]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agostyle.Makefile(5)
obrien [Fri, 12 Oct 2007 18:03:00 +0000 (18:03 +0000)]
style.Makefile(5)

16 years agoWelcome FreeBSD 8.
obrien [Fri, 12 Oct 2007 17:49:12 +0000 (17:49 +0000)]
Welcome FreeBSD 8.

16 years agoDo not attempt to make an NFS rpc call if using tftp
ps [Fri, 12 Oct 2007 17:09:43 +0000 (17:09 +0000)]
Do not attempt to make an NFS rpc call if using tftp

PR: kern/91720
Submitted by: Ruben Kerkhof

16 years agoFix a mistake made in the MPSAFE commit that caused CAM to serialize requests
scottl [Fri, 12 Oct 2007 17:03:41 +0000 (17:03 +0000)]
Fix a mistake made in the MPSAFE commit that caused CAM to serialize requests
to the controller.

16 years agoFor some blasted reason the SATA WC change frees a structure right in the
scottl [Fri, 12 Oct 2007 16:54:51 +0000 (16:54 +0000)]
For some blasted reason the SATA WC change frees a structure right in the
middle of using it.

16 years agoPlay better with CAM so that there aren't so many spurious warnings at boot.
scottl [Fri, 12 Oct 2007 16:52:55 +0000 (16:52 +0000)]
Play better with CAM so that there aren't so many spurious warnings at boot.

16 years agoDeferred interrupts don't work yet, disable them.
scottl [Fri, 12 Oct 2007 16:51:22 +0000 (16:51 +0000)]
Deferred interrupts don't work yet, disable them.

Submitted by: Manjunath Ranganathaiah

16 years agoTeach /etc/rc.d/ppp to start multiple instances of ppp.
emax [Fri, 12 Oct 2007 16:35:36 +0000 (16:35 +0000)]
Teach /etc/rc.d/ppp to start multiple instances of ppp.

ppp_profile variable can now contain multiple profiles.
Overrides for ppp mode and nat can go into ppp_$profile_mode
and ppp_$profile_nat variables respectively. If those are
not specified, defaults from ppp_mode and ppp_nat are used.

Submitted by: Yuri Kurenkov < y dot kurenkov at init dot ru >
Reviewed by: mtm
MFC after: 1 week

16 years agoBump version numbers to reflect that HEAD represents 8.0-CURRENT.
bmah [Fri, 12 Oct 2007 15:35:56 +0000 (15:35 +0000)]
Bump version numbers to reflect that HEAD represents 8.0-CURRENT.

16 years ago- Change the wakeup logic associated with having multiple sleepers
csjp [Fri, 12 Oct 2007 15:09:02 +0000 (15:09 +0000)]
- Change the wakeup logic associated with having multiple sleepers
  on multiple different audit pipes.  The old method used cv_signal()
  which would result in only one thread being woken up after we
  appended a record to it's queue.  This resulted in un-timely wake-ups
  when processing audit records real-time.

- Assign PSOCK priority to threads that have been sleeping on a read(2).
  This is the same priority threads are woken up with when they select(2)
  or poll(2).  This yields fairness between various forms of sleep on
  the audit pipes.

Obtained from: TrustedBSD Project
Discussed with: rwatson
MFC after: 1 week

16 years agoMake sure that we refresh the PID on read(2) and write(2) operations.
csjp [Fri, 12 Oct 2007 14:58:34 +0000 (14:58 +0000)]
Make sure that we refresh the PID on read(2) and write(2) operations.
This fixes the process portion of the bpf(4) stats if the peer forks
into the background after it's opened the descriptor.  This bug
results in the following behavior for netstat -B:

# netstat -B
  Pid  Netif  Flags      Recv      Drop     Match Sblen Hblen Command
netstat: kern.proc.pid failed: No such process
78023    em0 p--s--   2237404     43119   2237404 13986     0 ??????

MFC after: 1 week

16 years agoAdd a signal handler for SIGINT to make sure that the PID file
csjp [Fri, 12 Oct 2007 14:57:39 +0000 (14:57 +0000)]
Add a signal handler for SIGINT to make sure that the PID file
gets cleaned up upon receiving SIGINT.  This un-breaks subsequent
executions of ipfwpcap and helps when debugging network/divert
issues like this:

ipfwpcap -r 6000 - | tcpdump -r -

MFC after: 1 week

16 years agoRevision 1.12 of lockf.c fixed a "thundering herd" scenario when the
csjp [Fri, 12 Oct 2007 14:56:52 +0000 (14:56 +0000)]
Revision 1.12 of lockf.c fixed a "thundering herd" scenario when the
lock experienced contention a number of processes would race to acquire
lock when it was released.  This problem resulted in a lot of CPU
load as well as locks being picked up out of order.

Unfortunately, a regression snuck in which allowed multiple threads
to pickup the same lock when -k was not used.  This could occur when
multiple processes open a file descriptor to inode X (one process
will be blocked) and the file is unlinked on unlock (thereby removing
the directory entry allow another process to create a new directory
entry for the same file name and lock it).

This changes restores the old algorithm of: wait for the lock, then
acquire lock when we want to unlink the file on exit (specifically
when -k is not used) and keeps the new algorithm for when -k is used,
which yields fairness and improved performance.

Also, update the man page to inform users that if lockf(1) is being
used to facilitate concurrency between a number of processes, it
is recommended that -k be used to reduce CPU load and yeld
fairness with regard to lock ordering.

Collaborated with: jdp
PR: bin/114341
PR: bin/116543
PR: bin/111101
MFC after: 1 week

16 years agoAdd pts/pty to the un-hidden devices for logins. This un-breaks
csjp [Fri, 12 Oct 2007 14:55:41 +0000 (14:55 +0000)]
Add pts/pty to the un-hidden devices for logins.  This un-breaks
logins to jailed environments when the system is using PTS style
ptys (kern.pts.enable=1).

Discussed with: rwatson
MFc after: 1 week

16 years agoWhen pidfile is already locked and has zero length, do not return
kib [Fri, 12 Oct 2007 10:38:05 +0000 (10:38 +0000)]
When pidfile is already locked and has zero length, do not return
success and zero pid from pidfile_read(). Return EAGAIN instead. Sleep
up to three times for 5 ms while waiting for pidfile to be written.

mount(8) does the kill(mountpid, SIGHUP). If mountd pidfile is truncated,
that would result in the SIGHUP delivered to the mount' process group
instead of the mountd.

Found and analyzed by: Peter Holm
Tested by: Peter Holm, kris
Reviewed by: pjd
MFC after: 1 week

16 years agoStyle (whitespace-only) change.
ru [Fri, 12 Oct 2007 10:35:24 +0000 (10:35 +0000)]
Style (whitespace-only) change.

16 years agoRemove unneeded debug printf that is broken on 64bit arches.
thompsa [Fri, 12 Oct 2007 10:00:26 +0000 (10:00 +0000)]
Remove unneeded debug printf that is broken on 64bit arches.

16 years agoAdd another zydas device: Planex GW-US54GD (a.k.a Dempaotoko)
takawata [Fri, 12 Oct 2007 08:45:55 +0000 (08:45 +0000)]
Add another zydas device: Planex GW-US54GD (a.k.a Dempaotoko)

16 years agoBring in some new Sierra Wireless drivers running under pccard.
remko [Fri, 12 Oct 2007 08:26:15 +0000 (08:26 +0000)]
Bring in some new Sierra Wireless drivers running under pccard.

(AC710/AC750/AC850/A550/A555/A710/A750)

PR: kern/110190
Submitted by: Pascal Vizeli<pvizeli at yahoo dot de>
Approved by: imp (mentor)

MFC After: 1 week

16 years agoThough it was possible to configure our BIND to build even when
ru [Fri, 12 Oct 2007 08:03:51 +0000 (08:03 +0000)]
Though it was possible to configure our BIND to build even when
libpthread support isn't present, our maintainer felt it's an
overkill, so instead enforce the BIND dependency on libpthread.

16 years agoRemove default (compressed) manpages.
ru [Fri, 12 Oct 2007 07:58:08 +0000 (07:58 +0000)]
Remove default (compressed) manpages.

16 years agoSpelling fix for interupt -> interrupt
kevlo [Fri, 12 Oct 2007 06:03:46 +0000 (06:03 +0000)]
Spelling fix for interupt -> interrupt

16 years agoUpdate ipw to work with the new net80211 stack, plus other driver improvements.
thompsa [Fri, 12 Oct 2007 05:23:00 +0000 (05:23 +0000)]
Update ipw to work with the new net80211 stack, plus other driver improvements.

 - Add proper scanning support rather than letting the firmware grab the first
   access point
 - Overhaul state changes
 - Use macros for locking and provide _locked() versions of some functions
 - Increase debugging output
 - Use a callout rather than the old watchdog interface
 - Improve style, function names and defines
 - Add WPA (TKIP) support

Based heavily on a patchset provided by Sam Leffler.

16 years agoProtect struct seq with mutex.
mav [Fri, 12 Oct 2007 04:56:26 +0000 (04:56 +0000)]
Protect struct seq with mutex.

Approved by: glebius (mentor)

16 years agoRemove one unneded assertion. It is also checked in
mav [Fri, 12 Oct 2007 04:54:43 +0000 (04:54 +0000)]
Remove one unneded assertion. It is also checked in
ng_l2tp_seq_check().

Approved by: glebius (mentor)

16 years agoReplace single rcvdata with 3 distinct to simplify code and
mav [Fri, 12 Oct 2007 04:53:23 +0000 (04:53 +0000)]
Replace single rcvdata with 3 distinct to simplify code and
reduce stack usage.

Approved by: mentor (glebius)

16 years agoRemove duplicate variables.
mav [Fri, 12 Oct 2007 04:51:30 +0000 (04:51 +0000)]
Remove duplicate variables.

16 years agoNote getfacl(1) changes
kevlo [Fri, 12 Oct 2007 04:48:58 +0000 (04:48 +0000)]
Note getfacl(1) changes

Submitted by: Henrik Brix Andersen <henrik at brixandersen.dk>

16 years agoCorrect the cpio writers to not accept data for non-regular files.
kientzle [Fri, 12 Oct 2007 04:11:31 +0000 (04:11 +0000)]
Correct the cpio writers to not accept data for non-regular files.
In particular, the previous code led to archives that had
non-empty bodies following directory entries.  Not a fatal
problem, as bsdtar and GNU cpio are both happy to just skip
this bogus data, but it still shouldn't be there.

MFC after: 3 days

16 years agoCorrect the return values of the final zero-length block at EOF.
kientzle [Fri, 12 Oct 2007 04:08:28 +0000 (04:08 +0000)]
Correct the return values of the final zero-length block at EOF.
Return EOF immediately if an entry in a ZIP archive has no body.
In particular, the latter issue was causing bsdtar to emit spurious
warnings when extracting directory entries from ZIP archives.

MFC after: 3 days

16 years agoSet the NFS server sockbuf high watermarks to the system defaults
mohans [Fri, 12 Oct 2007 03:56:27 +0000 (03:56 +0000)]
Set the NFS server sockbuf high watermarks to the system defaults
(up form 32KB). The low highwatermark setting caused UDP fullsock
request drops, throttling thruput greatly.
Reported by: Kris Kennaway
Approved by: re@ (Ken Smith)

16 years agoNot all VIA Rhine chips support 256 register space. So touching
yongari [Fri, 12 Oct 2007 03:32:55 +0000 (03:32 +0000)]
Not all VIA Rhine chips support 256 register space. So touching
VR_STICKHW register would result in unexpected results on these
hardwares. wpaul said the following for the issue.

  The vr_attach() routine unconditionally does this for all supported
  chips:

/*
 * Windows may put the chip in suspend mode when it
 * shuts down. Be sure to kick it in the head to wake it
 * up again.
 */
VR_CLRBIT(sc, VR_STICKHW, (VR_STICKHW_DS0|VR_STICKHW_DS1));

  The problem is, the VR_STICKHW register is not valid on all Rhine
  devices. The VT86C100A chip, which is present on the D-Link DFE-530TX
  boards, doesn't support power management, and its register space is
  only 128 bytes wide. The VR_STICKHW register offset falls outside this
  range. This may go unnoticed in most scenarios, but if you happen to have
  another PCI device in your system which is assigned the register
  space immediately after that of the Rhine, the vr(4) driver will
  incorrectly stomp it. In my case, the BIOS on my test board decided
  to put the register space for my PRO/100 ethernet board right next
  to the Rhine, and the Rhine driver ended up clobbering the IMR register
  of the PRO/100 device. (Long story short: the board kept locking up on
  boot. Took me the better part of the morning suss out why.)

  The strictly correct thing to do would be to check the PCI config space
  to make sure the device supports the power management capability and only
  write to the VR_STICKHW register if it does.

Instead of inspecting chip revision numbers for the availability of
VR_STICKHW register, check the existence of power management capability
of the hardware as wpaul suggested.

Reported by: wpaul
Suggested by: wpaul
OK'ed by: jhb

16 years agoFix two panics in lagg.
thompsa [Fri, 12 Oct 2007 03:03:16 +0000 (03:03 +0000)]
Fix two panics in lagg.

1. The locking was changed to shared but roundrobin mode still updated a
   pointer in the softc with the next tx interface to use. This will panic
   under high load. Change this to an atomically incremented sequence number in
   order to choose the tx port in round robin.

2. IFQ_HANDOFF will free the mbuf if the queue is full, this will then be freed
   again by lagg_start() and panic.  Reorganised the error handling and freeing
   to fix this.

MFC after: 3 days

16 years ago- After gcc 4.2 import, include/objc/objc-api.h requires objc-decls.h which
rafan [Fri, 12 Oct 2007 02:52:46 +0000 (02:52 +0000)]
- After gcc 4.2 import, include/objc/objc-api.h requires objc-decls.h which
  we did not install. Install objc-decls.h to fix.

PR: 116943
Reported by: beech
Submitted by: vanilla on -current, kan
MFC after: 1 week

16 years ago - Mention that less v408 was MFC'ed.
delphij [Thu, 11 Oct 2007 20:44:38 +0000 (20:44 +0000)]
 - Mention that less v408 was MFC'ed.
 - Mention tzdata2007h update.  An in advance &merged; is
   added so that we don't have to change this again when
   it is actually MFC'ed.

Approved by: re (bmah)

16 years agoAllow recursion on the 'zones' internal UMA zone.
jhb [Thu, 11 Oct 2007 20:11:27 +0000 (20:11 +0000)]
Allow recursion on the 'zones' internal UMA zone.

Submitted by: thompsa
MFC after: 1 week
Approved by: re (kensmith)
Discussed with: jeff

16 years agoTo honor the birth of RELENG_7 bump HEAD to 8.0-CURRENT.
kensmith [Thu, 11 Oct 2007 04:28:08 +0000 (04:28 +0000)]
To honor the birth of RELENG_7 bump HEAD to 8.0-CURRENT.

Approved by: re (implicit)

16 years agoRemove file which was accidently repo-copied to the wrong destination
simon [Wed, 10 Oct 2007 17:34:29 +0000 (17:34 +0000)]
Remove file which was accidently repo-copied to the wrong destination
file-name.  The file will be nuked entirely from the repository later.

Approved by: re (kensmith)

16 years agoFor 7.0 make the shared lib "version" '3'.
obrien [Wed, 10 Oct 2007 16:59:14 +0000 (16:59 +0000)]
For 7.0 make the shared lib "version" '3'.
Approved by: re(kensmith)

16 years agoProperly regen src.conf(5).
ru [Wed, 10 Oct 2007 06:04:43 +0000 (06:04 +0000)]
Properly regen src.conf(5).

Approved by: re (kensmith)

16 years agoTweak the handling of "WITHOUT_LIBPTHREAD". Also remove the accidental
obrien [Tue, 9 Oct 2007 23:31:11 +0000 (23:31 +0000)]
Tweak the handling of "WITHOUT_LIBPTHREAD".  Also remove the accidental
treatment of 'LIBKSE' as an "old style" knob.

Submitted by: ru
Approved by: re(kensmith)

16 years agoAdd PCI ID for ICH9 AHCI w/ two adapters.
delphij [Tue, 9 Oct 2007 20:15:09 +0000 (20:15 +0000)]
Add PCI ID for ICH9 AHCI w/ two adapters.

Tested by: Abdullah Ibn Hamad Al-Marri <almarrie gmail com>
Approved by: sos
Approved by: re (kensmith)

16 years agoMdoc fix.
obrien [Tue, 9 Oct 2007 17:53:33 +0000 (17:53 +0000)]
Mdoc fix.

Submitted by: ru
Approved by: re(implicit)

16 years agoPulled the trigger 2hr 50min late (Pago Pago time).
obrien [Tue, 9 Oct 2007 17:51:11 +0000 (17:51 +0000)]
Pulled the trigger 2hr 50min late (Pago Pago time).
Submitted by: ru
Approved by: re (implicit)

16 years agoUpdate to version 3.70.03.007 from the vendor. This adds support for new
scottl [Tue, 9 Oct 2007 17:43:57 +0000 (17:43 +0000)]
Update to version 3.70.03.007 from the vendor.  This adds support for new
SAS-enabled cards.  It also makes the driver MPSAFE, eliminating some
problems that resulted from CAM becoming MPSAFE.  Many thanks to 3Ware/AMCC
for continuing to support FreeBSD.

Submitted by: Manjunath Ranganathaiah
Approved by: re

16 years agoCorrect lib name.
obrien [Tue, 9 Oct 2007 17:22:00 +0000 (17:22 +0000)]
Correct lib name.

Submitted by: ru
Approved by: re(kensmith)

16 years agoRepo copy libpthreads to libkse.
obrien [Tue, 9 Oct 2007 14:16:39 +0000 (14:16 +0000)]
Repo copy libpthreads to libkse.
This introduces the WITHOUT_LIBKSE nob,
and changes WITHOUT_LIBPTHREADS to mean with neither threading libs.
Approved by: re(kensmith)

16 years agoRepo copy libpthreads to libkse.
obrien [Tue, 9 Oct 2007 13:42:34 +0000 (13:42 +0000)]
Repo copy libpthreads to libkse.
This introduces the WITHOUT_LIBKSE nob,
and changes WITHOUT_LIBPTHREADS to mean with neither threading libs.
Approved by: re(kensmith)

16 years agoWhen the EC times out (common with Compaqs), it may report a design
njl [Tue, 9 Oct 2007 07:51:46 +0000 (07:51 +0000)]
When the EC times out (common with Compaqs), it may report a design
voltage of 0.  This can result in a divide by zero trap.  Add a guard
for this case.  The value of lfcap is checked in acpi_battery_bif_valid()
just before this, so it is safe.

Reportd by: sam
Approved by: re
MFC after: 3 days

16 years agoFix the HPET table probe routine to run from device_identify() instead
njl [Tue, 9 Oct 2007 07:48:07 +0000 (07:48 +0000)]
Fix the HPET table probe routine to run from device_identify() instead
of directly from acpi0.  Before it would attach prior to the sysresource
devices, causing the later allocation of its memory range to fail and
print a warning like "acpi0: reservation of fed00000, 1000 (3) failed".
Use an explicit define for our probe order base value of 10.

Help from: jhb
Tested by: Abdullah Ibn Hamad Al-Marri <almarrie / gmail.com>
MFC after: 3 days
Approved by: re

16 years agoDeprecate use of the early.sh script as advertised when the support for
dougb [Tue, 9 Oct 2007 07:30:14 +0000 (07:30 +0000)]
Deprecate use of the early.sh script as advertised when the support for
local rc.d scripts in the overall boot order was added.

Proper rc.d scripts are run by rc.subr in a subshell, whereas scripts that
end in .sh are sourced into rc's shell. The latter has potential to create
serious boot problems, and there is no reason that the same functionality
cannot be added by the user in the form of a proper rc.d script (as
opposed to being added by the user in the form of /etc/rc.early).

This script will be removed prior to the 8.0 branch.

Approved by: re (kensmith)

16 years agoRemove pre-rc.d compatibility shims that were added before the 5.0 branch
dougb [Tue, 9 Oct 2007 07:20:44 +0000 (07:20 +0000)]
Remove pre-rc.d compatibility shims that were added before the 5.0 branch
for pre-5.0 variable names.

Remove two dhcp compatibility variables added after the 5.1-RELEASE.

Remove the now-unused support for these shims.

Approved by: re (kensmith)

16 years ago - Fix from pr kern/115469; Don't redeliver a signal once it has been
jeff [Tue, 9 Oct 2007 00:03:39 +0000 (00:03 +0000)]
 - Fix from pr kern/115469; Don't redeliver a signal once it has been
   handled by the target process.

Contributed by: Tijl Coosemans <tijl@ulyssis.org>
Approved by: re

16 years ago - Bail out of tdq_idled if !smp_started or idle stealing is disabled. This
jeff [Mon, 8 Oct 2007 23:50:39 +0000 (23:50 +0000)]
 - Bail out of tdq_idled if !smp_started or idle stealing is disabled.  This
   fixes a bug on UP machines with SMP kernels where the idle thread
   constantly switches after trying to steal work from the local cpu.
 - Make the idle stealing code more robust against self selection.
 - Prefer to steal from the cpu with the highest load that has at least one
   transferable thread.  Before we selected the cpu with the highest
   transferable count which excludes bound threads.

Collaborated with: csjp
Approved by: re

16 years ago - Restore historical sched_yield() behavior by changing sched_relinquish()
jeff [Mon, 8 Oct 2007 23:45:24 +0000 (23:45 +0000)]
 - Restore historical sched_yield() behavior by changing sched_relinquish()
   to simply switch rather than lowering priority and switching.  This allows
   threads of equal priority to run but not lesser priority.

Discussed with: davidxu
Reported by: NIIMI Satoshi <sa2c@sa2c.net>
Approved by: re

16 years ago - Restore historical yield() behavior by manually lowering priority and
jeff [Mon, 8 Oct 2007 23:40:40 +0000 (23:40 +0000)]
 - Restore historical yield() behavior by manually lowering priority and
   switching.

Approved by: re

16 years ago - Fix ULE in kernels without PREEMPTION compiled in by always enabling the
jeff [Mon, 8 Oct 2007 23:37:28 +0000 (23:37 +0000)]
 - Fix ULE in kernels without PREEMPTION compiled in by always enabling the
   critical_exit() owepreempt check.  ULE will always use owepreempt to
   preempt the idle thread.  This change does not effect 4BSD since it will
   never set owepreempt without PREEMPTION enabled.
 - Remove some unused code from choosethread().

Discussed with: jhb
Approved by: re

16 years agoMFV for tzdata2007h
edwin [Mon, 8 Oct 2007 21:41:29 +0000 (21:41 +0000)]
MFV for tzdata2007h

Timezone data changes in the following locations:

- Egypt (push possible DST ending one year ahead)
- Iran gets DST again in 2008.
- Palestine DST time for this year fixed.
- Brasils DST rule change in 2008.
- Venezuela time moves half an hour back at the end of this year.

PR: conf/116900
Approved by: re (Ken Smith), grog@ (mentor)
MFC after: 1 week

16 years agoDescribe the process of importing new zoneinfo data and updating
edwin [Mon, 8 Oct 2007 21:16:38 +0000 (21:16 +0000)]
Describe the process of importing new zoneinfo data and updating
it into CVS.

PR: conf/116901
Approved by: re (Ken Smith), grog@ (mentor)

16 years agoDo not dereference NULL pointer.
kib [Mon, 8 Oct 2007 20:09:53 +0000 (20:09 +0000)]
Do not dereference NULL pointer.

Reported by: Peter Holm
Reviewed by: alc
Approved by: re (kensmith)

16 years agoDeny attempt to malloc unbounded amount of the memory.
kib [Mon, 8 Oct 2007 20:08:34 +0000 (20:08 +0000)]
Deny attempt to malloc unbounded amount of the memory.
Convert malloc()/bzero() to malloc(M_ZERO).

Obtained from:  OpenBSD
MFC after:      3 days
Approved by:    re (kensmith)

16 years agoDisable TLS for arm and sparc64 here as binutils 2.15 predate GNU TLS
marius [Mon, 8 Oct 2007 18:59:34 +0000 (18:59 +0000)]
Disable TLS for arm and sparc64 here as binutils 2.15 predate GNU TLS
support for these. This is in line with gnu/lib/libgomp/config.h and
gnu/lib/libstdc++/config.h.

Reviewed by: cognet, obrien
Approved by: re (kensmith)

16 years agoIn the rare case that vm_page_cache() actually frees the given page,
alc [Mon, 8 Oct 2007 18:01:38 +0000 (18:01 +0000)]
In the rare case that vm_page_cache() actually frees the given page,
it must first ensure that the page is no longer mapped.  This is
trivially accomplished by calling pmap_remove_all() a little earlier
in vm_page_cache().  While I'm in the neighborbood, make a related
panic message a little more useful.

Approved by: re (kensmith)
Reported by: Peter Holm and Konstantin Belousov
Reviewed by: Konstantin Belousov

16 years agoResolve conflicts to complete less v408 import.
delphij [Mon, 8 Oct 2007 16:17:42 +0000 (16:17 +0000)]
Resolve conflicts to complete less v408 import.

Approved by: re (kensmith)

16 years agoThis commit was generated by cvs2svn to compensate for changes in r172468,
delphij [Mon, 8 Oct 2007 16:14:52 +0000 (16:14 +0000)]
This commit was generated by cvs2svn to compensate for changes in r172468,
which included commits to RCS files with non-trunk default branches.

16 years agoVirgin import of less v408.
delphij [Mon, 8 Oct 2007 16:14:52 +0000 (16:14 +0000)]
Virgin import of less v408.

Approved by: re (kensmith)

16 years agoAdd FBSDID to all files in netinet so that people can more
silby [Sun, 7 Oct 2007 20:44:24 +0000 (20:44 +0000)]
Add FBSDID to all files in netinet so that people can more
easily include file version information in bug reports.

Approved by: re (kensmith)

16 years agoCorrect a lock assertion failure in sparc64's pmap_page_is_mapped() that is
alc [Sun, 7 Oct 2007 18:03:03 +0000 (18:03 +0000)]
Correct a lock assertion failure in sparc64's pmap_page_is_mapped() that is
a consequence of sparc64/sparc64/vm_machdep.c revision 1.76.  It occurs
when uma_small_free() frees a page.  The solution has two parts: (1) Mark
pages allocated with VM_ALLOC_NOOBJ as PG_UNMANAGED.  (2) Defer the lock
assertion in pmap_page_is_mapped() until after PG_UNMANAGED is tested.
This is safe because both PG_UNMANAGED and PG_FICTITIOUS are immutable
flags, i.e., they do not change state between the time that a page is
allocated and freed.

Approved by: re (kensmith)
PR: 116794

16 years agoImprove the debugging message:
silby [Sun, 7 Oct 2007 00:07:27 +0000 (00:07 +0000)]
Improve the debugging message:

TCP: [X.X.X.X]:X to [X.X.X.X]:X tcpflags 0x18<PUSH,ACK>; tcp_do_segment: FIN_WAIT_2: Received data after socket was closed, sending RST and removing tcpcb

So that it also includes how many bytes of data were received.  It now looks
like this:

TCP: [X.X.X.X]:X to [X.X.X.X]:X tcpflags 0x18<PUSH,ACK>; tcp_do_segment: FIN_WAIT_2: Received X bytes of data after socket was closed, sending RST and removing tcpcb

Approved by: re (gnn)

16 years agoThis patch adds an M_NOFREE flag which allows one to mark an mbuf as
kmacy [Sat, 6 Oct 2007 21:42:39 +0000 (21:42 +0000)]
This patch adds an M_NOFREE flag which allows one to mark an mbuf as
not being independently freeable. This allows one to embed an mbuf in
the cluster itself. This confers the benefits of the packet zone on
all cluster sizes. Embedded mbufs currently suffer from the same
limitation that packet zone mbufs do in that one cannot disconnect
them and pass them around independently of the cluster. It would
likely be possible to eliminate this limitation in the future by
adding a second reference for the mbuf itself.

Approved by: re(gnn)

16 years agoAllow drivers to free an mbuf without having the mbuf be touched if
kmacy [Sat, 6 Oct 2007 21:13:55 +0000 (21:13 +0000)]
Allow drivers to free an mbuf without having the mbuf be touched if
the driver has already freed any attached tags

Approved by: re(gnn)

16 years ago- Fix the one-2-one model to properly do a socantrecv()
rrs [Sat, 6 Oct 2007 13:23:42 +0000 (13:23 +0000)]
- Fix the one-2-one model to properly do a socantrecv()
Approved by: re@freeBSD.org (Ken Smith)

16 years agoAdd an MLINKS for pci_find_dbsf.9.
marius [Fri, 5 Oct 2007 22:50:44 +0000 (22:50 +0000)]
Add an MLINKS for pci_find_dbsf.9.

Submitted by: ru
Approved by: re (gnn)

16 years agoDisable TCP syncache debug logging by default. While useful in debugging
rwatson [Fri, 5 Oct 2007 22:39:44 +0000 (22:39 +0000)]
Disable TCP syncache debug logging by default.  While useful in debugging
problems with the syncache, it produces a lot of console noise and has led
to quite a few false positive bug reports.  It can be selectively
re-enabled when debugging specific problems by frobbing the same sysctl.

Discussed with: silby
Approved by: re (gnn)

16 years agoUse the correct pid when checking to see whether or not the /proc/<pid>
jhb [Fri, 5 Oct 2007 17:37:25 +0000 (17:37 +0000)]
Use the correct pid when checking to see whether or not the /proc/<pid>
directory itself (rather than any of its contents) is visible to the
current thread.

MFC after: 1 week
PR: kern/90063
Submitted by: john of 8192.net
Approved by: re (kensmith)

16 years agomanpage update for the recent commit to uscanner.c
luigi [Fri, 5 Oct 2007 15:17:14 +0000 (15:17 +0000)]
manpage update for the recent commit to uscanner.c
I also took the chance to make the list of supported devices a
bit more compact, as it was really long to read.

Even though re@ and Warner only saw the diffs for the code, i expect
their approval also covered the manpage update.

Approved by: re, imp (implicitly i hope)
MFC after: 3 days

16 years agoAdd entries for Epson multifunction scanner/printer/card readers,
luigi [Fri, 5 Oct 2007 07:26:39 +0000 (07:26 +0000)]
Add entries for Epson multifunction scanner/printer/card readers,
with all functions supported. This is done adding usb device IDs
to the table of recognised devices (because there is no standard
'scanner' class, so no other way to recognise them), and with
a small change to the uscanner attach routine that prevents
reconfiguring the whole USB device while we are dealing only with
one of its USB interfaces.

The latter part has been suggested by Steinar Hamre in
http://www.freebsd.org/cgi/query-pr.cgi?pr=107665 , i have
only added a bit of explaination to the code.

I have personally tried this on the Epson DX-5050 and DX-6000
devices (on the US market they have different names, CX-something).
I have good reasons to think that, possibly with the mere addition
of more USB ids to the table in uscanner.c, this should work with
all Epson multifunction devices in that family (from DX-3800 to
DX-7000 - these units are in the 50-120$ price range).
More details on related topics (SANE configuration, OCR, etc.)
at http://info.iet.unipi.it/~luigi/FreeBSD/dx5050.html

Manpage updates coming soon.

Approved by: re, imp
MFC after: 3 days

16 years agoAMD CS5536 and VIA 8237S support.
brueffer [Fri, 5 Oct 2007 07:06:51 +0000 (07:06 +0000)]
AMD CS5536 and VIA 8237S support.

Approved by: re (blanket)

16 years agoMake selector parsing accept the dot "." as an alternative selector
se [Thu, 4 Oct 2007 22:27:08 +0000 (22:27 +0000)]
Make selector parsing accept the dot "." as an alternative selector
with identical meaning as the colon ":". This is to support a syntax
that is more similar to a PCI device specification in the device hints
file. The selector is not fully compatible with the specification in
the hints file, since entries in that file use a different prefix,
which needs to be added to the getsel() routine, if full support of
that syntax is found to be desirable.

Approved by: re (Ken Smith)

16 years agoRestore compatibility with version before introduction of PCI domains.
se [Thu, 4 Oct 2007 22:18:53 +0000 (22:18 +0000)]
Restore compatibility with version before introduction of PCI domains.
PCI selectors with 2 or 3 elements behave exactly as before (i.e. the
domain is 0 and in the 2 element case, the function is also 0).
The form with 4 selector elements works as in the previous revision
and provides the PCI domain number as the left-most selector element.

This change allows old scripts (which used the 2 or 3 selector element
formats) to be kept. Without this patch, the 3 element form was parsed
as starting with a domain number (and the function was assumed to be 0),
with this patch, the domain is assumed to be 0 (and the last value is
used as the function number).

The man page is updated to describe the new selector semantics.

Approved by: re (Ken Smith)

16 years agoAdd support for the VIA 8237S
sos [Thu, 4 Oct 2007 19:17:16 +0000 (19:17 +0000)]
Add support for the VIA 8237S
Fix the LBA28/LBA48 crossover bug.

Approved by: re@

16 years agoAlso boot *.debug if everything else fails.
obrien [Thu, 4 Oct 2007 18:29:52 +0000 (18:29 +0000)]
Also boot *.debug if everything else fails.

Approved by: re(gnn)

16 years agoFix lock leak leading to the 'System call <name> returning with 1 locks held'
pjd [Thu, 4 Oct 2007 17:51:59 +0000 (17:51 +0000)]
Fix lock leak leading to the 'System call <name> returning with 1 locks held'
panic.

Reported by: kris
Approved by: re (kensmith)

16 years agoMFp4: Provide a dummy verb "export" to shut up the message
delphij [Thu, 4 Oct 2007 17:11:48 +0000 (17:11 +0000)]
MFp4: Provide a dummy verb "export" to shut up the message
showed up at start when NFS is enabled.

Reported by: rafan
Approved by: re (tmpfs blanket)

16 years agoAdditional work is still needed before we can claim that tmpfs
delphij [Thu, 4 Oct 2007 17:08:46 +0000 (17:08 +0000)]
Additional work is still needed before we can claim that tmpfs
is stable enough for production usage.  Warn user upon mount.

Approved by: re (tmpfs blanket)

16 years agoThe exit status of a case statement where none of the patterns is matched
stefanf [Thu, 4 Oct 2007 16:14:48 +0000 (16:14 +0000)]
The exit status of a case statement where none of the patterns is matched
is supposed to be 0, not the status of the previous command.

Reported by: Eygene Ryabinkin
PR: 116559
Approved by: re (gnn)

16 years agoFix the module name matching to the drivers present in the kernel. Previously
thompsa [Thu, 4 Oct 2007 09:45:41 +0000 (09:45 +0000)]
Fix the module name matching to the drivers present in the kernel. Previously
it would return true on a partial match where it would think the edsc module
was already present by having a positive match on 'ed'.  This changes it so
that it compares the full string including the nul terminators.

This also fixes a buffer overflow in the ifkind variable where the length of
the interface name in *argv wasnt checked for size.

Reviewed by: brooks
Approved by: re (gnn)

16 years ago- We should return error = 0 and the upper processing would
rrs [Thu, 4 Oct 2007 09:29:33 +0000 (09:29 +0000)]
- We should return error = 0 and the upper processing would
  return a zero length read. Otherwise we don't return the
  right error indication.

Approved by: re@freebsd.org (gnn)

16 years agoRecognize the CS5536 support chip for the AMD Geode LX CPU to enable
phk [Thu, 4 Oct 2007 06:21:54 +0000 (06:21 +0000)]
Recognize the CS5536 support chip for the AMD Geode LX CPU to enable
UDMA modes.

Please notice that Soekris NET5501 bios versions before 1.32f has a bug
that prevents this from working.

Approved by: re (gnn)
MFC: 2 weeks

16 years agoFreeBSD 6.2 is now known to mdoc.
ru [Thu, 4 Oct 2007 04:39:06 +0000 (04:39 +0000)]
FreeBSD 6.2 is now known to mdoc.

Approved by: re (kensmith)

16 years agoMFV: recent mdoc(7) changes.
ru [Thu, 4 Oct 2007 04:38:23 +0000 (04:38 +0000)]
MFV: recent mdoc(7) changes.

Approved by: re (kensmith)

16 years agoCorrect a buffer overflow in OpenSSL SSL_get_shared_ciphers().
simon [Wed, 3 Oct 2007 21:38:57 +0000 (21:38 +0000)]
Correct a buffer overflow in OpenSSL SSL_get_shared_ciphers().

Security: FreeBSD-SA-07:08.openssl
Approved by: re (security blanket)

16 years agoThis commit was generated by cvs2svn to compensate for changes in r172423,
ru [Wed, 3 Oct 2007 06:04:40 +0000 (06:04 +0000)]
This commit was generated by cvs2svn to compensate for changes in r172423,
which included commits to RCS files with non-trunk default branches.

16 years agoPull up current -mdoc version from vendor. This includes the
ru [Wed, 3 Oct 2007 06:04:40 +0000 (06:04 +0000)]
Pull up current -mdoc version from vendor.  This includes the
following changes:

: 2007-10-02  Ruslan Ermilov  <ru@FreeBSD.org>
:
:         * tmac/doc-common, tmac/groff_doc.man: Add FreeBSD 6.2, document
:         FreeBSD 5.5 and 7.0.
:
:         * tmac/doc-syms: Give better names for System V releases.

: 2007-05-30  Werner LEMBERG <wl@gnu.org>
:
:         * tmac/groff_mdoc.man: Convert `-' to `\-' where appropriate.
:         Recommend `tbl' instead of `-column' lists for more complicated
:         cases.

16 years agoSort as per README.
ru [Wed, 3 Oct 2007 05:51:20 +0000 (05:51 +0000)]
Sort as per README.

Approved by: re (kensmith)

16 years agoRemoved "tail +5" from the command used to sanity check changes to
ru [Wed, 3 Oct 2007 05:44:27 +0000 (05:44 +0000)]
Removed "tail +5" from the command used to sanity check changes to
mtree files -- the 5-line header is no longer printed when mtree(8)
is run with -n (as of mtree/create.c,v 1.34).

Approved by: re (kensmith)

16 years agoCorrection to the log for rev. 1.4:
obrien [Wed, 3 Oct 2007 01:26:24 +0000 (01:26 +0000)]
Correction to the log for rev. 1.4:
"One must now explicitly specify -ftree-vrp if one wants it".

Approved by: re(KenSmith)

16 years agoDo not imply -ftree-vrp with -O2 and above. One must implicitly specify
obrien [Tue, 2 Oct 2007 20:06:14 +0000 (20:06 +0000)]
Do not imply -ftree-vrp with -O2 and above.  One must implicitly specify
'-ftree-vrp' if one wants it.
Some bad code generation has been tracked to -ftree-vrp.  jdk1{5,6} are
notable examples.

Approved by: re(kensmith)

16 years agoFix usage example since we don't have wext driver.
kevlo [Tue, 2 Oct 2007 15:30:28 +0000 (15:30 +0000)]
Fix usage example since we don't have wext driver.

Reviewed by: sam
Approved by: re (kensmith)

16 years agoFix sx_try_slock(), so it only fails when there is an exclusive owner.
pjd [Tue, 2 Oct 2007 14:48:48 +0000 (14:48 +0000)]
Fix sx_try_slock(), so it only fails when there is an exclusive owner.
Before that fix, it was possible for the function to fail if number
of sharers changes between 'x = sx->sx_lock' step and atomic_cmpset_acq_ptr()
call.

This fixes ZFS problem when ZFS returns strange EIO errors under load.
In ZFS there is a code that depends on the fact that sx_try_slock() can
only fail if there is an exclusive owner.

Discussed with: attilio
Reviewed by: jhb
Approved by: re (kensmith)