]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agoAdd "IDS" command to freebsd-update. This was present in the original
cperciva [Sat, 2 Aug 2008 00:09:41 +0000 (00:09 +0000)]
Add "IDS" command to freebsd-update.  This was present in the original
version of freebsd-update, but I took it out when I rewrote everything
and added FreeBSD Update to the base system because I didn't think it
was useful.  It turns out that quite a few people liked it and wanted
it back.

Requested by: Royce Williams + others
MFC after: 2 weeks

16 years agoSlight wordsmithing. prompted by danger@
julian [Fri, 1 Aug 2008 23:31:28 +0000 (23:31 +0000)]
Slight wordsmithing.  prompted by danger@

16 years agoDocument the use of the tablearg keyword together with the skipto command.
julian [Fri, 1 Aug 2008 22:34:01 +0000 (22:34 +0000)]
Document the use of the tablearg keyword together with the skipto command.

16 years agoThe IPFW code accepts the use of the tablearg keyword along with the skipto
julian [Fri, 1 Aug 2008 22:21:03 +0000 (22:21 +0000)]
The IPFW code accepts the use of the tablearg keyword along with the skipto
keyword. But it doesn't work. Two options.. make it no longer accept it,
or actually make it work.. I chose the 2nd..

Allow the tablearg to be used to specify a skipto destination.

This is actually a very powerful construct if used correctly, or a sink
of cpu cycles if used badly.

changes t teh man page will follow.

16 years agoMake "1000baseT" the description and "1000baseTX" the alias for
antoine [Fri, 1 Aug 2008 22:13:39 +0000 (22:13 +0000)]
Make "1000baseT" the description and "1000baseTX" the alias for
IFM_1000_T instead of the reverse.  It is possible FreeBSD doesn't
even support 1000baseTX.
This changes ifconfig(8) output.

Requested by: gavin@ and bms@
See also: http://docs.freebsd.org/cgi/mid.cgi?20050307191901.H32508

16 years agoRemove trailing ';' in BPFD_LOCK_ASSERT macro.
antoine [Fri, 1 Aug 2008 22:08:14 +0000 (22:08 +0000)]
Remove trailing ';' in BPFD_LOCK_ASSERT macro.

MFC after: 1 month
X-MFC-to: stable/7, stable/6 has it right

16 years agoTweak the support for using ldd on 32-bit objects a bit further.
jhb [Fri, 1 Aug 2008 21:52:41 +0000 (21:52 +0000)]
Tweak the support for using ldd on 32-bit objects a bit further.
Specifically, build a 32-bit /usr/bin/ldd32 on amd64 which handles 32-bit
objects.  Since it is a 32-bit binary, it can fork a child process which
can dlopen() a 32-bit shared library.  The current 32-bit support in ldd
can't do this because it does the dlopen() from a 64-bit process.  In order
to preserve an intuitive interface for users, the ldd binary automatically
execs /usr/bin/ldd32 for 32-bit objects.  The end result is that ldd on
amd64 now transparently handles 32-bit shared libraries in addition to
32-bit binaries.

Submitted by: ps (indirectly)

16 years agoAnnotate why we do not call BPF_CHECK_DIRECTION() in this tapping routine.
csjp [Fri, 1 Aug 2008 21:38:46 +0000 (21:38 +0000)]
Annotate why we do not call BPF_CHECK_DIRECTION() in this tapping routine.
There is no way for the caller to tell us which direction this packet is
going.  With the bpf_mtap{2} routines, we can check the interface pointer.

MFC after: 2 weeks

16 years agoUse the new bus device_shutdown hook for performing the shutdown action
jhb [Fri, 1 Aug 2008 21:33:07 +0000 (21:33 +0000)]
Use the new bus device_shutdown hook for performing the shutdown action
on the ee16 parts rather than explicitly registering an event handler
that wasn't being torn down on detach.

16 years ago- Use an sx lock to serialize writes since they update the checksum.
jhb [Fri, 1 Aug 2008 20:39:18 +0000 (20:39 +0000)]
- Use an sx lock to serialize writes since they update the checksum.
- Remove D_NEEDGIANT as the rtc drivers already have their own locks, so
  this doesn't need Giant.

MFC after: 1 week

16 years agoAdjust comment. This stack is only used for booting now and not as an
jhb [Fri, 1 Aug 2008 20:10:47 +0000 (20:10 +0000)]
Adjust comment.  This stack is only used for booting now and not as an
idle stack.

16 years agoFix whitespace.
emaste [Fri, 1 Aug 2008 15:13:56 +0000 (15:13 +0000)]
Fix whitespace.

16 years agoChar is unsigned on arm, so is not suitable to store the return value of
cognet [Fri, 1 Aug 2008 13:12:06 +0000 (13:12 +0000)]
Char is unsigned on arm, so is not suitable to store the return value of
getopt(). Use an int instead.

Submitted by: Matthew Luckie
MFC after: 3 days

16 years agoEnable the support for G33/Q35/Q33 now that both the G33 and Q35 have been
jhb [Fri, 1 Aug 2008 13:11:23 +0000 (13:11 +0000)]
Enable the support for G33/Q35/Q33 now that both the G33 and Q35 have been
tested:

PR: amd64/126090
MFC after: 1 week

16 years agoRemove further trailing white space.
rwatson [Fri, 1 Aug 2008 09:41:45 +0000 (09:41 +0000)]
Remove further trailing white space.

16 years agoFix empty mailto (-m "") handling: somehow I missed all checks but the first,
marck [Fri, 1 Aug 2008 08:01:33 +0000 (08:01 +0000)]
Fix empty mailto (-m "") handling: somehow I missed all checks but the first,
hence output would be written to the wrong filehandle.

Submitted by: reg
Approved by: yar (implicit)
MFC after: ASAP
Pointy hat to: marck

16 years agoWhen using SRV records the protocols and services files need to be in the
dougb [Fri, 1 Aug 2008 06:11:33 +0000 (06:11 +0000)]
When using SRV records the protocols and services files need to be in the
chroot /etc directory.

PR: conf/121101
Submitted by: Stefan `Sec` Zehl <sec@42.org>

16 years agoAdd the -c option for named_flags (still commented out) that is
dougb [Fri, 1 Aug 2008 05:15:54 +0000 (05:15 +0000)]
Add the -c option for named_flags (still commented out) that is
relevant for ports users, and change the comment to match.

While I'm here fix the capitalization of the named_program comment.

16 years agoEnhance pmap_change_attr() with the ability to demote 1GB page mappings.
alc [Fri, 1 Aug 2008 04:55:38 +0000 (04:55 +0000)]
Enhance pmap_change_attr() with the ability to demote 1GB page mappings.

16 years agoUpgrade to OpenSSH 5.1p1.
des [Fri, 1 Aug 2008 02:48:36 +0000 (02:48 +0000)]
Upgrade to OpenSSH 5.1p1.

I have worked hard to reduce diffs against the vendor branch.  One
notable change in that respect is that we no longer prefer DSA over
RSA - the reasons for doing so went away years ago.  This may cause
some surprises, as ssh will warn about unknown host keys even for
hosts whose keys haven't changed.

MFC after: 6 weeks

16 years agoRemove svn:keywords except on files that need it. This makes diffs
des [Fri, 1 Aug 2008 02:06:48 +0000 (02:06 +0000)]
Remove svn:keywords except on files that need it.  This makes diffs
against the vendor branch much more readable.

16 years agoAnother file with no local changes.
des [Fri, 1 Aug 2008 01:50:55 +0000 (01:50 +0000)]
Another file with no local changes.

"This time for sure!"

16 years agoAnother file with no local changes.
des [Fri, 1 Aug 2008 01:48:33 +0000 (01:48 +0000)]
Another file with no local changes.

16 years agoAnother four files without local changes. This is driving me nuts -
des [Fri, 1 Aug 2008 01:45:56 +0000 (01:45 +0000)]
Another four files without local changes.  This is driving me nuts -
every time I think I got them all, another one pops up.

16 years agoYet another file with no local changes.
des [Fri, 1 Aug 2008 01:38:50 +0000 (01:38 +0000)]
Yet another file with no local changes.

16 years agoAccidentally mangled this one in the previous commit.
des [Fri, 1 Aug 2008 01:38:24 +0000 (01:38 +0000)]
Accidentally mangled this one in the previous commit.

16 years agoMore files which no longer have any local changes.
des [Fri, 1 Aug 2008 01:32:56 +0000 (01:32 +0000)]
More files which no longer have any local changes.

16 years agoThese two files have no local patches except to prevent expansion of the
des [Fri, 1 Aug 2008 01:30:26 +0000 (01:30 +0000)]
These two files have no local patches except to prevent expansion of the
original $FreeBSD$ keywords.  Revert those changes, and simply disable
keyword expansion.

16 years agoadd missing capability bits
sam [Fri, 1 Aug 2008 01:27:52 +0000 (01:27 +0000)]
add missing capability bits

16 years agoLast remains of old OPIE patch
des [Fri, 1 Aug 2008 01:24:42 +0000 (01:24 +0000)]
Last remains of old OPIE patch

16 years agoFix some problems with asinf(), acosf(), atanf(), and atan2f():
das [Fri, 1 Aug 2008 01:24:25 +0000 (01:24 +0000)]
Fix some problems with asinf(), acosf(), atanf(), and atan2f():

- Adjust several constants for float precision. Some thresholds
  that were appropriate for double precision were never changed
  when these routines were converted to float precision. This
  has an impact on performance but not accuracy. (Submitted by bde.)

- Reduce the degrees of the polynomials used. A smaller degree
  suffices for float precision.

- In asinf(), use double arithmetic in part of the calculation to
  avoid a corner case and some complicated arithmetic involving a
  division and some buggy constants. This improves performance and
  accuracy.

Max error (ulps):
         asinf  acosf  atanf
before   0.925  0.782  0.852
after    0.743  0.804  0.852

As bde points out, it's cheaper for asin*() and acos*() to use
polynomials instead of rational functions, but that's a task for
another day.

16 years agoIn function pthread_condattr_getpshared, store result correctly.
davidxu [Fri, 1 Aug 2008 01:21:49 +0000 (01:21 +0000)]
In function pthread_condattr_getpshared, store result correctly.

PR: kern/126128

16 years agoWe no longer have any local changes here.
des [Fri, 1 Aug 2008 01:19:13 +0000 (01:19 +0000)]
We no longer have any local changes here.

16 years agoConsistently set svn:eol-style.
des [Fri, 1 Aug 2008 01:13:41 +0000 (01:13 +0000)]
Consistently set svn:eol-style.

16 years agoTag expansion is no longer needed (svn handles them correctly).
des [Fri, 1 Aug 2008 01:05:27 +0000 (01:05 +0000)]
Tag expansion is no longer needed (svn handles them correctly).
Add svn command to diff against vendor branch.

16 years agoThis is no longer needed.
des [Fri, 1 Aug 2008 01:04:52 +0000 (01:04 +0000)]
This is no longer needed.

16 years agoCleanup.
des [Fri, 1 Aug 2008 01:00:34 +0000 (01:00 +0000)]
Cleanup.

16 years agoImplement ratelimiting for debug messages. For now, allow at most
emax [Fri, 1 Aug 2008 00:36:43 +0000 (00:36 +0000)]
Implement ratelimiting for debug messages. For now, allow at most
one message per second. In the future might add a sysctl knob for
each socket family to fine tune this.

MFC after: 1 week

16 years agoUgh. Set svn:mergeinfo correctly.
des [Fri, 1 Aug 2008 00:34:37 +0000 (00:34 +0000)]
Ugh.  Set svn:mergeinfo correctly.

16 years agoCatch up with reality.
des [Fri, 1 Aug 2008 00:28:50 +0000 (00:28 +0000)]
Catch up with reality.

16 years agoRevert an old hack I put in to replace S/Key with OPIE. We haven't used
des [Fri, 1 Aug 2008 00:27:48 +0000 (00:27 +0000)]
Revert an old hack I put in to replace S/Key with OPIE.  We haven't used
that code in ages - we use pam_opie(8) instead - so this is a NOP.

16 years agoIncrease maximum input queue size limit for raw Bluetooth HCI sockets.
emax [Fri, 1 Aug 2008 00:16:40 +0000 (00:16 +0000)]
Increase maximum input queue size limit for raw Bluetooth HCI sockets.

MFC after: 3 days

16 years agoFix locking bug, i.e. lock "wildcard" matched pcb before return.
emax [Fri, 1 Aug 2008 00:13:32 +0000 (00:13 +0000)]
Fix locking bug, i.e. lock "wildcard" matched pcb before return.

16 years agoAdd missing #include for strlen()
des [Thu, 31 Jul 2008 23:33:26 +0000 (23:33 +0000)]
Add missing #include for strlen()

16 years agounset TARGET_CPUTYPE and TARGET_BIG_ENDIAN when building the kernel
sam [Thu, 31 Jul 2008 23:04:00 +0000 (23:04 +0000)]
unset TARGET_CPUTYPE and TARGET_BIG_ENDIAN when building the kernel
to silence compiler complaints when cross-building

Reviewed by: phk
MFC after: 2 weeks

16 years agoEnhance pmap_change_attr(). Specifically, avoid 2MB page demotions, cache
alc [Thu, 31 Jul 2008 22:45:28 +0000 (22:45 +0000)]
Enhance pmap_change_attr().  Specifically, avoid 2MB page demotions, cache
mode changes, and cache and TLB invalidation when some or all of the
specified range is already mapped with the specified cache mode.

Submitted by: Magesh Dhasayyan

16 years agoAdd some tests for acos*(), asin*(), atan*(), and atan2*().
das [Thu, 31 Jul 2008 22:43:38 +0000 (22:43 +0000)]
Add some tests for acos*(), asin*(), atan*(), and atan2*().

16 years agoAdd implementations of acosl(), asinl(), atanl(), atan2l(),
das [Thu, 31 Jul 2008 22:41:26 +0000 (22:41 +0000)]
Add implementations of acosl(), asinl(), atanl(), atan2l(),
and cargl().

Reviewed by: bde
sparc64 testing resources from: remko

16 years agoOops, restore the recent changes to make startup messages quieter.
jhb [Thu, 31 Jul 2008 22:13:14 +0000 (22:13 +0000)]
Oops, restore the recent changes to make startup messages quieter.

16 years agoParse sysctl settings from /etc/sysctl.conf.local after /etc/sysctl.conf
jhb [Thu, 31 Jul 2008 21:57:35 +0000 (21:57 +0000)]
Parse sysctl settings from /etc/sysctl.conf.local after /etc/sysctl.conf
if it exists.  This mirrors similar behavior for /boot/loader.conf and
/etc/rc.conf.

Obtained from: Yahoo!
MFC after: 1 week

16 years agoIn mac_bsdextended's auditctl and acct policy access control checks,
rwatson [Thu, 31 Jul 2008 20:49:12 +0000 (20:49 +0000)]
In mac_bsdextended's auditctl and acct policy access control checks,
return success if the passed vnode pointer is NULL (rather than
panicking).  This can occur if either audit or accounting are
disabled while the policy is running.

Since the swapoff control has no real relevance to this policy,
which is concerned about intent to write rather than water under the
bridge, remove it.

PR:             kern/126100
Reported by:    Alan Amesbury <amesbury at umn dot edu>
MFC after:      3 days

16 years agoremove socketvar.h, add more selective includes
kmacy [Thu, 31 Jul 2008 20:28:58 +0000 (20:28 +0000)]
remove socketvar.h, add more selective includes

16 years agomove sockbuf locking macros in to sockbuf.h
kmacy [Thu, 31 Jul 2008 20:27:50 +0000 (20:27 +0000)]
move sockbuf locking macros in to sockbuf.h

16 years agoAdd thr_pread_{int,long,ptr} and thr_pwrite_{int,long,ptr} to
marcel [Thu, 31 Jul 2008 20:25:52 +0000 (20:25 +0000)]
Add thr_pread_{int,long,ptr} and thr_pwrite_{int,long,ptr} to
help abstract the sizes and endianness of the primary types of
the target.  These currently use the native characteristics.

16 years agoSet WARNS=1.
das [Thu, 31 Jul 2008 20:11:37 +0000 (20:11 +0000)]
Set WARNS=1.

I believe I've committed all the bits necessary to make this compile
on all supported architectures. :crosses fingers:

16 years agoThe high part of the mantissa is 64 bits on sparc64.
das [Thu, 31 Jul 2008 20:09:47 +0000 (20:09 +0000)]
The high part of the mantissa is 64 bits on sparc64.

16 years agoAs in other parts of libm, mark a few constants as volatile to prevent
das [Thu, 31 Jul 2008 19:57:50 +0000 (19:57 +0000)]
As in other parts of libm, mark a few constants as volatile to prevent
spurious optimizations. gcc doesn't support FENV_ACCESS, so when it
folds constants, it assumes that the rounding mode is always the
default and floating point exceptions never matter.

16 years agoTry to make this code slightly less painful to read.
des [Thu, 31 Jul 2008 17:15:21 +0000 (17:15 +0000)]
Try to make this code slightly less painful to read.

16 years agoCurrently, BSM audit pathname token generation for chrooted or jailed
csjp [Thu, 31 Jul 2008 16:57:41 +0000 (16:57 +0000)]
Currently, BSM audit pathname token generation for chrooted or jailed
processes are not producing absolute pathname tokens.  It is required
that audited pathnames are generated relative to the global root mount
point.  This modification changes our implementation of audit_canon_path(9)
and introduces a new function: vn_fullpath_global(9) which performs a
vnode -> pathname translation relative to the global mount point based
on the contents of the name cache.  Much like vn_fullpath,
vn_fullpath_global is a wrapper function which called vn_fullpath1.

Further, the string parsing routines have been converted to use the
sbuf(9) framework.  This change also removes the conditional acquisition
of Giant, since the vn_fullpath1 method will not dip into file system
dependent code.

The vnode locking was modified to use vhold()/vdrop() instead the vref()
and vrele().  This will modify the hold count instead of modifying the
user count.  This makes more sense since it's the kernel that requires
the reference to the vnode.  This also makes sure that the vnode does not
get recycled we hold the reference to it. [1]

Discussed with: rwatson
Reviewed by: kib [1]
MFC after: 2 weeks

16 years agoCleanup for WARNS 3.
marcel [Thu, 31 Jul 2008 16:26:58 +0000 (16:26 +0000)]
Cleanup for WARNS 3.

16 years agoMFp4 (//depot/projects/tcpecn/):
rpaulo [Thu, 31 Jul 2008 15:10:09 +0000 (15:10 +0000)]
MFp4 (//depot/projects/tcpecn/):

  TCP ECN support. Merge of my GSoC 2006 work for NetBSD.
  TCP ECN is defined in RFC 3168.

Partly reviewed by: dwmalone, silby
Obtained from: NetBSD

16 years agoAdds support for the SCTP_PORT_REUSE option
rrs [Thu, 31 Jul 2008 11:08:30 +0000 (11:08 +0000)]
Adds support for the SCTP_PORT_REUSE option
Fixes a refcount bug found in the process

Obtained from: With the help of Michael Tuexen

16 years agoFurther synchronization of copyrights, licenses, white space, etc from
rwatson [Thu, 31 Jul 2008 09:54:35 +0000 (09:54 +0000)]
Further synchronization of copyrights, licenses, white space, etc from
Apple and from the OpenBSM vendor tree.

Obtained from: Apple Inc., TrustedBSD Project
MFC after: 3 days

16 years agoCleanup for WARNS 2.
marcel [Thu, 31 Jul 2008 05:25:52 +0000 (05:25 +0000)]
Cleanup for WARNS 2.

16 years agoEliminate recomputation of the PDE by pmap_pde_attr().
alc [Thu, 31 Jul 2008 04:42:42 +0000 (04:42 +0000)]
Eliminate recomputation of the PDE by pmap_pde_attr().

16 years agoFix LINT
emax [Thu, 31 Jul 2008 03:51:53 +0000 (03:51 +0000)]
Fix LINT

MFC after: 3 months

16 years agoData type fix
jfv [Thu, 31 Jul 2008 02:22:53 +0000 (02:22 +0000)]
Data type fix

16 years agoUnbreak the build by creating opt_nfs.h
ps [Thu, 31 Jul 2008 02:13:36 +0000 (02:13 +0000)]
Unbreak the build by creating opt_nfs.h

16 years agoUnbreak the build by including sys/socketvar.h
ps [Thu, 31 Jul 2008 01:52:04 +0000 (01:52 +0000)]
Unbreak the build by including sys/socketvar.h

16 years agoAdd a -q[uiet] flag for scripts.
obrien [Thu, 31 Jul 2008 00:55:29 +0000 (00:55 +0000)]
Add a -q[uiet] flag for scripts.

16 years agoInclude netinet/tcp_lro.h, unbreak the build
ps [Wed, 30 Jul 2008 23:54:23 +0000 (23:54 +0000)]
Include netinet/tcp_lro.h, unbreak the build

16 years agoHook up Bluetooth SCO sockets code to the build
emax [Wed, 30 Jul 2008 22:42:17 +0000 (22:42 +0000)]
Hook up Bluetooth SCO sockets code to the build

MFC after: 3 months

16 years agoIntroduce support for Bluetooth SCO sockets. This is based on older
emax [Wed, 30 Jul 2008 22:41:23 +0000 (22:41 +0000)]
Introduce support for Bluetooth SCO sockets. This is based on older
code that was revisted.

MFC after: 3 months

16 years agoAdd igb driver to the default kernel
jfv [Wed, 30 Jul 2008 22:30:49 +0000 (22:30 +0000)]
Add igb driver to the default kernel

16 years agoAdd igb to the default kernel
jfv [Wed, 30 Jul 2008 22:27:38 +0000 (22:27 +0000)]
Add igb to the default kernel

MFC after:ASAP

16 years agoChange Makefile to reflect new directory structure
jfv [Wed, 30 Jul 2008 22:06:38 +0000 (22:06 +0000)]
Change Makefile to reflect new directory structure

MFC after:ASAP

16 years agoChange to build module with new directory tree
jfv [Wed, 30 Jul 2008 22:05:06 +0000 (22:05 +0000)]
Change to build module with new directory tree

MFC after:ASAP

16 years agoAlter kernel build to work with new dev/e1000 structure.
jfv [Wed, 30 Jul 2008 22:01:32 +0000 (22:01 +0000)]
Alter kernel build to work with new dev/e1000 structure.
This makes both em and igb, or either alone build and
work in the static kernel.

MFC after:ASAP

16 years agoMerge of the source for igb and em into dev/e1000, this
jfv [Wed, 30 Jul 2008 21:56:53 +0000 (21:56 +0000)]
Merge of the source for igb and em into dev/e1000, this
proved to be necessary to make the static drivers work
in EITHER/OR or BOTH configurations. Modules will still
build in sys/modules/igb or em as before.

This also updates the igb driver for support for the 82576
adapter, adds shared code fixes, and etc....

MFC after: ASAP

16 years agoOne more whitespace nit.
jhb [Wed, 30 Jul 2008 21:23:32 +0000 (21:23 +0000)]
One more whitespace nit.

16 years agoAdd a regression test for r181017.
stefanf [Wed, 30 Jul 2008 21:18:52 +0000 (21:18 +0000)]
Add a regression test for r181017.

Submitted by: simon

16 years agoAdd POSIX -p flag to make(1).
ed [Wed, 30 Jul 2008 21:18:38 +0000 (21:18 +0000)]
Add POSIX -p flag to make(1).

This article [1] describes the -p flag for make(1):

Write to standard output the complete set of macro definitions and
target descriptions. The output format is unspecified.

We already support a similar flag (-d g1), but unlike -p, it still
executes commands. Our implementation just turns it into -d g1, but also
sets flag `printGraphOnly', which will cause make(1) to skip execution.

[1] http://www.opengroup.org/onlinepubs/009695399/utilities/make.html

Reviewed by: imp
PR: standards/99960

16 years agoA few more whitespace fixes.
jhb [Wed, 30 Jul 2008 21:18:08 +0000 (21:18 +0000)]
A few more whitespace fixes.

16 years agoIf the kernel has run out of metadata for swap, then explicitly panic()
jhb [Wed, 30 Jul 2008 21:12:15 +0000 (21:12 +0000)]
If the kernel has run out of metadata for swap, then explicitly panic()
instead of emitting a warning before deadlocking.

MFC after: 1 month

16 years agoWhitespace tweak.
jhb [Wed, 30 Jul 2008 21:07:56 +0000 (21:07 +0000)]
Whitespace tweak.

16 years agoPass the correct flags to expandarg() for NFROMFD and NTOFD. This fixes a
stefanf [Wed, 30 Jul 2008 21:07:04 +0000 (21:07 +0000)]
Pass the correct flags to expandarg() for NFROMFD and NTOFD.  This fixes a
segmentation fault when the argument expands to an empty string.

Reported by: simon
MFC after: 3 weeks

16 years agoTrim some noise from some #ifdef's. This had leaked into the compat32
jhb [Wed, 30 Jul 2008 21:01:51 +0000 (21:01 +0000)]
Trim some noise from some #ifdef's.  This had leaked into the compat32
support for bpf(4) due to hacks in the Y! tree for a truss32 binary
(since superseded by native support for 32-bit binaries in truss itself).

MFC after: 1 week

16 years agofix includes for post sockbuf re-factor
kmacy [Wed, 30 Jul 2008 20:08:34 +0000 (20:08 +0000)]
fix includes for post sockbuf re-factor

16 years agoUse single byte read and write operations to toggle the BIOS and OS
jhb [Wed, 30 Jul 2008 19:16:53 +0000 (19:16 +0000)]
Use single byte read and write operations to toggle the BIOS and OS
semaphores rather than 4 byte operations.

MFC after: 1 month

16 years agoThe original adduser/rmuser scripts in Perl used to modify the PATH
jhb [Wed, 30 Jul 2008 18:37:21 +0000 (18:37 +0000)]
The original adduser/rmuser scripts in Perl used to modify the PATH
setting.  When the scripts were converted to Bourne shell, this was
removed.  The adduser script was changed to use an explicit path for
the pw(8) command so that /usr/sbin did not have to be in the user's
PATH.  The rmuser script continued to assume that /usr/sbin was in the
user's path, however.  This fixes the rmuser script to use an explicit
path for pw(8) similar to adduser.

MFC after: 2 weeks

16 years agoThe behaviour of the lockmgr going back at least to the 4.4BSD-Lite2 was
kib [Wed, 30 Jul 2008 18:16:06 +0000 (18:16 +0000)]
The behaviour of the lockmgr going back at least to the 4.4BSD-Lite2 was
to downgrade the exclusive lock to shared one when exclusive lock owner
requested shared lock. New lockmgr panics instead.

The vnode_pager_lock function requests shared lock on the vnode backing
the OBJT_VNODE, and can be called when the current thread already holds
an exlcusive lock on the vnode. For instance, it happens when handling
page fault from the VOP_WRITE() uiomove that writes to the file, with
the faulted in page fetched from the vm object backed by the same file.
We then get the situation described above.

Verify whether the vnode is already exclusively locked by the curthread
and request recursed exclusive vnode lock instead of shared, if true.

Reported by: gallatin
Discussed with: attilio

16 years agoThis updates the ixgbe driver to Intel internal version 1.4.7
jfv [Wed, 30 Jul 2008 18:15:18 +0000 (18:15 +0000)]
This updates the ixgbe driver to Intel internal version 1.4.7
Shared code changes, core driver fixes, vlan event/filter support

Also Kip Macy's fix to allow any number of queues, thanks Kip!

16 years agoSet all of the "optimum performance" PHY registers for the 15D parts as
jhb [Wed, 30 Jul 2008 17:28:49 +0000 (17:28 +0000)]
Set all of the "optimum performance" PHY registers for the 15D parts as
well as the 15C since it seems to be required in practice.  The Linux
natsemi.c driver mostly does this as well.

PR: kern/112179
Submitted by: Mark Willson  mark - hydrus org uk
MFC after: 1 week

16 years agoSIS_SETBIT() already does a read/modify/write operation, so there isn't
jhb [Wed, 30 Jul 2008 17:21:32 +0000 (17:21 +0000)]
SIS_SETBIT() already does a read/modify/write operation, so there isn't
a reason to read the register twice.

MFC after: 1 week

16 years agocorrect description of how to clear a desired channel
sam [Wed, 30 Jul 2008 16:22:52 +0000 (16:22 +0000)]
correct description of how to clear a desired channel

Pointed out by: Chris Buechler
MFC after: 1 week

16 years agoAccording to a proposed Norwegian standard for alphabetization of which
des [Wed, 30 Jul 2008 15:37:13 +0000 (15:37 +0000)]
According to a proposed Norwegian standard for alphabetization of which
I've obtained a draft, <u:> is indeed equivalent to u (to my surprise),
and <th> sorts immediately after z.

The correct ordering is algorithmic (based on the EOR) and can not be
accurately represented as a table.

16 years agoRemove the use of lbolt from the VFS syncer.
ed [Wed, 30 Jul 2008 12:39:18 +0000 (12:39 +0000)]
Remove the use of lbolt from the VFS syncer.

It seems we only use `lbolt' inside the VFS syncer and the TTY layer
now.  Because I'm planning to replace the TTY layer next month, there's
no reason to keep `lbolt' if it's only used in a single thread inside
the kernel.

Because the syncer code wanted to wake up the syncer thread before the
timeout, it called sleepq_remove(). Because we now just use a condvar(9)
with a timeout value of `hz', we can wake it up using cv_broadcast()
without waking up any unrelated threads.

Reviewed by: phk

16 years agoDon't make subr_clist.c depend on the TTY layer.
ed [Wed, 30 Jul 2008 12:32:42 +0000 (12:32 +0000)]
Don't make subr_clist.c depend on the TTY layer.

After the import of the new TTY layer, the TTY_QUOTE definition will not
be present anymore. To make sure clists will still work as expected,
introduce an internal definition called QUOTEMASK.

Maybe we can decide to remove the quote bits entirely, but we still have
to look into this. There may be drivers that still use the quote bits.

Obtained from: //depot/projects/mpsafetty

16 years agoUpdate routed to use the RFC 3678 protocol-independent multicast API. Use
phk [Wed, 30 Jul 2008 11:56:15 +0000 (11:56 +0000)]
Update routed to use the RFC 3678 protocol-independent multicast API. Use
IP_MULTICAST_IF with struct ip_mreqn (obtained from Linux) to tell the
stack which interface index to use for sending IPv4 datagrams.

Submitted by: bms
Tested by: phk

16 years agoBring back the save/restore of the %ds, %es, %fs and %gs registers for
kib [Wed, 30 Jul 2008 11:30:55 +0000 (11:30 +0000)]
Bring back the save/restore of the %ds, %es, %fs and %gs registers for
the 32bit images on amd64.

Change the semantic of the PCB_32BIT pcb flag to request the context
switch code to operate on the segment registers. Its previous meaning
of saving or restoring the %gs base offset is assigned to the new
PCB_GS32BIT flag.

FreeBSD 32bit image activator sets the PCB_32BIT flag, while Linux 32bit
emulation sets PCB_32BIT | PCB_GS32BIT.

Reviewed by: peter
MFC after: 2 weeks