 1.845-1.849

Obtained from: The FreeBSD Simplified Chinese Project CVS

19 years agoAdd convenience APIs pmc_width() and pmc_capabilities() to -lpmc.
Joseph Koshy [Sun, 1 May 2005 14:11:49 +0000 (14:11 +0000)]
Add convenience APIs pmc_width() and pmc_capabilities() to -lpmc.
Have pmcstat(8) and pmccontrol(8) use these APIs.

Return PMC class-related constants (PMC widths and capabilities)
with the OP GETCPUINFO call leaving OP PMCINFO to return only the
dynamic information associated with a PMC (i.e., whether enabled,
owner pid, reload count etc.).

Allow pmc_read() (i.e., OPS PMCRW) on active self-attached PMCs to
get upto-date values from hardware since we can guarantee that the
hardware is running the correct PMC at the time of the call.

Bug fixes:
 - (x86 class processors) Fix a bug that prevented an RDPMC
   instruction from being recognized as permitted till after the
   attached process had context switched out and back in again after
   a pmc_start() call.

   Tighten the rules for using RDPMC class instructions: a GETMSR
   OP is now allowed only after an OP ATTACH has been done by the
   PMC's owner to itself.  OP GETMSR is not allowed for PMCs that
   track descendants, for PMCs attached to processes other than
   their owner processes.

 - (P4/HTT processors only) Fix a bug that caused the MI and MD
   layers to get out of sync.  Add a new MD operation 'get_config()'
   as part of this fix.

 - Allow multiple system-mode PMCs at the same row-index but on
   different CPUs to be allocated.

 - Reject allocation of an administratively disabled PMC.

Misc. code cleanups and refactoring.  Improve a few comments.

19 years agoRemove now unused inirw variable from previous use of COMMON_END().
Robert Watson [Sun, 1 May 2005 14:01:38 +0000 (14:01 +0000)]
Remove now unused inirw variable from previous use of COMMON_END().

Reported by: csjp

19 years agoCatch up with latest ATA newbus commits.
Peter Grehan [Sun, 1 May 2005 13:11:29 +0000 (13:11 +0000)]
Catch up with latest ATA newbus commits.

19 years agoFix typo in last commit.
Peter Grehan [Sun, 1 May 2005 13:06:05 +0000 (13:06 +0000)]
Fix typo in last commit.

Approved by: rwatson

19 years agooops, we don't need previous change.
Hajimu UMEMOTO [Sun, 1 May 2005 12:37:12 +0000 (12:37 +0000)]
oops, we don't need previous change.

19 years agoUpdate on the last commit, the dma* funciton needs to be called with
Søren Schmidt [Sun, 1 May 2005 12:24:45 +0000 (12:24 +0000)]
Update on the last commit, the dma* funciton needs to be called with
a channel device, not an ata device, or we'll be out of luck in
reset/timeout where we dont have a device.

19 years agomake it compilable without YP definition.
Hajimu UMEMOTO [Sun, 1 May 2005 12:08:57 +0000 (12:08 +0000)]
make it compilable without YP definition.

19 years ago - All buffers should either be clean or dirty. If neither of these flags
Jeff Roberson [Sun, 1 May 2005 12:00:36 +0000 (12:00 +0000)]
 - All buffers should either be clean or dirty.  If neither of these flags
   are set when we attempt to remove a buffer from a queue we should panic.
   Hopefully this will catch the source of the wrong bufobj panics.

Sponsored by: Isilon Systems, Inc.

19 years agoSlide unlocking of the tcbinfo lock earlier in tcp_usr_send(), as it's
Robert Watson [Sun, 1 May 2005 11:11:38 +0000 (11:11 +0000)]
Slide unlocking of the tcbinfo lock earlier in tcp_usr_send(), as it's
needed only for implicit connect cases.  Under load, especially on SMP,
this can greatly reduce contention on the tcbinfo lock.

NB: Ambiguities about the state of so_pcb need to be resolved so that
all use of the tcbinfo lock in non-implicit connection cases can be
eliminated.

Submited by: Kazuaki Oda <kaakun at highway dot ne dot jp>

19 years ago- Merged some missing changes from fdisk/fdisk.c.
Yoshihiro Takahashi [Sun, 1 May 2005 10:08:35 +0000 (10:08 +0000)]
- Merged some missing changes from fdisk/fdisk.c.
- Remove ifdef PC98.
- Make WARN=4 clean.

19 years agocosmetic changes.
Yoshihiro Takahashi [Sun, 1 May 2005 09:50:02 +0000 (09:50 +0000)]
cosmetic changes.

19 years agoAdd some convenience defines.
Yoshihiro Takahashi [Sun, 1 May 2005 09:48:24 +0000 (09:48 +0000)]
Add some convenience defines.

19 years agoFix signed vs unsigned warning.
Yoshihiro Takahashi [Sun, 1 May 2005 09:44:50 +0000 (09:44 +0000)]
Fix signed vs unsigned warning.

19 years agoGo back to the old way of finding the Promise metadata, the new way was
Søren Schmidt [Sun, 1 May 2005 08:45:12 +0000 (08:45 +0000)]
Go back to the old way of finding the Promise metadata, the new way was
too simple causing older controllers metadata to get lost.

19 years ago - document powerd_flags too
Alexander Leidinger [Sun, 1 May 2005 08:34:01 +0000 (08:34 +0000)]
 - document powerd_flags too
 - add a reference to powerd(8)

Approved by: mentor (joerg)

19 years agoAdd ntp.jst.mfeed.ad.jp and ntp1.v6.mfeed.ad.jp (IPv6 only) to
Makoto Matsushita [Sun, 1 May 2005 08:19:54 +0000 (08:19 +0000)]
Add ntp.jst.mfeed.ad.jp and ntp1.v6.mfeed.ad.jp (IPv6 only) to
the NTP server list.

The servers are by Internet Multifeed Co., and connected to their
IX service JPNAP.  It would be a good NTP server for machines in Japan.

See also http://www.jst.mfeed.ad.jp/ for more details (Japanese only).

Reviewed by: nork

19 years agodon't see RES_USE_INET6 when called from getipnodeby*().
Hajimu UMEMOTO [Sun, 1 May 2005 07:39:45 +0000 (07:39 +0000)]
don't see RES_USE_INET6 when called from getipnodeby*().

19 years agoMerged more changes from fdisk/fdisk.c.
Yoshihiro Takahashi [Sun, 1 May 2005 05:04:05 +0000 (05:04 +0000)]
Merged more changes from fdisk/fdisk.c.

19 years agoUse NDOSPART macro instead of a magic number.
Yoshihiro Takahashi [Sun, 1 May 2005 05:01:55 +0000 (05:01 +0000)]
Use NDOSPART macro instead of a magic number.

19 years agoMerged from src/sbin/fdisk/fdisk.c revision 1.77.
Yoshihiro Takahashi [Sun, 1 May 2005 04:14:01 +0000 (04:14 +0000)]
Merged from src/sbin/fdisk/fdisk.c revision 1.77.

19 years agoMFi386: revision 1.1198 (add KDB_STOP_NMI option).
Yoshihiro Takahashi [Sun, 1 May 2005 04:00:12 +0000 (04:00 +0000)]
MFi386: revision 1.1198 (add KDB_STOP_NMI option).

19 years agoMFi386: revision 1.220 (add KDB_STOP_NMI option).
Yoshihiro Takahashi [Sun, 1 May 2005 03:59:25 +0000 (03:59 +0000)]
MFi386: revision 1.220 (add KDB_STOP_NMI option).

19 years agoupdate for recent api changes
Sam Leffler [Sun, 1 May 2005 02:12:44 +0000 (02:12 +0000)]
update for recent api changes

19 years agoTurn on libobjc.so for all platforms on -current. It didn't work back
Peter Wemm [Sun, 1 May 2005 01:59:48 +0000 (01:59 +0000)]
Turn on libobjc.so for all platforms on -current.  It didn't work back
in 1996.

19 years ago - Remove spls and comments relating to them.
Jeff Roberson [Sun, 1 May 2005 01:01:17 +0000 (01:01 +0000)]
 - Remove spls and comments relating to them.

19 years ago - Remove an old splcam hack.
Jeff Roberson [Sun, 1 May 2005 00:59:55 +0000 (00:59 +0000)]
 - Remove an old splcam hack.

19 years ago - Remove unnecessary spls.
Jeff Roberson [Sun, 1 May 2005 00:59:34 +0000 (00:59 +0000)]
 - Remove unnecessary spls.

19 years ago - Return EACCES if we're trying to exec on a vp with no object.
Jeff Roberson [Sun, 1 May 2005 00:58:19 +0000 (00:58 +0000)]
 - Return EACCES if we're trying to exec on a vp with no object.

Errno supplied by: cperciva

19 years ago - In devfs_open() and devfs_close() grab Giant if the driver sets NEEDGIANT.
Jeff Roberson [Sun, 1 May 2005 00:56:34 +0000 (00:56 +0000)]
 - In devfs_open() and devfs_close() grab Giant if the driver sets NEEDGIANT.
   We still have to DROP_GIANT and PICKUP_GIANT when NEEDGIANT is not set
   because vfs is still sometime entered with Giant held.

19 years agoo enable shutdown of taskqueue threads; the thread servicing the queue checks
Sam Leffler [Sun, 1 May 2005 00:38:11 +0000 (00:38 +0000)]
o enable shutdown of taskqueue threads; the thread servicing the queue checks
  a new entry in the taskqueue struct each time it wakes up to see if it
  should terminate
o adjust TASKQUEUE_DEFINE_THREAD & co. to record the thread/proc identity for
  the shutdown rendezvous
o replace wakeup after adding a task to a queue with wakeup_one; this helps
  queues where multiple threads are used to service tasks (e.g. acpi)
o remove NULL check of tq_enqueue method; it should never be NULL

Reviewed by: dfr, njl

19 years agooops, gethostbyaddr(3) must return h_addr as an IPv4-mapped
Hajimu UMEMOTO [Sat, 30 Apr 2005 20:07:01 +0000 (20:07 +0000)]
oops, gethostbyaddr(3) must return h_addr as an IPv4-mapped
IPv6 address when RES_USE_INET6 was set, according to RFC 2133
section 6.2.

19 years agoImplement an alternate method to stop CPUs when entering DDB. Normally we use
Doug White [Sat, 30 Apr 2005 20:01:00 +0000 (20:01 +0000)]
Implement an alternate method to stop CPUs when entering DDB. Normally we use
a regular IPI vector, but this vector is blocked when interrupts are disabled.
With "options KDB_STOP_NMI" and debug.kdb.stop_cpus_with_nmi set, KDB will
send an NMI to each CPU instead. The code also has a context-stuffing
feature which helps ddb extract the state of processes running on the
stopped CPUs.

KDB_STOP_NMI is only useful with SMP and complains if SMP is not defined.
This feature only applies to i386 and amd64 at the moment, but could be
used on other architectures with the appropriate MD bits.

Submitted by: ups

19 years agohandling RES_USE_INET6 better.
Hajimu UMEMOTO [Sat, 30 Apr 2005 19:28:31 +0000 (19:28 +0000)]
handling RES_USE_INET6 better.

19 years ago_ht_gethostbyaddr didn't handle RES_USE_INET6 correctly.
Hajimu UMEMOTO [Sat, 30 Apr 2005 18:46:27 +0000 (18:46 +0000)]
_ht_gethostbyaddr didn't handle RES_USE_INET6 correctly.

19 years ago_gethostbynisname() didn't support RES_USE_INET6.
Hajimu UMEMOTO [Sat, 30 Apr 2005 18:37:35 +0000 (18:37 +0000)]
_gethostbynisname() didn't support RES_USE_INET6.

19 years agoReduce diffs against actual production configuration.
Dag-Erling Smørgrav [Sat, 30 Apr 2005 18:26:05 +0000 (18:26 +0000)]
Reduce diffs against actual production configuration.

19 years agoRemove a line which merely echoed the builtin default.
Dag-Erling Smørgrav [Sat, 30 Apr 2005 18:25:01 +0000 (18:25 +0000)]
Remove a line which merely echoed the builtin default.

19 years agoSplit the 'clean' command into 'cleansrc', 'cleanobj' and 'cleanroot'.
Dag-Erling Smørgrav [Sat, 30 Apr 2005 18:22:12 +0000 (18:22 +0000)]
Split the 'clean' command into 'cleansrc', 'cleanobj' and 'cleanroot'.
Specifying 'clean' is equivalent to specifying all three.

19 years agoAdd 'cleanobj' to the default target list.
Dag-Erling Smørgrav [Sat, 30 Apr 2005 17:57:11 +0000 (17:57 +0000)]
Add 'cleanobj' to the default target list.

19 years ago_ht_gethostbyname didn't handle RES_USE_INET6 correctly.
Hajimu UMEMOTO [Sat, 30 Apr 2005 17:01:18 +0000 (17:01 +0000)]
_ht_gethostbyname didn't handle RES_USE_INET6 correctly.

19 years agoFix an old pasto.
Dag-Erling Smørgrav [Sat, 30 Apr 2005 16:27:20 +0000 (16:27 +0000)]
Fix an old pasto.

19 years agoTake newbusification one step further, ie use the device_t more consequently
Søren Schmidt [Sat, 30 Apr 2005 16:22:07 +0000 (16:22 +0000)]
Take newbusification one step further, ie use the device_t more consequently
all way through the code down the layers, instead of the mix'n'match that
resulted from the conversion done earlier.

Sponsored by: pair.com

19 years agoAdd some subversive code to cleandir: remove the sbin/ipf tree
Warner Losh [Sat, 30 Apr 2005 16:09:23 +0000 (16:09 +0000)]
Add some subversive code to cleandir: remove the sbin/ipf tree
entirely to fix the tinderbox machine.  This change will be reverted
once the tinderboxen are tindering again.

19 years ago - Remove long dead splbio() calls and comments relating to the old
Jeff Roberson [Sat, 30 Apr 2005 12:18:50 +0000 (12:18 +0000)]
 - Remove long dead splbio() calls and comments relating to the old
   synchronization mechanism.

19 years ago - Don't acquire Giant before calling b_biodone, individual consumers are
Jeff Roberson [Sat, 30 Apr 2005 11:44:22 +0000 (11:44 +0000)]
 - Don't acquire Giant before calling b_biodone, individual consumers are
   now required to do so themselves.

Sponsored by: Isilon Systems, Inc.

19 years ago - Use bdone() directly instead of calling it indirectly through
Jeff Roberson [Sat, 30 Apr 2005 11:28:19 +0000 (11:28 +0000)]
 - Use bdone() directly instead of calling it indirectly through
   ffs_rawreaddone().

Sponsored by: Isilon Systems, Inc.

19 years ago - Acquire Giant in AIO's iodone routine. VFS will no longer do it for us
Jeff Roberson [Sat, 30 Apr 2005 11:27:31 +0000 (11:27 +0000)]
 - Acquire Giant in AIO's iodone routine.  VFS will no longer do it for us
   soon.

Sponsored by: Isilon Systems, Inc.

19 years ago - Call VM_LOCK_GIANT in cluster_callback() to protect some pmap calls. VFS
Jeff Roberson [Sat, 30 Apr 2005 11:26:58 +0000 (11:26 +0000)]
 - Call VM_LOCK_GIANT in cluster_callback() to protect some pmap calls.  VFS
   will not be acquiring Giant before calling this function anymore.

Sponsored by: Isilon Systems, Inc.

19 years ago - VM_LOCK_GIANT in the swap pager's iodone routine as VFS will soon call it
Jeff Roberson [Sat, 30 Apr 2005 11:25:49 +0000 (11:25 +0000)]
 - VM_LOCK_GIANT in the swap pager's iodone routine as VFS will soon call it
   without Giant.

Sponsored by: Isilon Systems, Inc.

19 years ago - Mark devfs as MNTK_MPSAFE as I belive it does not require Giant.
Jeff Roberson [Sat, 30 Apr 2005 11:24:17 +0000 (11:24 +0000)]
 - Mark devfs as MNTK_MPSAFE as I belive it does not require Giant.

Sponsored by: Isilon Systems, Inc.
Agreed in principle by: phk

19 years ago - In vnlru_free() remove the vnode from the free list before we call
Jeff Roberson [Sat, 30 Apr 2005 11:22:40 +0000 (11:22 +0000)]
 - In vnlru_free() remove the vnode from the free list before we call
   vtryrecycle().  We could sometimes get into situations where two threads
   could try to recycle the same vnode before this.
 - vtryrecycle() is now responsible for returning the vnode to the free list
   if it fails and someone else hasn't done it.
 - Make a new function vfreehead() which moves a vnode to the head of the
   free list and use it in vgone() to clean up that code a bit.

Sponsored by: Isilon Systems, Inc.
Reported by: pho, kkenn

19 years agoFix spelling errors.
Joel Dahl [Sat, 30 Apr 2005 09:26:22 +0000 (09:26 +0000)]
Fix spelling errors.

Approved by: brueffer (mentor)

19 years agoDo not unconditionally mount devfs to ${jail_devdir}/dev. First check
Christian S.J. Peron [Sat, 30 Apr 2005 00:16:00 +0000 (00:16 +0000)]
Do not unconditionally mount devfs to ${jail_devdir}/dev. First check
to see if a prior devfs has been mounted. If no devfs is mounted on
${jail_devdir}/dev then proceed. This will prevent the stack up of
multiple devfs mounts on the same mount point.

Discussed with: pjd
MFC after: 1 week

19 years agoTo allow /etc to be as minimal as possible in a diskless setup, we need
Brooks Davis [Fri, 29 Apr 2005 23:02:56 +0000 (23:02 +0000)]
To allow /etc to be as minimal as possible in a diskless setup, we need
to run initdiskless before we run rcorder on /etc/rc.d.  To allow this,
move /etc/rc.d/initdiskless to /etc/rc.initdiskless and run it directly
from /etc/rc.

Remove /etc/rc.d/preseedrandom as it is no longer necessicary (we start
with entropy unblocked) and was only used by initdiskless when it
was needed.

Discussed on: freebsd-rc
Repocopy by: peter

19 years agoAdd PROC_UNLOCK(p) to PTRACESTOP_SC(p, td, flag). This is necessary
Colin Percival [Fri, 29 Apr 2005 22:38:02 +0000 (22:38 +0000)]
Add PROC_UNLOCK(p) to PTRACESTOP_SC(p, td, flag).  This is necessary
due to a change made in revision 1.284 of sys/kern/kern_sig.c in August
2004 which made ptracestop() return with the process still locked.

Submitted by: Mauritz Sundell
MFC After: 3 days

19 years agoIf we resort to opening a slice instead of the base device, do so with
Poul-Henning Kamp [Fri, 29 Apr 2005 21:24:27 +0000 (21:24 +0000)]
If we resort to opening a slice instead of the base device, do so with
the correct mode for our arguments.

19 years agoMake call to tw_cl_deferred_interrupt in twa_poll, not dependent on the
Vinod Kashyap [Fri, 29 Apr 2005 20:03:20 +0000 (20:03 +0000)]
Make call to tw_cl_deferred_interrupt in twa_poll, not dependent on the
return value from tw_cl_interrupt.

19 years ago- do validation check and IPv4-mapped IPv6 address handling before
Hajimu UMEMOTO [Fri, 29 Apr 2005 19:55:23 +0000 (19:55 +0000)]
- do validation check and IPv4-mapped IPv6 address handling before
  any query.
- don't query against IPv6 link-local address.
- use IN6_IS_ADDR_V4{MAPPED,COMPAT} macros.
- use memcpy() instead of bcopy().

Inspired by: NetBSD

19 years agoModify UMA to use critical sections to protect per-CPU caches, rather than
Robert Watson [Fri, 29 Apr 2005 18:56:36 +0000 (18:56 +0000)]
Modify UMA to use critical sections to protect per-CPU caches, rather than
mutexes, which offers lower overhead on both UP and SMP.  When allocating
from or freeing to the per-cpu cache, without INVARIANTS enabled, we now
no longer perform any mutex operations, which offers a 1%-3% performance
improvement in a variety of micro-benchmarks.  We rely on critical
sections to prevent (a) preemption resulting in reentrant access to UMA on
a single CPU, and (b) migration of the thread during access.  In the event
we need to go back to the zone for a new bucket, we release the critical
section to acquire the global zone mutex, and must re-acquire the critical
section and re-evaluate which cache we are accessing in case migration has
occured, or circumstances have changed in the current cache.

Per-CPU cache statistics are now gathered lock-free by the sysctl, which
can result in small races in statistics reporting for caches.

Reviewed by: bmilekic, jeff (somewhat)
Tested by: rwatson, kris, gnn, scottl, mike at sentex dot net, others

19 years agoEnable dc(4) and rl(4) in the PAE kernel.
David E. O'Brien [Fri, 29 Apr 2005 18:12:22 +0000 (18:12 +0000)]
Enable dc(4) and rl(4) in the PAE kernel.
Both have been busdma'ed for use and tested in the Sparc64 kenrel.

19 years agoNETDB_INTERNAL is not fit, here. return NO_RECOVERY for h_errno.
Hajimu UMEMOTO [Fri, 29 Apr 2005 17:36:54 +0000 (17:36 +0000)]
NETDB_INTERNAL is not fit, here.  return NO_RECOVERY for h_errno.

19 years agoRename the directive_hash.[ch] files to hash_tables.[ch] now
Hartmut Brandt [Fri, 29 Apr 2005 15:15:28 +0000 (15:15 +0000)]
Rename the directive_hash.[ch] files to hash_tables.[ch] now
that there are more than one hash table in them. There is no
history to preserve here, so go without a repo-copy.

Asked for by: Max Okumoto <okumoto@ucsd.edu>

19 years agoRemove the Netgear WG311 from the list. The v1 card is supported, but
Brooks Davis [Fri, 29 Apr 2005 15:07:44 +0000 (15:07 +0000)]
Remove the Netgear WG311 from the list.  The v1 card is supported, but
the v2 card is a TI.  Since we're not attempting to keep this list
complete, removing this is best.

Reported by: Brian Candler <B dot Candler at pobox dot com>
Evan Dower <evantd at hotmail dot com>
MFC After: 1 day

19 years agoToggle on warnings. This resolves the problem with building old releases
Hartmut Brandt [Fri, 29 Apr 2005 14:44:10 +0000 (14:44 +0000)]
Toggle on warnings. This resolves the problem with building old releases
(getting zillions of warnings). Building an old release uses that release's
sys.mk which does not switch on these warnings, so make will be silent.
They can be switch on on the command line with the -x option to make.
This has been tested by building RELENG_5_4 on CURRENT.

19 years agoour get{addr,name}info() is considered thread-safe.
Hajimu UMEMOTO [Fri, 29 Apr 2005 14:37:51 +0000 (14:37 +0000)]
our get{addr,name}info() is considered thread-safe.

19 years agoImplement a pseudo-target .WARN that allows toggeling the warning flags
Hartmut Brandt [Fri, 29 Apr 2005 14:37:44 +0000 (14:37 +0000)]
Implement a pseudo-target .WARN that allows toggeling the warning flags
for the current make. This does not override flags specified on the
command line and these settings are not passed to sub-makes.

19 years agowe cannot use inet_ntoa(3), here. so, use inet_ntop(3), instead.
Hajimu UMEMOTO [Fri, 29 Apr 2005 12:01:36 +0000 (12:01 +0000)]
we cannot use inet_ntoa(3), here.  so, use inet_ntop(3), instead.

19 years agoNow that probing is working in the new fashion, we need to go back to
Søren Schmidt [Fri, 29 Apr 2005 11:30:03 +0000 (11:30 +0000)]
Now that probing is working in the new fashion, we need to go back to
having ata_getparm issue an ata_request and not fool around with the HW
on its own.
Needed for new HW support.

19 years agoo Typo: s/teminal/terminal/.
Maxim Konovalov [Fri, 29 Apr 2005 11:10:27 +0000 (11:10 +0000)]
o Typo: s/teminal/terminal/.

Submitted by: Michal Varga

19 years agoThis commit was generated by cvs2svn to compensate for changes in r145673,
Hartmut Brandt [Fri, 29 Apr 2005 11:02:18 +0000 (11:02 +0000)]
This commit was generated by cvs2svn to compensate for changes in r145673,
which included commits to RCS files with non-trunk default branches.

19 years agoVendor patch: fix a bug that was introduced when moving from libisc to
Hartmut Brandt [Fri, 29 Apr 2005 11:02:18 +0000 (11:02 +0000)]
Vendor patch: fix a bug that was introduced when moving from libisc to
libbegemot: the rpoll_start_timer function needs the timeout value in
milli-seconds, not the absolute time when the timer should tick.

19 years agoAdd flag to choose whether to use getgrouplist(3) or getgroups(2)
Robert Drehmel [Fri, 29 Apr 2005 10:11:18 +0000 (10:11 +0000)]
Add flag to choose whether to use getgrouplist(3) or getgroups(2)
to the id_print() function.

Use getgrouplist(3) for the case when an user was specified,
and getgroups(2) when no user was given.
That reverts to the expected behaviour and makes it easy to
implement an option later to force using getgrouplist(3).

19 years agoFix spelling error.
Robert Drehmel [Fri, 29 Apr 2005 08:37:52 +0000 (08:37 +0000)]
Fix spelling error.

19 years agoRemove trailing spaces.
Pawel Jakub Dawidek [Fri, 29 Apr 2005 07:57:50 +0000 (07:57 +0000)]
Remove trailing spaces.

19 years agoAdd a detach for pci bridge and pci bus drivers. This allows one to
Warner Losh [Fri, 29 Apr 2005 06:22:41 +0000 (06:22 +0000)]
Add a detach for pci bridge and pci bus drivers.  This allows one to
theoretically unload pci bridges or pci drivers.  It will also allow
detach to work if one needed to detach a subtree.

This is inspired by looking at the p4 commits from bms to his 5.4
tree, but I didn't look at the final results.

19 years agoFix the following warnings on amd64:
Ruslan Ermilov [Fri, 29 Apr 2005 05:57:17 +0000 (05:57 +0000)]
Fix the following warnings on amd64:

/usr/src/sbin/ipf/ipftest/../../../sys/contrib/ipfilter/netinet/ip_frag.c: In function `fr_ipid_newfrag':
/usr/src/sbin/ipf/ipftest/../../../sys/contrib/ipfilter/netinet/ip_frag.c:397: warning: cast to pointer from integer of different size
/usr/src/sbin/ipf/ipftest/../../../sys/contrib/ipfilter/netinet/ip_frag.c: In function `fr_ipid_knownfrag':
/usr/src/sbin/ipf/ipftest/../../../sys/contrib/ipfilter/netinet/ip_frag.c:582: warning: cast from pointer to integer of different size

19 years agoOnly create the rdpti alias if the asr device creation succeeds.
Scott Long [Fri, 29 Apr 2005 04:47:11 +0000 (04:47 +0000)]
Only create the rdpti alias if the asr device creation succeeds.

19 years agoFor whatever reason, we don't allow USB on PAE. Since it's a dependency
Scott Long [Fri, 29 Apr 2005 02:59:19 +0000 (02:59 +0000)]
For whatever reason, we don't allow USB on PAE.  Since it's a dependency
for EHCI, exclude that driver also.

19 years agoDon't bother pretending that CAM will send CAM_DATA_PHYS pointers. It's
Scott Long [Fri, 29 Apr 2005 02:58:23 +0000 (02:58 +0000)]
Don't bother pretending that CAM will send CAM_DATA_PHYS pointers.  It's
a concept that is fundamentally broken with PAE.

19 years agoUpdate the file.* entries for the new home of hwpmc
Scott Long [Fri, 29 Apr 2005 02:40:16 +0000 (02:40 +0000)]
Update the file.* entries for the new home of hwpmc