]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoFix support for Acer chips. The UDMA enable regs wasn't set properly.
Søren Schmidt [Tue, 25 Feb 2003 21:22:27 +0000 (21:22 +0000)]
Fix support for Acer chips. The UDMA enable regs wasn't set properly.

21 years agoMove most everything back to a MI NOTES, and use "nodevice" in MD NOTES
David E. O'Brien [Tue, 25 Feb 2003 20:59:23 +0000 (20:59 +0000)]
Move most everything back to a MI NOTES, and use "nodevice" in MD NOTES
Where needed.  Use 'sed' for now in place of "nooptions".  Add a sparc64
MD NOTES.

Reviewed by: arch@

21 years agoLimit the maximum I/O size to 8 pages for version 2 controllers. This
John Baldwin [Tue, 25 Feb 2003 20:47:22 +0000 (20:47 +0000)]
Limit the maximum I/O size to 8 pages for version 2 controllers.  This
fixes problems with some mlx(4) cards in Alpha machines.

Reviewed by: msmith (ages ago)

21 years agoFix a small fat-finger: s/mlxd/mxld/.
John Baldwin [Tue, 25 Feb 2003 20:35:56 +0000 (20:35 +0000)]
Fix a small fat-finger: s/mlxd/mxld/.

Reported by: Bob Willcox <bob@immure.com>

21 years ago- Added inlines pmap_is_current, pmap_is_alternate and pmap_set_alternate
Jake Burkholder [Tue, 25 Feb 2003 19:40:21 +0000 (19:40 +0000)]
- Added inlines pmap_is_current, pmap_is_alternate and pmap_set_alternate
  for testing and setting the current and alternate address spaces.
- Changed PTDpde and APTDpde to arrays to support multiple page directory
  pages.

ponsored by: DARPA, Network Associates Laboratories

21 years agoFixed CLEANFILES.
Ruslan Ermilov [Tue, 25 Feb 2003 15:41:49 +0000 (15:41 +0000)]
Fixed CLEANFILES.

Submitted by: cron

21 years agoConvert to new disk API.
Søren Schmidt [Tue, 25 Feb 2003 15:33:36 +0000 (15:33 +0000)]
Convert to new disk API.

Prodded by: phk

21 years agoAllocate struct ipx_ifaddrs with an initial reference count of 1, not 0.
Tim J. Robbins [Tue, 25 Feb 2003 15:10:23 +0000 (15:10 +0000)]
Allocate struct ipx_ifaddrs with an initial reference count of 1, not 0.
The wrong reference count was causing them to get freed too early and
have their contents scrambled.

21 years agoFix a long-standing bug where if the package being deleted had no
Dag-Erling Smørgrav [Tue, 25 Feb 2003 15:01:54 +0000 (15:01 +0000)]
Fix a long-standing bug where if the package being deleted had no
post-deinstall script, the variable intended to hold the name of that
script would be used uninitialized.  In some cases, fexists() would
succeed, causing pkg_delete to try to chmod +x it, then execute it,
resulting in bizarre error messages such as:

.//: Permission denied

This bug would normally only occur when multiple packages were
specified on the command line; otherwise post_script would be located
in a previously unused part of the stack, and implicitly (but quite
accidentally) initialized to all-zeros.

MFC after: 3 days

21 years agoProperly teardown the interrupt so we wont panic on detach.
Søren Schmidt [Tue, 25 Feb 2003 14:46:30 +0000 (14:46 +0000)]
Properly teardown the interrupt so we wont panic on detach.

21 years agoRemoved the ISA only hooks here, no longer needed..
Søren Schmidt [Tue, 25 Feb 2003 14:24:42 +0000 (14:24 +0000)]
Removed the ISA only hooks here, no longer needed..

21 years agoWhen generating a TCP response to a connection, not only test if the
Robert Watson [Tue, 25 Feb 2003 14:08:41 +0000 (14:08 +0000)]
When generating a TCP response to a connection, not only test if the
tcpcb is NULL, but also its connected inpcb, since we now allow
elements of a TCP connection to hang around after other state, such
as the socket, has been recycled.

Tested by: dcs
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoBump __FreeBSD_version for the d_mmap_t interface change.
Maxime Henrion [Tue, 25 Feb 2003 13:38:09 +0000 (13:38 +0000)]
Bump __FreeBSD_version for the d_mmap_t interface change.

21 years agoConvert one more d_mmap_t consumer I missed in my previous commit.
Maxime Henrion [Tue, 25 Feb 2003 13:30:50 +0000 (13:30 +0000)]
Convert one more d_mmap_t consumer I missed in my previous commit.

21 years agoFix support for the old CMD 646 chips, interrupt setup was wrong.
Søren Schmidt [Tue, 25 Feb 2003 12:56:37 +0000 (12:56 +0000)]
Fix support for the old CMD 646 chips, interrupt setup was wrong.

21 years agostyle(9): join lines.
Maxim Konovalov [Tue, 25 Feb 2003 11:53:11 +0000 (11:53 +0000)]
style(9): join lines.

21 years agoIp reassembly queue structure has ipq_nfrags now. Count a number of
Maxim Konovalov [Tue, 25 Feb 2003 11:49:01 +0000 (11:49 +0000)]
Ip reassembly queue structure has ipq_nfrags now.  Count a number of
dropped ip fragments precisely.

Reviewed by: silby

21 years agoRemove an unsafe KASSERT.
David Xu [Tue, 25 Feb 2003 11:23:17 +0000 (11:23 +0000)]
Remove an unsafe KASSERT.

21 years agoCatch up with change to kse_release syscall.
Julian Elischer [Tue, 25 Feb 2003 09:49:46 +0000 (09:49 +0000)]
Catch up with change to kse_release syscall.
The background info in this man page needs rewriting
in some parts since the last major changes
to the code, however it still accuratly reflects how to use the
API.

21 years ago - Properly handle the vnode interlock in nfs_fsync.
Jeff Roberson [Tue, 25 Feb 2003 08:50:21 +0000 (08:50 +0000)]
 - Properly handle the vnode interlock in nfs_fsync.

Reported by: phk

21 years ago - Add the missing NULL interlock argument to a recently added BUF_LOCK.
Jeff Roberson [Tue, 25 Feb 2003 08:23:11 +0000 (08:23 +0000)]
 - Add the missing NULL interlock argument to a recently added BUF_LOCK.

21 years agoRevert the probe code change.
Søren Schmidt [Tue, 25 Feb 2003 08:22:01 +0000 (08:22 +0000)]
Revert the probe code change.
It helped those fake devices probe delays to go away, but as a side
effect it also made some not so conforming CD/DVD drives go away...

21 years agoNO_GEOM cleanup:
Poul-Henning Kamp [Tue, 25 Feb 2003 08:02:57 +0000 (08:02 +0000)]
NO_GEOM cleanup:

Move to "struct disk *" centric API.
Add NULL check to a malloc(9) return value.

Ok'ed by: sos

21 years agoNO_GEOM cleanup:
Poul-Henning Kamp [Tue, 25 Feb 2003 07:41:51 +0000 (07:41 +0000)]
NO_GEOM cleanup:

Move to new "struct disk *" centric API.
Retire major #147 (twed)

21 years agoMore fixes.
Ruslan Ermilov [Tue, 25 Feb 2003 07:41:33 +0000 (07:41 +0000)]
More fixes.

Submitted by: Andy Farkas <andyf@speednet.com.au>
Reviewed by: das

21 years agoNO_GEOM cleanup:
Poul-Henning Kamp [Tue, 25 Feb 2003 07:35:41 +0000 (07:35 +0000)]
NO_GEOM cleanup:

Move to "struct disk*" centric API.
Retire major #131 (mlxd)

21 years agoRetire major #133 (amrd).
Poul-Henning Kamp [Tue, 25 Feb 2003 07:24:51 +0000 (07:24 +0000)]
Retire major #133 (amrd).

21 years agoFuse two #ifdefs with identical conditions.
Alan Cox [Tue, 25 Feb 2003 06:46:08 +0000 (06:46 +0000)]
Fuse two #ifdefs with identical conditions.

21 years agoPrevent large files from monopolizing the system buffers. Keep
Kirk McKusick [Tue, 25 Feb 2003 06:44:42 +0000 (06:44 +0000)]
Prevent large files from monopolizing the system buffers. Keep
track of the number of dirty buffers held by a vnode. When a
bdwrite is done on a buffer, check the existing number of dirty
buffers associated with its vnode. If the number rises above
vfs.dirtybufthresh (currently 90% of vfs.hidirtybuffers), one
of the other (hopefully older) dirty buffers associated with
the vnode is written (using bawrite). In the event that this
approach fails to curb the growth in it the vnode's number of
dirty buffers (due to soft updates rollback dependencies),
the more drastic approach of doing a VOP_FSYNC on the vnode
is used. This code primarily affects very large and actively
written files such as snapshots. This change should eliminate
hanging when taking snapshots or doing background fsck on
very large filesystems.

Hopefully, one day it will be possible to cache filesystem
metadata in the VM cache as is done with file data. As it
stands, only the buffer cache can be used which limits total
metadata storage to about 20Mb no matter how much memory is
available on the system. This rather small memory gets badly
thrashed causing a lot of extra I/O. For example, taking a
snapshot of a 1Tb filesystem minimally requires about 35,000
write operations, but because of the cache thrashing (we only
have about 350 buffers at our disposal) ends up doing about
237,540 I/O's thus taking twenty-five minutes instead of four
if it could run entirely in the cache.

Reported by: Attila Nagy <bra@fsn.hu>
Sponsored by:   DARPA & NAI Labs.

21 years agoRemove a bogus comment.
David Xu [Tue, 25 Feb 2003 05:17:18 +0000 (05:17 +0000)]
Remove a bogus comment.

21 years agoRemove a never true condition.
David Xu [Tue, 25 Feb 2003 05:14:18 +0000 (05:14 +0000)]
Remove a never true condition.

21 years ago - Add an interlock argument to BUF_LOCK and BUF_TIMELOCK.
Jeff Roberson [Tue, 25 Feb 2003 03:37:48 +0000 (03:37 +0000)]
 - Add an interlock argument to BUF_LOCK and BUF_TIMELOCK.
 - Remove the buftimelock mutex and acquire the buf's interlock to protect
   these fields instead.
 - Hold the vnode interlock while locking bufs on the clean/dirty queues.
   This reduces some cases from one BUF_LOCK with a LK_NOWAIT and another
   BUF_LOCK with a LK_TIMEFAIL to a single lock.

Reviewed by: arch, mckusick

21 years agoCleanup of the d_mmap_t interface.
Maxime Henrion [Tue, 25 Feb 2003 03:21:22 +0000 (03:21 +0000)]
Cleanup of the d_mmap_t interface.

- Get rid of the useless atop() / pmap_phys_address() detour.  The
  device mmap handlers must now give back the physical address
  without atop()'ing it.
- Don't borrow the physical address of the mapping in the returned
  int.  Now we properly pass a vm_offset_t * and expect it to be
  filled by the mmap handler when the mapping was successful.  The
  mmap handler must now return 0 when successful, any other value
  is considered as an error.  Previously, returning -1 was the only
  way to fail.  This change thus accidentally fixes some devices
  which were bogusly returning errno constants which would have been
  considered as addresses by the device pager.
- Garbage collect the poorly named pmap_phys_address() now that it's
  no longer used.
- Convert all the d_mmap_t consumers to the new API.

I'm still not sure wheter we need a __FreeBSD_version bump for this,
since and we didn't guarantee API/ABI stability until 5.1-RELEASE.

Discussed with: alc, phk, jake
Reviewed by: peter
Compile-tested on: LINT (i386), GENERIC (alpha and sparc64)
Runtime-tested on: i386

21 years agoHold the TCP protocol lock while modifying the connection hash table.
Jeffrey Hsu [Tue, 25 Feb 2003 01:32:03 +0000 (01:32 +0000)]
Hold the TCP protocol lock while modifying the connection hash table.

21 years agoRemove support for running in SimOS. The support has rotted over
Marcel Moolenaar [Tue, 25 Feb 2003 00:42:40 +0000 (00:42 +0000)]
Remove support for running in SimOS. The support has rotted over
time and there's no indication that it will improve anytime soon.
By removing support for SimOS it is possible to build LINT on
Alpha, which is considered more important at the moment.

Not objected to on: alpha@

21 years agoOnly apply rev 1.10 (which hacks around the i386 boot2 being too big for
David E. O'Brien [Tue, 25 Feb 2003 00:10:20 +0000 (00:10 +0000)]
Only apply rev 1.10 (which hacks around the i386 boot2 being too big for
both ufs1 and ufs2 support) on i386.

21 years agomdoc(7) police: Tidy up.
Ruslan Ermilov [Mon, 24 Feb 2003 23:02:37 +0000 (23:02 +0000)]
mdoc(7) police: Tidy up.

21 years agoFixed a typo in MLINKS.
Ruslan Ermilov [Mon, 24 Feb 2003 22:59:01 +0000 (22:59 +0000)]
Fixed a typo in MLINKS.

21 years agoInstall the gcov(1) manpage; kernbb(8) references it.
Ruslan Ermilov [Mon, 24 Feb 2003 22:58:19 +0000 (22:58 +0000)]
Install the gcov(1) manpage; kernbb(8) references it.

21 years agoSo we don't use this header anymore.
Ruslan Ermilov [Mon, 24 Feb 2003 22:57:06 +0000 (22:57 +0000)]
So we don't use this header anymore.

21 years agoFixed copyright.
Ruslan Ermilov [Mon, 24 Feb 2003 22:55:33 +0000 (22:55 +0000)]
Fixed copyright.
Tidy up the markup.
Only describe the new, post-2.0 behavior.
Added the RETURN VALUES and ERRORS sections.

21 years agoKilled one more instance of pseudo-devices.
Ruslan Ermilov [Mon, 24 Feb 2003 22:54:26 +0000 (22:54 +0000)]
Killed one more instance of pseudo-devices.

21 years agomdoc(7) police: Scheduled sweep.
Ruslan Ermilov [Mon, 24 Feb 2003 22:53:26 +0000 (22:53 +0000)]
mdoc(7) police: Scheduled sweep.

21 years agoNO_GEOM cleanup:
Poul-Henning Kamp [Mon, 24 Feb 2003 21:49:27 +0000 (21:49 +0000)]
NO_GEOM cleanup:

Move to new "struct disk *" centered API".

OK'ed by: emoore

21 years ago- Removed UMAXPTDI and UMAXPTEOFF.
Jake Burkholder [Mon, 24 Feb 2003 20:29:52 +0000 (20:29 +0000)]
- Removed UMAXPTDI and UMAXPTEOFF.
- Changed VM_MAXUSER_ADDRESS to be defined in terms of PTDPTDI.  In order for
  assumptions about the recursive page table map to work it must be the base
  of the recursive map.  Any pte offset that's not NPTEPG will break these
  assumptions.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoEnsure that build-tools end up in the object directory different
Ruslan Ermilov [Mon, 24 Feb 2003 20:09:16 +0000 (20:09 +0000)]
Ensure that build-tools end up in the object directory different
from the source directory.  (This mostly affects the RELENG_4's
``make release'' release.5 target, where "rtermcap" build-tool
for release/sysinstall ends up in the source directory and later
steps of release.5 wipe it out.)

Spotted by: jhay

21 years agoEMACS joke (no, not EMACS itself).
Daniel C. Sobral [Mon, 24 Feb 2003 18:04:20 +0000 (18:04 +0000)]
EMACS joke (no, not EMACS itself).

21 years agoFix a grammar bogon.
Jens Schweikhardt [Mon, 24 Feb 2003 17:52:02 +0000 (17:52 +0000)]
Fix a grammar bogon.

Submitted by: Seth Hieronymus <sethh@principia.edu>
X-MFC after: 4.8 is out the door.

21 years agoNew errata: SA-03:02, SA-03:03.
Bruce A. Mah [Mon, 24 Feb 2003 17:50:27 +0000 (17:50 +0000)]
New errata:  SA-03:02, SA-03:03.

21 years agoIn vm_page_dirty(), assert that the page is not in the free queue(s).
Alan Cox [Mon, 24 Feb 2003 17:30:45 +0000 (17:30 +0000)]
In vm_page_dirty(), assert that the page is not in the free queue(s).

21 years agoFix whitespace bogon introduced in previous commit...no content
Bruce A. Mah [Mon, 24 Feb 2003 17:23:03 +0000 (17:23 +0000)]
Fix whitespace bogon introduced in previous commit...no content
changes.

21 years agoNew release notes: SA-03:02, SA-03:03.
Bruce A. Mah [Mon, 24 Feb 2003 17:21:15 +0000 (17:21 +0000)]
New release notes:  SA-03:02, SA-03:03.

21 years agoFix style bugs in previous commit.
Mike Barcroft [Mon, 24 Feb 2003 17:10:32 +0000 (17:10 +0000)]
Fix style bugs in previous commit.

21 years agoCorrectly map SIGSYS signal to/from Linux.
Alexander Kabaev [Mon, 24 Feb 2003 16:16:45 +0000 (16:16 +0000)]
Correctly map SIGSYS signal to/from Linux.

Submitted by:   "Georg-W. Koltermann" <g.w.k@web.de>

21 years agoThe mpbiosreason variable does not used for pc98.
Yoshihiro Takahashi [Mon, 24 Feb 2003 14:36:03 +0000 (14:36 +0000)]
The mpbiosreason variable does not used for pc98.

21 years agoFix printf format error.
Yoshihiro Takahashi [Mon, 24 Feb 2003 14:32:27 +0000 (14:32 +0000)]
Fix printf format error.

21 years agoHacks to get i386 and pc98 building:
Dag-Erling Smørgrav [Mon, 24 Feb 2003 10:48:01 +0000 (10:48 +0000)]
Hacks to get i386 and pc98 building:
 - TARGET=${TARGET_MACHINE}
 - BOOT2_UFS=UFS1_ONLY

21 years agoClarify about NSWAPDEV, add a DIAGNOSTICS section pointing to
David Schultz [Mon, 24 Feb 2003 08:52:06 +0000 (08:52 +0000)]
Clarify about NSWAPDEV, add a DIAGNOSTICS section pointing to
swapon(2), and correct HISTORY.

Reviewed by: mike (mentor)

21 years ago- Add a clearer error message for the case where swapon hits the
David Schultz [Mon, 24 Feb 2003 08:51:31 +0000 (08:51 +0000)]
- Add a clearer error message for the case where swapon hits the
  NSWAPDEV limit.
- Don't warn about devices that are not in use in 'swapoff -a'.
- Re-add behavior mistakenly removed in revision 1.44:
  If using 'swapon -a', do not warn if the device is already in use.

PR: 46633
Submitted by: Andy Farkas <andyf@speednet.com.au> (in part)
Reviewed by: mike (mentor)

21 years agoExpand the reference count on struct dquot to 32 bits.
David Schultz [Mon, 24 Feb 2003 08:49:59 +0000 (08:49 +0000)]
Expand the reference count on struct dquot to 32 bits.
This fixes a panic on large systems where a single user
may have more than 64K active or inactive vnodes.

PR: 48234
Reviewed by: mike (mentor)

21 years agoThird attempt at removing machdep.h and using ALIGN from <sys/param.h>.
Marcel Moolenaar [Mon, 24 Feb 2003 08:07:05 +0000 (08:07 +0000)]
Third attempt at removing machdep.h and using ALIGN from <sys/param.h>.
The initial stack_block is staticly allocated and will be aligned
according to the alignment requirements of pointers, which does not
necessarily match the alignment enforced by ALIGN. To solve this a
more involved change is required: remove the static initial stack
and deal with an initial condition of not having a stack at all. This
change is therefore more risky than the previous ones, but unavoidable
(other than not using the platform default alignment).

Discussed with: tjr
Approved and reviewed by: tjr
Tested on: alpha, i386, ia64 and sparc64

21 years agoWhen removing the last item from a non-empty worklist, the worklist
Kirk McKusick [Mon, 24 Feb 2003 07:28:41 +0000 (07:28 +0000)]
When removing the last item from a non-empty worklist, the worklist
tail pointer must be updated.

Reported by: Kris Kennaway <kris@obsecurity.org>
Sponsored by:   DARPA & NAI Labs.

21 years ago<machine/vmparam.h> is the right place to get KERNBASE, so don't special
David E. O'Brien [Mon, 24 Feb 2003 06:56:18 +0000 (06:56 +0000)]
<machine/vmparam.h> is the right place to get KERNBASE, so don't special
platform case it (or use alternate spellings).

21 years agoo add -b option to enable batching
Sam Leffler [Mon, 24 Feb 2003 06:10:50 +0000 (06:10 +0000)]
o add -b option to enable batching
o add #ifdef __FreeBSD__'s around the profiling so the code compiles on openbsd
o update a comment about expected performance for Broadcom hardware

21 years agoo instead of applying arbitrary tunables, just honor the COP_F_BATCH
Sam Leffler [Mon, 24 Feb 2003 06:03:13 +0000 (06:03 +0000)]
o instead of applying arbitrary tunables, just honor the COP_F_BATCH
  flag that can be marked on each symmetric op
o eliminate hw.ubsec.maxbatch and hw.ubsec.maxaggr since they are not
  needed anymore
o change ubsec_feed to return void instead of int since zero is always
  returned and noone ever looked at the return value

21 years agoReturn bits which were removed in revision 1.22.
Tom Rhodes [Mon, 24 Feb 2003 05:59:12 +0000 (05:59 +0000)]
Return bits which were removed in revision 1.22.

Requested by: imp

21 years agoCheck whether setmode(3) fails and deallocate its return value after
Mike Heffner [Mon, 24 Feb 2003 05:56:36 +0000 (05:56 +0000)]
Check whether setmode(3) fails and deallocate its return value after
calling getmode(3).

Approved by: wollman
MFC after: 1 week

21 years agoReturn bits which were removed in revision 1.28.
Tom Rhodes [Mon, 24 Feb 2003 05:53:27 +0000 (05:53 +0000)]
Return bits which were removed in revision 1.28.

Requested by: imp

21 years agoDon't NULL out p_fd until after closefd() has been called. This isn't
Scott Long [Mon, 24 Feb 2003 05:46:55 +0000 (05:46 +0000)]
Don't NULL out p_fd until after closefd() has been called.  This isn't
totally correct, but it has caused breakage for too long.  I welcome
someone with more fd fu to fix it correctly.

21 years agopst(4) should be portable across all our platforms.
David E. O'Brien [Mon, 24 Feb 2003 05:12:24 +0000 (05:12 +0000)]
pst(4) should be portable across all our platforms.

21 years agoRevert to old (broken for over 1.5Tb filesystems) version of cgbase
Kirk McKusick [Mon, 24 Feb 2003 04:57:01 +0000 (04:57 +0000)]
Revert to old (broken for over 1.5Tb filesystems) version of cgbase
so that boot loader once again will fit.

Sponsored by:   DARPA & NAI Labs.

21 years agoWrap the static endian functions to shutup GCC.
David E. O'Brien [Mon, 24 Feb 2003 04:44:53 +0000 (04:44 +0000)]
Wrap the static endian functions to shutup GCC.

21 years agoBe more specific about supported cards.
Tom Rhodes [Mon, 24 Feb 2003 04:23:56 +0000 (04:23 +0000)]
Be more specific about supported cards.

Submitted by: imp

21 years agoFix a comment which didn't match the new cookie behavior.
Mike Silbersack [Mon, 24 Feb 2003 03:15:48 +0000 (03:15 +0000)]
Fix a comment which didn't match the new cookie behavior.

Submitted by: Scott Renfro <scott@renfro.org>
MFC after: 1 day

21 years agoCatch up with ATAng changes
Peter Grehan [Mon, 24 Feb 2003 03:12:30 +0000 (03:12 +0000)]
Catch up with ATAng changes

21 years agoI thought groff was wrapped with NO_CXX. Oh well, it is now.
David E. O'Brien [Mon, 24 Feb 2003 02:40:56 +0000 (02:40 +0000)]
I thought groff was wrapped with NO_CXX.  Oh well, it is now.

21 years agoCheck the value given for number-of-logs to keep in config-file entries,
Garance A Drosehn [Mon, 24 Feb 2003 02:28:46 +0000 (02:28 +0000)]
Check the value given for number-of-logs to keep in config-file entries,
and error-out if the value is negative (avoiding an infinite loop).

PR: bin/30654 (part of it)
MFC after: 3 weeks

21 years agoRemove a comment about pccardd(8).
Tom Rhodes [Mon, 24 Feb 2003 02:20:20 +0000 (02:20 +0000)]
Remove a comment about pccardd(8).

Discussed with: imp

21 years ago1) Fix so -a honors the -n ("do nothing") option. Before, if the directory
Garance A Drosehn [Mon, 24 Feb 2003 02:09:02 +0000 (02:09 +0000)]
1) Fix so -a honors the -n ("do nothing") option.  Before, if the directory
   given for -a did not exist, then newsyslog would always try to create
   it, even if -n was specified.
2) When -a processing *does* create the directory, have it check the result
   from mkdir(), and immediately error-out if that failed.

PR: bin/46974
MFC after: 3 weeks

21 years agoFix another case for timewait.
Jonathan Lemon [Mon, 24 Feb 2003 02:06:50 +0000 (02:06 +0000)]
Fix another case for timewait.

21 years agoUpdate to work with the new timewait state.
Jonathan Lemon [Mon, 24 Feb 2003 02:06:02 +0000 (02:06 +0000)]
Update to work with the new timewait state.

Reviewed by: sam

21 years agoAdd device wlan as its now required.
Tom Rhodes [Mon, 24 Feb 2003 01:45:44 +0000 (01:45 +0000)]
Add device wlan as its now required.

Discussed with: imp

21 years agoRemove a XXXKSE. kg_completed now needs proc lock.
David Xu [Mon, 24 Feb 2003 01:28:10 +0000 (01:28 +0000)]
Remove a XXXKSE. kg_completed now needs proc lock.

21 years agoMinor style fix. Add parenthesis around 'return' values.
Garance A Drosehn [Mon, 24 Feb 2003 01:05:05 +0000 (01:05 +0000)]
Minor style fix.  Add parenthesis around 'return' values.

Reviewed by: md5
MFC after: 3 weeks

21 years agotcp_twstart() need to be called with the TCP protocol lock held to avoid
Jeffrey Hsu [Mon, 24 Feb 2003 00:52:03 +0000 (00:52 +0000)]
tcp_twstart() need to be called with the TCP protocol lock held to avoid
a race condition with the TCP timer routines.

21 years agoAdd the idea of a "default rotate action", which will be used if a file is
Garance A Drosehn [Mon, 24 Feb 2003 00:51:41 +0000 (00:51 +0000)]
Add the idea of a "default rotate action", which will be used if a file is
specified at runtime, but that filename is not listed in the newsyslog.conf
file.  This default-action can be changed by having a line in newsyslog.conf
with the filename of "<default>".  Before this change, the program would
quietly ignore the given file.  An update to the man page will be written
after I finish some other updates to newsyslog.c.

Reviewed by: no objections from freebsd-arch
MFC after: 3 weeks

21 years agoBackout last surplus commit. That day just wasn't my day.
David Xu [Mon, 24 Feb 2003 00:49:55 +0000 (00:49 +0000)]
Backout last surplus commit. That day just wasn't my day.

21 years agoPass the right function to callout_reset() for a compressed
Jeffrey Hsu [Mon, 24 Feb 2003 00:48:12 +0000 (00:48 +0000)]
Pass the right function to callout_reset() for a compressed
TIME-WAIT control block.

21 years agoUse the direct mapping of IdlePTD setup in locore for proc0's page directory,
Jake Burkholder [Mon, 24 Feb 2003 00:39:50 +0000 (00:39 +0000)]
Use the direct mapping of IdlePTD setup in locore for proc0's page directory,
instead of allocating another page of kva and mapping it in again.  This was
likely an oversight in revision 1.174 (cut and paste from pmap_pinit).

Discussed with: peter, tegge
Sponsored by: DARPA, Network Associates Laboratories

21 years agoAdd apple partition map GEOM module
Peter Grehan [Mon, 24 Feb 2003 00:16:51 +0000 (00:16 +0000)]
Add apple partition map GEOM module

21 years agoMake the 'a' parameter of bus_space_write_multi_stream_*() a const pointer.
David E. O'Brien [Mon, 24 Feb 2003 00:11:15 +0000 (00:11 +0000)]
Make the 'a' parameter of bus_space_write_multi_stream_*() a const pointer.

21 years agoAllow machines with one CPU and a valid mp table to boot an SMP kernel.
Tor Egge [Sun, 23 Feb 2003 23:49:57 +0000 (23:49 +0000)]
Allow machines with one CPU and a valid mp table to boot an SMP kernel.

21 years agoAdd support for Peppercon ROL-F Card.
Dan Moschuk [Sun, 23 Feb 2003 23:35:35 +0000 (23:35 +0000)]
Add support for Peppercon ROL-F Card.

Submitted by: Sascha Holzeiter <sascha@root-login.org>
PR: 48559

21 years agoUse the correct size for reading and writing the PCI config space.
Marcel Moolenaar [Sun, 23 Feb 2003 23:09:17 +0000 (23:09 +0000)]
Use the correct size for reading and writing the PCI config space.
Reading the PCI config space with the wrong (larger) size is not
a problem in this case, but writing can be as it clobbers unrelated
registers. In this case the clobbering is for reserved fields, which
too is mostly harmless... for now. Hence, this change is mostly
preventive in nature.

21 years agoSync new socket nonblocking/async state with file flags in accept().
Tor Egge [Sun, 23 Feb 2003 23:00:28 +0000 (23:00 +0000)]
Sync new socket nonblocking/async state with file flags in accept().

PR: 1775
Reviewed by: mbr

21 years agoThere is no reason to be cute with ntohl(). Just call it directly rather
David E. O'Brien [Sun, 23 Feb 2003 22:26:39 +0000 (22:26 +0000)]
There is no reason to be cute with ntohl().  Just call it directly rather
than use a macro that tries to do conversions in place.

Compile tested on: sparc64

21 years agoExplain what exca is.
Warner Losh [Sun, 23 Feb 2003 22:22:29 +0000 (22:22 +0000)]
Explain what exca is.

21 years agoPrevious commit missed a 1 that should be NGPTD, and an NPDEPG that should
Jake Burkholder [Sun, 23 Feb 2003 22:12:08 +0000 (22:12 +0000)]
Previous commit missed a 1 that should be NGPTD, and an NPDEPG that should
be NPDEPTD.  Grumble.

Sponsored by: DARPA, Network Associates Laboratories

21 years agoThis commit was generated by cvs2svn to compensate for changes in r111368,
David E. O'Brien [Sun, 23 Feb 2003 21:58:31 +0000 (21:58 +0000)]
This commit was generated by cvs2svn to compensate for changes in r111368,
which included commits to RCS files with non-trunk default branches.

21 years agoImport Linux's linux/include/asm-sparc64/bitopts.h.
David E. O'Brien [Sun, 23 Feb 2003 21:58:31 +0000 (21:58 +0000)]
Import Linux's linux/include/asm-sparc64/bitopts.h.
This is taken from the 2.4.3 Linux sources as shipped on Red Hat 7.1 Alpha.