]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoThis commit was generated by cvs2svn to compensate for changes in r161701,
Warner Losh [Mon, 28 Aug 2006 17:26:38 +0000 (17:26 +0000)]
This commit was generated by cvs2svn to compensate for changes in r161701,
which included commits to RCS files with non-trunk default branches.

17 years agoImport on vendor branch two files that have been tweaked to unbreak
Warner Losh [Mon, 28 Aug 2006 17:26:38 +0000 (17:26 +0000)]
Import on vendor branch two files that have been tweaked to unbreak
the build.  The openbsm folks are free to fix it in any other way they
see fit once they resurface.

Basically, make everything always be const char **, even though const
char ** usually should be 'const char * const *' in most cases.  This
makes the three different definitions consistant and allows world to
build.

17 years agoFREE -> free
Suleiman Souhlal [Mon, 28 Aug 2006 13:52:27 +0000 (13:52 +0000)]
FREE -> free

Submitted by: rdivacky

17 years agoMFi386 parts of rev 1.55 (modulo real MD parts):
Alexander Leidinger [Mon, 28 Aug 2006 13:09:24 +0000 (13:09 +0000)]
MFi386 parts of rev 1.55 (modulo real MD parts):
 - implement CLONE_PARENT semantic
 - lock proc in the currently disabled part of CLONE_THREAD

Submitted by: rdivacky

17 years agopthread_sigmask is in thr_sig.c, remove this file.
David Xu [Mon, 28 Aug 2006 12:29:54 +0000 (12:29 +0000)]
pthread_sigmask is in thr_sig.c, remove this file.

17 years agoImprove input parsing:
Poul-Henning Kamp [Mon, 28 Aug 2006 08:27:02 +0000 (08:27 +0000)]
Improve input parsing:

Add "-C <column>" and "-d <delims>" options to chop up input lines.

Make '#' a comment character, rest of line is ignored.

Submitted by: Dmitry Morozovsky <marck@rinet.ru>

17 years agoProvide the installation script for the lib32 distribution.
Ruslan Ermilov [Mon, 28 Aug 2006 08:13:56 +0000 (08:13 +0000)]
Provide the installation script for the lib32 distribution.

17 years agoMove some historical artefacts into the attic.
Ruslan Ermilov [Mon, 28 Aug 2006 08:12:49 +0000 (08:12 +0000)]
Move some historical artefacts into the attic.

17 years agoAllow the DESTDIR to be specified without a terminating slash
Ruslan Ermilov [Mon, 28 Aug 2006 08:08:57 +0000 (08:08 +0000)]
Allow the DESTDIR to be specified without a terminating slash
(all other install scripts allow this).

17 years agoCosmetics.
Ruslan Ermilov [Mon, 28 Aug 2006 08:06:21 +0000 (08:06 +0000)]
Cosmetics.

17 years agoUpdate comments about interrupted mutex locking.
David Xu [Mon, 28 Aug 2006 07:09:27 +0000 (07:09 +0000)]
Update comments about interrupted mutex locking.

17 years ago1. Attempt to take one bullet out of the foot-shooting gun by silently
Doug Barton [Mon, 28 Aug 2006 06:41:50 +0000 (06:41 +0000)]
1. Attempt to take one bullet out of the foot-shooting gun by silently
ignoring errors when sourcing rc.conf* files. The most common error
occurs when users put a command of some sort into those files.
(ifconfig is a popular choice)

2. Make the file rotation logic simpler by starting one down from
the "top" of the list, rather than at the top.

3. Try to make file rotation more secure by calling unlink(1) on all
new file names before rotating an old file to the new name, rather than
merely calling 'rm -f' on any files that exceed the number of files
to save.

17 years agoKill unused files.
David Xu [Mon, 28 Aug 2006 05:01:31 +0000 (05:01 +0000)]
Kill unused files.

17 years agoUse umutex APIs to implement pthread_mutex, member pp_mutexq is added
David Xu [Mon, 28 Aug 2006 04:52:50 +0000 (04:52 +0000)]
Use umutex APIs to implement pthread_mutex, member pp_mutexq is added
into pthread structure to keep track of locked PTHREAD_PRIO_PROTECT mutex,
no real mutex code is changed, the mutex locking and unlocking code should
has same performance as before.

17 years agoAdd umutex APIs.
David Xu [Mon, 28 Aug 2006 04:47:27 +0000 (04:47 +0000)]
Add umutex APIs.

17 years agoRegenerate.
David Xu [Mon, 28 Aug 2006 04:28:25 +0000 (04:28 +0000)]
Regenerate.

17 years agoThis is initial version of POSIX priority mutex support, a new userland
David Xu [Mon, 28 Aug 2006 04:24:51 +0000 (04:24 +0000)]
This is initial version of POSIX priority mutex support, a new userland
mutex structure is added as following:
struct umutex {
        __lwpid_t       m_owner;
        uint32_t        m_flags;
        uint32_t        m_ceilings[2];
        uint32_t        m_spare[4];
};
The m_owner represents owner thread, it is a thread id, in non-contested
case, userland can simply use atomic_cmpset_int to lock the mutex, if the
mutex is contested, high order bit will be set, and userland should do locking
and unlocking via kernel syscall. Flag UMUTEX_PRIO_INHERIT represents
pthread's PTHREAD_PRIO_INHERIT mutex, which when contention happens, kernel
should do priority propagating. Flag UMUTEX_PRIO_PROTECT indicates it is
pthread's PTHREAD_PRIO_PROTECT mutex, userland should initialize m_owner
to contested state UMUTEX_CONTESTED, then atomic_cmpset_int will be failure
and kernel syscall should be invoked to do locking, this becauses
for such a mutex, kernel should always boost the thread's priority before
it can lock the mutex, m_ceilings is used by PTHREAD_PRIO_PROTECT mutex,
the first element is used to boost thread's priority when it locked the mutex,
second element is used when the mutex is unlocked, the PTHREAD_PRIO_PROTECT
mutex's link list is kept in userland, the m_ceiling[1] is managed by thread
library so kernel needn't allocate memory to keep the link list, when such
a mutex is unlocked, kernel reset m_owner to UMUTEX_CONTESTED.
Flag USYNC_PROCESS_SHARED indicate if the synchronization object is process
shared, if the flag is not set, it saves a vm_map_lookup() call.

The umtx chain is still used as a sleep queue, when a thread is blocked on
PTHREAD_PRIO_INHERIT mutex, a umtx_pi is allocated to support priority
propagating, it is dynamically allocated and reference count is used,
it is not optimized but works well in my tests, while the umtx chain has
its own locking protocol, the priority propagating protocol are all protected
by sched_lock because priority propagating function is called with sched_lock
held from scheduler.

No visible performance degradation is found which these changes. Some parameter
names in _umtx_op syscall are renamed.

17 years agoImplement casuword32, compare and set user integer, thank Marcel Moolenarr
David Xu [Mon, 28 Aug 2006 02:28:15 +0000 (02:28 +0000)]
Implement casuword32, compare and set user integer, thank Marcel Moolenarr
who wrote the IA64 version of casuword32.

17 years agoRefactor vm_page_sleep_if_busy() so that the test for a busy page is
Alan Cox [Sun, 27 Aug 2006 19:50:13 +0000 (19:50 +0000)]
Refactor vm_page_sleep_if_busy() so that the test for a busy page is
inlined and a procedure call is made in the rare case, i.e., when it is
necessary to sleep.  In this case, inlining the test actually makes the
kernel smaller.

17 years agoFix video playing and network connections in realplayer (and most likely
Alexander Leidinger [Sun, 27 Aug 2006 18:51:32 +0000 (18:51 +0000)]
Fix video playing and network connections in realplayer (and most likely
other stuff) in the osrelease=2.6.16 case:
 - implement CLONE_PARENT semantic
 - fix TLS loading in clone CLONE_SETTLS
 - lock proc in the currently disabled part of CLONE_THREAD

I suggest to not unload the linux module after testing this, there are
some "<defunct>" processes hanging around after exiting (they aren't
with osrelease=2.4.2) and they may panic your kernel when unloading the
linux module. They are in state Z and some of them consume CPU according
to ps. But I don't trust the CPU part, the idle threads gets too much CPU
that this may be possible (accumulating idle, X and 2 defunct processes
results in 104.7%, this looks to much to be a rounding error).

Noticed by: Intron <mag@intron.ac>
Submitted by: rdivacky (in collaboration with Intron)
Tested by: Intron, netchild
Reviewed by: jhb (previous version)

17 years agoNew release notes: GCC 3.4.6, OpenSSL 0.9.8b.
Bruce A. Mah [Sun, 27 Aug 2006 16:18:09 +0000 (16:18 +0000)]
New release notes:  GCC 3.4.6, OpenSSL 0.9.8b.

17 years agovisionary thoughts..
Wilko Bulte [Sun, 27 Aug 2006 14:29:10 +0000 (14:29 +0000)]
visionary thoughts..

MFC after: 1 week

17 years agoFill in dump_avail[] before pmap_boostrap() is called so that
Olivier Houchard [Sun, 27 Aug 2006 13:23:51 +0000 (13:23 +0000)]
Fill in dump_avail[] before pmap_boostrap() is called so that
ARM_USE_SMALL_ALLOC work.

17 years agoResurrect reference to (contemporary) kern.ipc.nmbclusters.
Dmitry Morozovsky [Sun, 27 Aug 2006 12:57:37 +0000 (12:57 +0000)]
Resurrect reference to (contemporary) kern.ipc.nmbclusters.

Suggested by: ru

MFC after: 3 days

17 years agoWhen stopping powerd, set the CPU frequency back to its maximum value
Colin Percival [Sun, 27 Aug 2006 11:04:39 +0000 (11:04 +0000)]
When stopping powerd, set the CPU frequency back to its maximum value
(i.e., what it was almost certainly at before powerd was started).

Submitted by: R.B. Riddick
MFC after: 3 days

17 years agoregen
Alexander Leidinger [Sun, 27 Aug 2006 08:58:00 +0000 (08:58 +0000)]
regen

17 years agoAdd the linux statfs64 call. This allows Tivoli backup to proceed a little
Alexander Leidinger [Sun, 27 Aug 2006 08:56:54 +0000 (08:56 +0000)]
Add the linux statfs64 call. This allows Tivoli backup to proceed a little
but further on -current (still not successful, but a step into the right
direction).

Sponsored by: Google SoC 2006
Submitted by: rdivacky
Tested by: Paul Mather <paul@gromit.dlib.vt.edu>

17 years agoUse ports INDEX-7 instead of INDEX-6
Doug Barton [Sun, 27 Aug 2006 08:12:53 +0000 (08:12 +0000)]
Use ports INDEX-7 instead of INDEX-6

Submitted by: Niclas Zeising <lothrandil@n00b.apagnu.se>

17 years ago- Add new service class definitions and a new attribute identifier definition
Markus Brueffer [Sat, 26 Aug 2006 23:16:35 +0000 (23:16 +0000)]
- Add new service class definitions and a new attribute identifier definition
- Update URL of the Assigned Numbers document for SDP

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

17 years agoGCC 3.4.6 gets confused on this file and produces bogus warning.
Alexander Kabaev [Sat, 26 Aug 2006 21:48:00 +0000 (21:48 +0000)]
GCC 3.4.6 gets confused on this file and produces bogus warning.
Shut it up.

17 years agoResolve conflicts after GCC 3.4.6 20060825 import.
Alexander Kabaev [Sat, 26 Aug 2006 21:37:21 +0000 (21:37 +0000)]
Resolve conflicts after GCC 3.4.6 20060825 import.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r161657,
Alexander Kabaev [Sat, 26 Aug 2006 21:30:30 +0000 (21:30 +0000)]
This commit was generated by cvs2svn to compensate for changes in r161657,
which included commits to RCS files with non-trunk default branches.

17 years agoGcc 3.4.6 F77 runtime support bits (as of 2006/08/25 #116475).
Alexander Kabaev [Sat, 26 Aug 2006 21:30:30 +0000 (21:30 +0000)]
Gcc 3.4.6 F77 runtime support bits (as of 2006/08/25 #116475).
.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r161655,
Alexander Kabaev [Sat, 26 Aug 2006 21:30:26 +0000 (21:30 +0000)]
This commit was generated by cvs2svn to compensate for changes in r161655,
which included commits to RCS files with non-trunk default branches.

17 years agoGcc 3.4.6 Objective C support bits (as of 2006/08/25 #116475).
Alexander Kabaev [Sat, 26 Aug 2006 21:30:26 +0000 (21:30 +0000)]
Gcc 3.4.6 Objective C support bits (as of 2006/08/25 #116475).

17 years agoThis commit was generated by cvs2svn to compensate for changes in r161653,
Alexander Kabaev [Sat, 26 Aug 2006 21:29:46 +0000 (21:29 +0000)]
This commit was generated by cvs2svn to compensate for changes in r161653,
which included commits to RCS files with non-trunk default branches.

17 years agoGcc 3.4.6 C++ support bits (as of 2006/08/25 #116475).
Alexander Kabaev [Sat, 26 Aug 2006 21:29:46 +0000 (21:29 +0000)]
Gcc 3.4.6 C++ support bits (as of 2006/08/25 #116475).

17 years agoThis commit was generated by cvs2svn to compensate for changes in r161651,
Alexander Kabaev [Sat, 26 Aug 2006 21:29:10 +0000 (21:29 +0000)]
This commit was generated by cvs2svn to compensate for changes in r161651,
which included commits to RCS files with non-trunk default branches.

17 years agoGcc 3.4.6 as of 2006/08/25 #116475.
Alexander Kabaev [Sat, 26 Aug 2006 21:29:10 +0000 (21:29 +0000)]
Gcc 3.4.6 as of 2006/08/25 #116475.

17 years agoChase after phk@: remove reference to (now obsoleted) NMBCLUSTERS.
Dmitry Morozovsky [Sat, 26 Aug 2006 21:24:22 +0000 (21:24 +0000)]
Chase after phk@: remove reference to (now obsoleted) NMBCLUSTERS.

MFC after: 1 week

17 years agoFix panic associated with file creation via RPC/NFS when the MLS policy
Christian S.J. Peron [Sat, 26 Aug 2006 20:13:35 +0000 (20:13 +0000)]
Fix panic associated with file creation via RPC/NFS when the MLS policy
is loaded. This problem stems from the fact that the policy is not properly
initializing the mac label associated with the NFS daemon.

Obtained from: TrustedBSD Project
Discussed with: rwatson

17 years agoAdd 2400 f/w support.
Matt Jacob [Sat, 26 Aug 2006 18:40:25 +0000 (18:40 +0000)]
Add 2400 f/w support.

17 years agoAdd QLogic 2400 (4Gb) firmware.
Matt Jacob [Sat, 26 Aug 2006 18:39:18 +0000 (18:39 +0000)]
Add QLogic 2400 (4Gb) firmware.

17 years agoAllow the user process to query the kernel's notion of a maximum
Robert Watson [Sat, 26 Aug 2006 17:59:31 +0000 (17:59 +0000)]
Allow the user process to query the kernel's notion of a maximum
audit record size at run-time, which can be used by the user
process to size the user space buffer it reads into from the audit
pipe.

Perforce change: 105098
Obtained from: TrustedBSD Project

17 years agoFix for a bug that causes the computation of "len" in tcp_output() to
Mohan Srinivasan [Sat, 26 Aug 2006 17:53:19 +0000 (17:53 +0000)]
Fix for a bug that causes the computation of "len" in tcp_output() to
get messed up, resulting in an inconsistency between the TCP state
and so_snd.

17 years agoFix another bug introduced with rev. 1.204; in vfs_donmount() if
Marius Strobl [Sat, 26 Aug 2006 16:28:19 +0000 (16:28 +0000)]
Fix another bug introduced with rev. 1.204; in vfs_donmount() if
the 'vfs_getopt(optlist, "errmsg", (void **)&errmsg, &errmsg_len)'
call fails, 'errmsg' is left uninitialized, making the later tests
against NULL meaningless, and the uses bogus. Thus initialize
'errmsg' to NULL beforehand. [1]
While at it, remove the superfluous assignment of 0 to 'errmsg_len'
if the above mentioned call fails as it's already initialized to 0.

Submitted by: Michael Plass [1]

17 years agoNew release notes: amdsmb(4), nfsmb(4), bpf(4) timestamp fix (+MFC),
Bruce A. Mah [Sat, 26 Aug 2006 14:00:42 +0000 (14:00 +0000)]
New release notes:  amdsmb(4), nfsmb(4), bpf(4) timestamp fix (+MFC),
logger(1) -P (+MFC).

MFCs noted:  em(4) 6.1.4, traceroute(8) -e, traceroute (8) complete
ICMP unreacable decoding.

Modified release notes:  OpenBSM 1.0alpha9.

17 years agoUpdate.
Pawel Jakub Dawidek [Sat, 26 Aug 2006 11:32:40 +0000 (11:32 +0000)]
Update.

17 years agoThe "taskqueue_fast" spinlocks were renamed to "fast_taskqueue" in
Suleiman Souhlal [Sat, 26 Aug 2006 11:21:25 +0000 (11:21 +0000)]
The "taskqueue_fast" spinlocks were renamed to "fast_taskqueue" in
subr_taskqueue.c:r1.32

Reported by: rdivacky

17 years agoCorrect the number of retries in a futex_wake() call.
Alexander Leidinger [Sat, 26 Aug 2006 10:36:16 +0000 (10:36 +0000)]
Correct the number of retries in a futex_wake() call.

Sponsored by: Google SoC 2006
Submitted by: rdivacky

17 years agoUpdate FreeBSD upgrade notes for OpenBSM.
Robert Watson [Sat, 26 Aug 2006 10:35:54 +0000 (10:35 +0000)]
Update FreeBSD upgrade notes for OpenBSM.

Obtained from: TrustedBSD Project

17 years agoUpdate kernel OpenBSM parts, especially src/sys/bsm, for the OpenBSM
Robert Watson [Sat, 26 Aug 2006 08:17:58 +0000 (08:17 +0000)]
Update kernel OpenBSM parts, especially src/sys/bsm, for the OpenBSM
1.0 alpha 9 import.  See the OpenBSM import commit message for a
detailed summary of changes.

Obtained from:  TrustedBSD Project

17 years agoUpdate for OpenBSM 1.0 alpha 9.
Robert Watson [Sat, 26 Aug 2006 08:12:17 +0000 (08:12 +0000)]
Update for OpenBSM 1.0 alpha 9.

Obtained from: TrustedBSD Project

17 years agoResolove conflicts from OpenBSM 1.0 alpha 9 into audit_event.
Robert Watson [Sat, 26 Aug 2006 08:08:47 +0000 (08:08 +0000)]
Resolove conflicts from OpenBSM 1.0 alpha 9 into audit_event.

Obtained from: TrustedBSD Project

17 years agoVendor import of OpenBSM 1.0 alpha 9, with the following change history
Robert Watson [Sat, 26 Aug 2006 08:04:15 +0000 (08:04 +0000)]
Vendor import of OpenBSM 1.0 alpha 9, with the following change history
notes since the last import:

OpenBSM 1.0 alpha 9

- Rename many OpenBSM-specific constants and API elements containing the
  strings "BSM" and "bsm" to "AUDIT" and "audit", observing that this is true
  for almost all existing constants and APIs.
- Instead of passing a per-instance cookie directly into all audit filter
  APIs, pass in the audit filter daemon state pointer, which is then used by
  the module using an audit_filter_{get,set}cookie() API.  This will allow
  future service APIs provided by the filter daemon to maintain their own
  state -- for example, per-module preselection state.

OpenBSM 1.0 alpha 8

- Correct typo in definition of AUR_INT.
- Adopt OpenSolaris constant values for AUDIT_* configuration flags.
- Arguments to au_to_exec_args() and au_to_exec_env() no longer const.
- Add kernel versions of au_to_exec_args() and au_to_exec_env().
- Fix exec argument type that is printed for env strings from 'arg' to 'env'.
- New OpenBSM token version number assigned, constants added for other
  commonly seen version numbers.
- OpenBSM-specific events assigned numbers in the 43xxx range to avoid future
  collisions with Solaris.  Darwin events renamed to AUE_DARWIN_foo, as they
  are now deprecated numberings.
- autoconf now detects clock_gettime(), which is not available on Darwin.
- praudit output fixes relating to arg32 and arg64 tokens.
- Maximum record size updated to 64k-1 to match Solaris record size limit.
- Various style and comment cleanups in include files.

This is an MFC candidate to RELENG_6.

Obtained from: TrustedBSD Project

17 years agoThis commit was generated by cvs2svn to compensate for changes in r161630,
Robert Watson [Sat, 26 Aug 2006 08:04:15 +0000 (08:04 +0000)]
This commit was generated by cvs2svn to compensate for changes in r161630,
which included commits to RCS files with non-trunk default branches.

17 years agoPrevent a call to contigmalloc() that asks for more physical memory than
Alan Cox [Sat, 26 Aug 2006 02:43:23 +0000 (02:43 +0000)]
Prevent a call to contigmalloc() that asks for more physical memory than
the machine has from causing a panic.

Submitted by: Michael Plass
PR: 101668
MFC after: 3 days

17 years agoEliminate unused definitions. (They came from NetBSD.)
Alan Cox [Fri, 25 Aug 2006 23:51:11 +0000 (23:51 +0000)]
Eliminate unused definitions.  (They came from NetBSD.)

Discussed with: cognet, grehan, marcel

17 years agoPass whatever the value of NM down to lorder. This allows one to
Warner Losh [Fri, 25 Aug 2006 23:50:05 +0000 (23:50 +0000)]
Pass whatever the value of NM down to lorder.  This allows one to
override NM in Makefiles when, for example, cross compiling and have
that value be used by lorder.  NM normally isn't defined, so we pass a
null value to lorder.  lorder says 'NM=${NM-nm}' which causes it to
pickup the default value.

17 years agoMove the bridge hook after the loopback check so that IFF_SIMPLEX is honoured
Andrew Thompson [Fri, 25 Aug 2006 20:16:39 +0000 (20:16 +0000)]
Move the bridge hook after the loopback check so that IFF_SIMPLEX is honoured
on member interfaces. This makes us the same as OpenBSD/NetBSD.

MFC after: 3 days

17 years agoThe bridge cant hear its own transmissions so set IFF_SIMPLEX.
Andrew Thompson [Fri, 25 Aug 2006 20:11:56 +0000 (20:11 +0000)]
The bridge cant hear its own transmissions so set IFF_SIMPLEX.

PR: kern/102361
Tested by: Radim Kolar <hsn@netmag.cz>
MFC after: 3 days

17 years ago- Talk about memory allocation in interrupt context
Pav Lucistnik [Fri, 25 Aug 2006 19:04:42 +0000 (19:04 +0000)]
- Talk about memory allocation in interrupt context
- Add uam to see also

Submitted by: Devon H. O'Dell <devon.odell@coyotepoint.com>

17 years ago- Catch up with ongoing rwatson's socket work;
Maksim Yevmenkin [Fri, 25 Aug 2006 17:53:13 +0000 (17:53 +0000)]
- Catch up with ongoing rwatson's socket work;

- Fix a couple of LORs and panics;

- Temporarily remove the code that tries to cleanup sockets that stuck
  on accepting queues (both complete and incomplete). I'm taking an ostrich
  approach here until I find a better way to deal with sockets that were
  disconnected before accepting (i.e. while socket was on complete or
  incomplete accept queue).

17 years agoUse the pcb in stoppcbs[] if it is present for threads that were running
John Baldwin [Fri, 25 Aug 2006 16:20:17 +0000 (16:20 +0000)]
Use the pcb in stoppcbs[] if it is present for threads that were running
on other CPUs in system when a dump is written.

Submitted by: ups
Reviewed by: marcel
MFC after: 3 days

17 years agoNote that the system only allows a maximum of kern.kq_calloutmax timers.
Ceri Davies [Fri, 25 Aug 2006 15:19:47 +0000 (15:19 +0000)]
Note that the system only allows a maximum of kern.kq_calloutmax timers.

PR: docs/102353
Submitted by: phk
MFC after: 1 week

17 years agoFix comment.
Pawel Jakub Dawidek [Fri, 25 Aug 2006 15:13:49 +0000 (15:13 +0000)]
Fix comment.

17 years agoExplicitely set the "allocbuffer" field to NULL when creating a new dmamap.
Olivier Houchard [Fri, 25 Aug 2006 15:10:45 +0000 (15:10 +0000)]
Explicitely set the "allocbuffer" field to NULL when creating a new dmamap.

17 years agoDo not create dma maps with bus_dmamap_create, as we call
Olivier Houchard [Fri, 25 Aug 2006 13:38:42 +0000 (13:38 +0000)]
Do not create dma maps with bus_dmamap_create, as we call
bus_dmamem_alloc later which will overwrite the value, leading to a small
memory leak.

17 years agoEmulate what vfork does instead of using it in linux_vfork. This way
Alexander Leidinger [Fri, 25 Aug 2006 11:59:56 +0000 (11:59 +0000)]
Emulate what vfork does instead of using it in linux_vfork. This way
we can do the stuff we need to do with linux processes at fork and
don't panic the kernel at exit of the child.

Submitted by: rdivacky
Tested with: tst-vfork* (glibc regression tests)
Tested by: netchild

17 years agoDon't call suser_cred() directly from linux_sethostname(), as it just
Robert Watson [Fri, 25 Aug 2006 11:02:42 +0000 (11:02 +0000)]
Don't call suser_cred() directly from linux_sethostname(), as it just
wraps userland_sysctl(), which performs necessary privilege checks as
part of its normal operation.

MFC after: 1 week

17 years agoSame as previous change, the user provided priority should be reversed
David Xu [Fri, 25 Aug 2006 10:05:30 +0000 (10:05 +0000)]
Same as previous change, the user provided priority should be reversed
too.

17 years ago- Fix options order.
Ruslan Ermilov [Fri, 25 Aug 2006 09:58:13 +0000 (09:58 +0000)]
- Fix options order.
- Touch manpage's document date.

17 years agoRemove a stray -a option that probably sneaked in from julian's
Ruslan Ermilov [Fri, 25 Aug 2006 09:42:16 +0000 (09:42 +0000)]
Remove a stray -a option that probably sneaked in from julian's
attempt to enter append mode twice in vi(1).  :-)

17 years agoProperly initialize and destroy the RAID lock.
Søren Schmidt [Fri, 25 Aug 2006 09:33:56 +0000 (09:33 +0000)]
Properly initialize and destroy the RAID lock.
Also dont mess with RAID's thats not attached yet and avoid panic.

17 years agoo Fix style(9) for previous.
Maxim Konovalov [Fri, 25 Aug 2006 09:14:23 +0000 (09:14 +0000)]
o Fix style(9) for previous.

17 years agoFix spelling.
Andrew Thompson [Fri, 25 Aug 2006 08:25:35 +0000 (08:25 +0000)]
Fix spelling.

17 years agoAdd login.conf checking to periodic security scripts. If the login.conf file
Tom Rhodes [Fri, 25 Aug 2006 07:34:36 +0000 (07:34 +0000)]
Add login.conf checking to periodic security scripts.  If the login.conf file
is not UID/GID 0, limits will be ignored and a strange error sent to auth.log.

Head nod: ru, rwatson

17 years agoRemove $P4$ from this file; other then temporarily P4-local work in
Robert Watson [Fri, 25 Aug 2006 07:30:23 +0000 (07:30 +0000)]
Remove $P4$ from this file; other then temporarily P4-local work in
progress the kernel audit code in CVS is considered authoritative.
This will ease $P4$-related merging issues during the CVS loopback.

Obtained from: TrustedBSD Project

17 years agoInitialize kg_base_user_pri.
David Xu [Fri, 25 Aug 2006 06:29:16 +0000 (06:29 +0000)]
Initialize kg_base_user_pri.

17 years agoAdd user priority loaning code to support priority propagation for
David Xu [Fri, 25 Aug 2006 06:12:53 +0000 (06:12 +0000)]
Add user priority loaning code to support priority propagation for
1:1 threading's POSIX priority mutexes, the code is no-op unless
priority-aware umtx code is committed.

17 years agoA bunch of fixes from NetBSD:
Maxim Konovalov [Fri, 25 Aug 2006 05:46:47 +0000 (05:46 +0000)]
A bunch of fixes from NetBSD:

o Restore owner/group/mode/atime/mtime of symbolic links, rev. 1.30.
o Extract file flags of symbolic link, rev. 1.42.
o Call getfile() before altering file attributes.
  Open file with mode 0600 instead of 0666 so that file won't remain
  group or world readable/writable even if getfile() terminated.
  Move skipfile() before altering file attributes in IF{CHR,BLK} and
  IFIFO case for symmetry, rev. 1.32.
o Use file mode 0600 when creating special file or fifo, revs. 1.33, 1.34.

o Remove redundant -N check.

PR: bin/101660
Submitted by: Andrey V. Elsukov
Obtained from: NetBSD, enami@netbsd
MFC after: 6 weeks

17 years agoThe return value from vm_pageq_add_new_page() is not used. Eliminate it.
Alan Cox [Fri, 25 Aug 2006 04:36:19 +0000 (04:36 +0000)]
The return value from vm_pageq_add_new_page() is not used.  Eliminate it.

17 years agoAdd member kg_base_user_pri and flag TDF_UBORROWING, they will be used
David Xu [Fri, 25 Aug 2006 03:15:27 +0000 (03:15 +0000)]
Add member kg_base_user_pri and flag TDF_UBORROWING, they will be used
to support userland priority propagation for 1:1 threading.

17 years agoSend more Alpha bits to the bin.
Tom Rhodes [Fri, 25 Aug 2006 00:36:59 +0000 (00:36 +0000)]
Send more Alpha bits to the bin.

17 years agoRemove file that snuck in accidentally in Marcel's gdb commit.
Peter Grehan [Fri, 25 Aug 2006 00:21:48 +0000 (00:21 +0000)]
Remove file that snuck in accidentally in Marcel's gdb commit.

17 years agoFinally bring it support for the i80219 XScale processor.
Olivier Houchard [Thu, 24 Aug 2006 23:51:28 +0000 (23:51 +0000)]
Finally bring it support for the i80219 XScale processor.

Submitted by: Max M. Boyarov <m.boyarov bsd by>

17 years agoUse ELFDATA2MSB if we're building big endian.
Olivier Houchard [Thu, 24 Aug 2006 23:00:03 +0000 (23:00 +0000)]
Use ELFDATA2MSB if we're building big endian.

Noticed by: Oleksandr Tymoshenko <gonzo freebsd org>

17 years agoRemove the DPMS code in creator_blank_display(), as it causes some
Marius Strobl [Thu, 24 Aug 2006 22:00:24 +0000 (22:00 +0000)]
Remove the DPMS code in creator_blank_display(), as it causes some
LCDs to blink in the V_DISPLAY_ON case, at least in combination with
some 13W3-VGA-adaptors (what's exactly going on is unclear though,
as it happens when all of H-sync, V-sync and video output are enabled
and not touching the sync bits from the preset fixes it). Thus
creator_blank_display() now is reduced to turning the video output
on/off.
Although that DPMS code did what the XFree86/Xorg sunffb(4x) does,
it was questionable in the first place, as both implementations
also turn(ed) off the video output on standby and suspend, thus most
likely causing the monitor to turn off instead of entering standby
or suspend as intended (at least my monitors don't).

Reported and tested by: Patrick Reich
MFC after: 3 days

17 years agoAdd initial support for kgdb(1) on PowerPC.
Marcel Moolenaar [Thu, 24 Aug 2006 21:53:49 +0000 (21:53 +0000)]
Add initial support for kgdb(1) on PowerPC.

17 years agoAdd skeletal support for GDB. In particular gdb_cpu_getreg() needs
Marcel Moolenaar [Thu, 24 Aug 2006 21:52:11 +0000 (21:52 +0000)]
Add skeletal support for GDB. In particular gdb_cpu_getreg() needs
implementing to make GDB support usable.

17 years agoFix typo in a comment: DEFINE_CLASSx => DEFINE_CLASS_x.
Roman Kurakin [Thu, 24 Aug 2006 21:09:39 +0000 (21:09 +0000)]
Fix typo in a comment: DEFINE_CLASSx => DEFINE_CLASS_x.

MFC after: 1 week

17 years agoAdd an option to allow copying of a hierarchy while linking he regular files.
Julian Elischer [Thu, 24 Aug 2006 20:45:38 +0000 (20:45 +0000)]
Add an option to allow copying of a hierarchy while linking he regular files.
Bikeshedded to death on: hackers
Submitted by:andersonatcenttech.com
MFC in: 1 month

17 years agoFix another fallout from the IF_LLADDR() type change.
Ruslan Ermilov [Thu, 24 Aug 2006 19:50:00 +0000 (19:50 +0000)]
Fix another fallout from the IF_LLADDR() type change.

Spotted by: mwlucas

17 years agoFix a bug introduced with rev. 1.204; in vfs_donmount() use
Marius Strobl [Thu, 24 Aug 2006 18:52:28 +0000 (18:52 +0000)]
Fix a bug introduced with rev. 1.204; in vfs_donmount() use
copyout(9) instead of copystr(9) for copying the errmsg from
kernel- to user-space. This fixes a panic on sparc64 when
using the nmount(2)-converted mountd(8).
While at it, use bcopy(3) instead of strncpy(3) in the kernel-
to kernel-space case for consistency with vfs_buildopts() and
between kernel- to user-space and kernel- to kernel-space case.

17 years agoDon't read in /etc/src.conf when building ports.
David E. O'Brien [Thu, 24 Aug 2006 18:04:49 +0000 (18:04 +0000)]
Don't read in /etc/src.conf when building ports.

Reported by: obrien
Submitted by: ru

17 years agoAdd kqueue support to audit pipe pseudo-devices.
Robert Watson [Thu, 24 Aug 2006 17:42:38 +0000 (17:42 +0000)]
Add kqueue support to audit pipe pseudo-devices.

Obtained from: TrustedBSD Project

17 years ago- add note about IPSEC_FILTERGIF to fast_ipsec(4) and let the users know
Daniel Gerzo [Thu, 24 Aug 2006 17:07:19 +0000 (17:07 +0000)]
- add note about IPSEC_FILTERGIF to fast_ipsec(4) and let the users know
  that it is not possible to use Fast IPsec in conjuction with KAME IPsec
- add available kernel options to ipsec(4)
- add reference for fast_ipsec(4) to ipsec(4)

Reviewed by: trhodes (mentor), keramida (mentor)
Approved by: keramida (mentor)

17 years agoAlways make obj when building the libraries. This never hurts, and
Warner Losh [Thu, 24 Aug 2006 17:02:26 +0000 (17:02 +0000)]
Always make obj when building the libraries.  This never hurts, and
helps some cross-architecture building tool installation patches that
I'm developing.

17 years agoDefine mtu as u_int16_t not as int. This should fix problem with rfcomm
Maksim Yevmenkin [Thu, 24 Aug 2006 16:51:02 +0000 (16:51 +0000)]
Define mtu as u_int16_t not as int. This should fix problem with rfcomm
on sparc64.

Reported by: Andrew Belashov <bel at orel dot ru>
Tested by: Andrew Belashov <bel at orel dot ru>
MFC after: 3 days

17 years agoBuild gdb on PowerPC.
Marcel Moolenaar [Thu, 24 Aug 2006 15:46:29 +0000 (15:46 +0000)]
Build gdb on PowerPC.