]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agominor grammar fixes
Bill Fumerola [Thu, 14 Mar 2002 10:09:19 +0000 (10:09 +0000)]
minor grammar fixes

22 years agoAdd an ENVIRONMENT section, and document TARGET_ARCH and NO_WERROR.
Murray Stokely [Thu, 14 Mar 2002 09:38:24 +0000 (09:38 +0000)]
Add an ENVIRONMENT section, and document TARGET_ARCH and NO_WERROR.
Add a cross-build example.

22 years ago* Use a mutex to protect the RID allocator.
Doug Rabson [Thu, 14 Mar 2002 09:28:05 +0000 (09:28 +0000)]
* Use a mutex to protect the RID allocator.
* Use ptc.g instead of ptc.l so that TLB shootdowns are broadcast to the
  coherence domain.
* Use smp_rendezvous for pmap_invalidate_all to ensure it happens on all
  cpus.
* Dike out a DIAGNOSTIC printf which didn't compile.
* Protect the internals of pmap_install with cpu_critical_enter/exit.

22 years agoMove the call to pmap_bootstrap to after the initialisation of thread0.
Doug Rabson [Thu, 14 Mar 2002 09:20:07 +0000 (09:20 +0000)]
Move the call to pmap_bootstrap to after the initialisation of thread0.
This allows us to use mutexes in pmap safely. Also initialise fpcurthread
for cpu0 so that ia64_fpstate_check doesn't barf during boot.

22 years agoMissed this file for select SMP fixes associated with rev 1.93 of
Alfred Perlstein [Thu, 14 Mar 2002 05:16:18 +0000 (05:16 +0000)]
Missed this file for select SMP fixes associated with rev 1.93 of
kern/sys_generic.c

22 years agoMissed this file for select SMP fixes associated with rev 1.93 of
Alfred Perlstein [Thu, 14 Mar 2002 04:47:08 +0000 (04:47 +0000)]
Missed this file for select SMP fixes associated with rev 1.93 of
kern/sys_generic.c

22 years agoVarious mdoc fixes.
Murray Stokely [Thu, 14 Mar 2002 04:21:50 +0000 (04:21 +0000)]
Various mdoc fixes.

Submitted by: ru

22 years agoTune WARNS for Alpha.
David E. O'Brien [Thu, 14 Mar 2002 02:24:25 +0000 (02:24 +0000)]
Tune WARNS for Alpha.

22 years agoDocument faultstate.lookup_still_valid more than none.
Brian Feldman [Thu, 14 Mar 2002 02:10:14 +0000 (02:10 +0000)]
Document faultstate.lookup_still_valid more than none.

Requested by: alfred

22 years agoActivate the sparc64 installation document.
Bruce A. Mah [Thu, 14 Mar 2002 01:47:56 +0000 (01:47 +0000)]
Activate the sparc64 installation document.

22 years agoWhitespace, no content or markup changes.
Bruce A. Mah [Thu, 14 Mar 2002 01:40:47 +0000 (01:40 +0000)]
Whitespace, no content or markup changes.

22 years agos/Sparc/SPARC/
Bruce A. Mah [Thu, 14 Mar 2002 01:34:26 +0000 (01:34 +0000)]
s/Sparc/SPARC/

22 years agoFixes to make select/poll mpsafe.
Alfred Perlstein [Thu, 14 Mar 2002 01:32:30 +0000 (01:32 +0000)]
Fixes to make select/poll mpsafe.

Problem:
  selwakeup required calling pfind which would cause lock order
  reversals with the allproc_lock and the per-process filedesc lock.
Solution:
  Instead of recording the pid of the select()'ing process into the
  selinfo structure, actually record a pointer to the thread.  To
  avoid dereferencing a bad address all the selinfo structures that
  are in use by a thread are kept in a list hung off the thread
  (protected by sellock).  When a selwakeup occurs the selinfo is
  removed from that threads list, it is also removed on the way out
  of select or poll where the thread will traverse its list removing
  all the selinfos from its own list.

Problem:
  Previously the PROC_LOCK was used to provide the mutual exclusion
  needed to ensure proper locking, this couldn't work because there
  was a single condvar used for select and poll and condvars can
  only be used with a single mutex.
Solution:
  Introduce a global mutex 'sellock' which is used to provide mutual
  exclusion when recording events to wait on as well as performing
  notification when an event occurs.

Interesting note:
  schedlock is required to manipulate the per-thread TDF_SELECT
  flag, however if given its own field it would not need schedlock,
  also because TDF_SELECT is only manipulated under sellock one
  doesn't actually use schedlock for syncronization, only to protect
  against corruption.

Proc locks are no longer used in select/poll.

Portions contributed by: davidc

22 years agoAdd a warning that URLs in this document are subject to change (in
Bruce A. Mah [Thu, 14 Mar 2002 01:29:31 +0000 (01:29 +0000)]
Add a warning that URLs in this document are subject to change (in
fact, they're incorrect right now).  Point readers to the list
archives for current locations.

22 years agoThis corrects the first of two known deadlock conditions that
Kirk McKusick [Thu, 14 Mar 2002 01:21:13 +0000 (01:21 +0000)]
This corrects the first of two known deadlock conditions that
come from the presence of a snapshot file.

22 years agoDon't restore r13 when returning to kernel mode. We may have migrated to
Doug Rabson [Thu, 14 Mar 2002 00:28:10 +0000 (00:28 +0000)]
Don't restore r13 when returning to kernel mode. We may have migrated to
a different cpu since the exception_save and r13 needs to point at the
current cpu's pcpu structure.

22 years agoRename SI_SUB_MUTEX to SI_SUB_MTX_POOL to make the name at all accurate.
Brian Feldman [Wed, 13 Mar 2002 23:48:08 +0000 (23:48 +0000)]
Rename SI_SUB_MUTEX to SI_SUB_MTX_POOL to make the name at all accurate.
While doing this, move it earlier in the sysinit boot process so that the
VM system can use it.

After that, the system is now able to use sx locks instead of lockmgr
locks in the VM system.  To accomplish this, some of the more
questionable uses of the locks (such as testing whether they are
owned or not, as well as allowing shared+exclusive recursion) are
removed, and simpler logic throughout is used so locks should also be
easier to understand.

This has been tested on my laptop for months, and has not shown any
problems on SMP systems, either, so appears quite safe.  One more
user of lockmgr down, many more to go :)

22 years agoDBA update: pick the right DBA for various LOMAC copyrights.
Robert Watson [Wed, 13 Mar 2002 22:55:20 +0000 (22:55 +0000)]
DBA update: pick the right DBA for various LOMAC copyrights.

22 years agoRemove a bogus cast and deconstification that broke the Alpha kernel build.
Dag-Erling Smørgrav [Wed, 13 Mar 2002 19:16:41 +0000 (19:16 +0000)]
Remove a bogus cast and deconstification that broke the Alpha kernel build.

22 years agoBack out previous commit, it was the wrong patch!
Dag-Erling Smørgrav [Wed, 13 Mar 2002 19:13:41 +0000 (19:13 +0000)]
Back out previous commit, it was the wrong patch!

22 years agoRemove a bogus cast that breaks the Alpha kernel build due to -Werror.
Dag-Erling Smørgrav [Wed, 13 Mar 2002 19:12:31 +0000 (19:12 +0000)]
Remove a bogus cast that breaks the Alpha kernel build due to -Werror.

22 years agoClarify the cylinder > 1023 issue, now that EDD support is automatic.
Robert Nordier [Wed, 13 Mar 2002 19:00:19 +0000 (19:00 +0000)]
Clarify the cylinder > 1023 issue, now that EDD support is automatic.
Also drop a note about undocumented features that was intended to
cover the since-documented slice-selection syntax.

22 years agoFix syntax error.
Giorgos Keramidas [Wed, 13 Mar 2002 18:37:40 +0000 (18:37 +0000)]
Fix syntax error.

22 years agoSupport linking with -z combreloc (combine and sort reloc sections) and with
David E. O'Brien [Wed, 13 Mar 2002 18:26:59 +0000 (18:26 +0000)]
Support linking with -z combreloc (combine and sort reloc sections) and with
--shared -z combreloc (shared library, combine & sort relocs)

Submitted by: peter
Bribed by: peter
Fixes: IA-64, Kaffe, and QT
Doesn't fix ports by: kris
Tested on: make(1) release w/docproj
Desired for: 5.0-DP#1
Approved by: murray
Seconded by: jhb

22 years agoAdd 5th parameter to pci_read_device specifying the size of the object
Warner Losh [Wed, 13 Mar 2002 16:32:11 +0000 (16:32 +0000)]
Add 5th parameter to pci_read_device specifying the size of the object
to create.

22 years agoSGML fixes, part2: block elements like lists do not belong into inline
Udo Erdelhoff [Wed, 13 Mar 2002 15:29:51 +0000 (15:29 +0000)]
SGML fixes, part2: block elements like lists do not belong into inline
elements like paragraphs

This is a markup-only commit (just like the last one)

MFC after: 14 days

22 years agoDocument the "lomac_enable" rc.conf switch.
Robert Watson [Wed, 13 Mar 2002 15:28:37 +0000 (15:28 +0000)]
Document the "lomac_enable" rc.conf switch.

Reminded by: Makoto Matsushita <matusita@jp.FreeBSD.org>

22 years agoUse the dx register for the destination port address. A immediate port
Yoshihiro Takahashi [Wed, 13 Mar 2002 13:59:41 +0000 (13:59 +0000)]
Use the dx register for the destination port address. A immediate port
address of outb is only 8 bits.

Submitted by: chi@bd.mbn.or.jp (Chiharu Shibata)
MFC after: 3 days

22 years agoMFi386: revision 1.502
Yoshihiro Takahashi [Wed, 13 Mar 2002 13:45:00 +0000 (13:45 +0000)]
MFi386: revision 1.502

22 years agoFix some of the worst sgml mistakes in this document:
Udo Erdelhoff [Wed, 13 Mar 2002 13:10:22 +0000 (13:10 +0000)]
Fix some of the worst sgml mistakes in this document:
- paragraphs are closed with </para>, not <para>
- for each opening tag, there should be a closing tag

Not fixed (yet): block elements inside inline elements
Noticed during translation, hunted down with suppe/slashexpand

22 years agoDocument the "slice" syntax.
Ruslan Ermilov [Wed, 13 Mar 2002 12:30:46 +0000 (12:30 +0000)]
Document the "slice" syntax.

Prodded by: ambrisko

22 years agoImplement -m and -p loader(8) "boot" command options in boot2.
Ruslan Ermilov [Wed, 13 Mar 2002 11:03:36 +0000 (11:03 +0000)]
Implement -m and -p loader(8) "boot" command options in boot2.
(This is more useful for 4.x where boot blocks can still load
kernels, modulo the PR kern/17422.)

22 years agoFurther document "console" (nullconsole).
Ruslan Ermilov [Wed, 13 Mar 2002 10:55:22 +0000 (10:55 +0000)]
Further document "console" (nullconsole).

22 years agoTokens after #endif are not strict ISO.
David E. O'Brien [Wed, 13 Mar 2002 10:29:06 +0000 (10:29 +0000)]
Tokens after #endif are not strict ISO.

22 years agoUse the return value from snprintf() to keep a track of the length of
Brian Somers [Wed, 13 Mar 2002 10:21:19 +0000 (10:21 +0000)]
Use the return value from snprintf() to keep a track of the length of
the display string in MPPEDispOpts.

PR: 35836
MFC After: 2 weeks

22 years ago* Simplify the synopsis.
Murray Stokely [Wed, 13 Mar 2002 09:54:31 +0000 (09:54 +0000)]
* Simplify the synopsis.
* Remove trailing whitespace at EOL.
* Various grammar cleanups.
* Note that MAKE_ISOS is disabled by default.
* Use more descriptive mdoc markup.
* Use proper references for the online FDP documents.

Submitted by: dd

22 years agoAdd a readonly sysctl variable of type string, kern.bootp_cookie,
Luigi Rizzo [Wed, 13 Mar 2002 09:23:11 +0000 (09:23 +0000)]
Add a readonly sysctl variable of type string, kern.bootp_cookie,
which is initialized with whatever string a dhcp/bootp server passes
as vendor tag 134.
There is no standard tag that I know with this information, and
no vendor-defined tag that applies to FreeBSD that I could find
doing the same thing.

The intended use is to pass information to userland for run-time
configuration of a diskless client without having to run a bootp/dhcp
client for the third time (after the one in pxeboot/etherboot, and
the one in the kernel bootp), also because these clients generally
screwup the interface configuration, which is not exactly what you
want when you have your disks nfs-mounted.

Manpage update to follow soon.

MFC-after: 3 days

22 years agoAdd information about multi-session CDs from the log message of r1.4.
Murray Stokely [Wed, 13 Mar 2002 08:27:27 +0000 (08:27 +0000)]
Add information about multi-session CDs from the log message of r1.4.
Also, add an examples section and show how to mount a Kodak Photo-CD.

PR: docs/35206

22 years agoAdd encounter between Flinders and Baudin.
Greg Lehey [Wed, 13 Mar 2002 07:57:24 +0000 (07:57 +0000)]
Add encounter between Flinders and Baudin.

22 years agoAdd note about genscripts.sh.
David E. O'Brien [Wed, 13 Mar 2002 07:16:00 +0000 (07:16 +0000)]
Add note about genscripts.sh.

22 years agoReference release(7) in SEE ALSO.
Murray Stokely [Wed, 13 Mar 2002 06:59:39 +0000 (06:59 +0000)]
Reference release(7) in SEE ALSO.

22 years agoAdd a makefile to document the release building process.
Murray Stokely [Wed, 13 Mar 2002 06:33:26 +0000 (06:33 +0000)]
Add a makefile to document the release building process.
Specifically, this documents the available targets and relevant
environment variables for "make release".  LOCAL_PATCHES,
RELEASEDISTFILES, RELEASENOUPDATE, etc.. are covered.

A future commit should add more information about drivers.conf,
boot_crunch.conf, and other less well-known aspects of the release
build.

Reviewed and history section added by: phk

22 years agoFix ifdef LOCORE protection.
Jake Burkholder [Wed, 13 Mar 2002 06:04:36 +0000 (06:04 +0000)]
Fix ifdef LOCORE protection.

22 years agoAdd a DEBUGGER_ON_POWERFAIL option. This makes the power button on ultra 10s
Jake Burkholder [Wed, 13 Mar 2002 05:58:45 +0000 (05:58 +0000)]
Add a DEBUGGER_ON_POWERFAIL option.  This makes the power button on ultra 10s
work like an NMI button.

22 years agoFix braino.
Jake Burkholder [Wed, 13 Mar 2002 05:54:00 +0000 (05:54 +0000)]
Fix braino.

22 years agoNew release notes: sparc64 support and SMP (better late than never),
Bruce A. Mah [Wed, 13 Mar 2002 05:41:41 +0000 (05:41 +0000)]
New release notes:  sparc64 support and SMP (better late than never),
via82c686 sound driver supports VIA VT8233, ich sound driver
supports SiS 7012, libz 1.1.4.

Belatedly note the MFC of the ich sound driver (also better late
than never).

22 years agoCleanup the recent cardbus cleanups. This fixes some of the panics
Warner Losh [Wed, 13 Mar 2002 05:38:19 +0000 (05:38 +0000)]
Cleanup the recent cardbus cleanups.  This fixes some of the panics
that I introduced with -v.  However, other problems still remain (including
the loss of interrupts).

22 years agoTurn up WARNS as high as possible.
David E. O'Brien [Wed, 13 Mar 2002 05:14:31 +0000 (05:14 +0000)]
Turn up WARNS as high as possible.

Tested by: gcc31, gcc32, /usr/bin/cc

22 years agoAdd support for starting and stopping cpus with ipis.
Jake Burkholder [Wed, 13 Mar 2002 04:59:01 +0000 (04:59 +0000)]
Add support for starting and stopping cpus with ipis.
Stop the other cpus when shutting down or entering the debugger.

Submitted by: tmm

22 years agoUse intr_disable/intr_restore instead of doing it manually.
Jake Burkholder [Wed, 13 Mar 2002 04:43:45 +0000 (04:43 +0000)]
Use intr_disable/intr_restore instead of doing it manually.

Submitted by: tmm

22 years agoRemove bogon from bad merge.
David E. O'Brien [Wed, 13 Mar 2002 04:39:40 +0000 (04:39 +0000)]
Remove bogon from bad merge.

22 years agoAdd support for driving the clocks on secondary cpus.
Jake Burkholder [Wed, 13 Mar 2002 04:38:33 +0000 (04:38 +0000)]
Add support for driving the clocks on secondary cpus.

Submitted by: tmm

22 years agoFix a bug where the wrong number of windows were copied for a failed fill
Jake Burkholder [Wed, 13 Mar 2002 04:02:27 +0000 (04:02 +0000)]
Fix a bug where the wrong number of windows were copied for a failed fill
on return to user mode.  We may not have frame pointers setup for more
than 1 on return from exec.

22 years agoWhite space.
Jake Burkholder [Wed, 13 Mar 2002 03:55:28 +0000 (03:55 +0000)]
White space.

22 years agoMake IPI_WAIT use a bit mask of the cpus that a pmap is active on and only
Jake Burkholder [Wed, 13 Mar 2002 03:43:00 +0000 (03:43 +0000)]
Make IPI_WAIT use a bit mask of the cpus that a pmap is active on and only
wait for those cpus, instead of all of them by using a count.  Oops.
Make the pointer to the mask that the primary cpu spins on volatile, so
gcc doesn't optimize out an important load.  Oops again.
Activate tlb shootdown ipi synchronization now that it works.  We have
all involved cpus wait until all the others are done.  This may not be
necessary, it is mostly for sanity.
Make the trigger level interrupt ipi handler work.

Submitted by: tmm

22 years agoAdd an ATOMIC_CLEAR_INT macro.
Jake Burkholder [Wed, 13 Mar 2002 03:28:47 +0000 (03:28 +0000)]
Add an ATOMIC_CLEAR_INT macro.

Submitted by: tmm

22 years agoUse the deo hosted toolchain by default.
Jake Burkholder [Wed, 13 Mar 2002 03:23:17 +0000 (03:23 +0000)]
Use the deo hosted toolchain by default.

22 years agoLABELOFFSET is 128 for sparc64.
Jake Burkholder [Wed, 13 Mar 2002 03:17:29 +0000 (03:17 +0000)]
LABELOFFSET is 128 for sparc64.

Submitted by: tmm
Forgotten by: obrien

22 years agortld support for sparc64.
Jake Burkholder [Wed, 13 Mar 2002 02:40:39 +0000 (02:40 +0000)]
rtld support for sparc64.

Largely obtained from: netbsd
Submitted by: jake, tmm

22 years agoAdd realloc() and reallocf(), and make free(NULL, ...) acceptable.
Archie Cobbs [Wed, 13 Mar 2002 01:42:33 +0000 (01:42 +0000)]
Add realloc() and reallocf(), and make free(NULL, ...) acceptable.

Reviewed by: alfred

22 years agoAdd a bugs notice indicating that the SEESENT flag may behave incorrectly
Robert Watson [Wed, 13 Mar 2002 00:04:53 +0000 (00:04 +0000)]
Add a bugs notice indicating that the SEESENT flag may behave incorrectly
on interfaces with hardware loopback, and point-to-point interfaces
(unclear why that is, may need to be tracked down).

22 years agoAllow LOMAC to be loaded as part of the boot scripts using "lomac_enable"
Robert Watson [Tue, 12 Mar 2002 21:47:31 +0000 (21:47 +0000)]
Allow LOMAC to be loaded as part of the boot scripts using "lomac_enable"
setting in rc.conf.

Extracted from the still clammy hands of: green
Sponsored by: DARPA, NAI Labs

22 years agoNo need to explicitly check for both cases when using grep -i.
Robert Watson [Tue, 12 Mar 2002 21:44:33 +0000 (21:44 +0000)]
No need to explicitly check for both cases when using grep -i.

22 years agoRun dumpon(8) early so crashes during startup can be caught. In
Crist J. Clark [Tue, 12 Mar 2002 20:59:35 +0000 (20:59 +0000)]
Run dumpon(8) early so crashes during startup can be caught. In
addition, take out the checks on the $dumpdev. dumpon(8) behaves well
if given a non-existent filename. It gives a nice error message which
is better rather than the current silent failure.

Reviewed by: des

22 years agoTGA and TGA2 clarification
Wilko Bulte [Tue, 12 Mar 2002 20:50:06 +0000 (20:50 +0000)]
TGA and TGA2 clarification
AS500 onboard sound.

22 years ago- TGA clarification
Wilko Bulte [Tue, 12 Mar 2002 20:44:03 +0000 (20:44 +0000)]
- TGA clarification
- TGA2 VGA compatibility clarification
- Sound on AS500

Thanks to: "Andrew M. Miklic" <AndrwMklc@cs.com> (TGA2)

22 years agoThe reload of ipf(8) rules should depend on $ipfilter_enable, not
Crist J. Clark [Tue, 12 Mar 2002 20:25:25 +0000 (20:25 +0000)]
The reload of ipf(8) rules should depend on $ipfilter_enable, not
$ipfilter_active. $ipfilter_enable is set to "NO" if modules fail to
load, and $ipfilter_active can be "YES" when we are not using ipf(8).

MFC after: 3 days

22 years agoRETURN VALUES is for sections 2, 3 and 9, but here in section 1 man page,
Philippe Charnier [Tue, 12 Mar 2002 20:00:52 +0000 (20:00 +0000)]
RETURN VALUES is for sections 2, 3 and 9, but here in section 1 man page,
DIAGNOSTICS should be used instead.

Reviewed by: wosch

22 years agoSimple fix so the 'LOGIN FAILURE' message send to syslog will include
Garance A Drosehn [Tue, 12 Mar 2002 19:48:32 +0000 (19:48 +0000)]
Simple fix so the 'LOGIN FAILURE' message send to syslog will include
the correct userid, instead of random garbage.  This bug does not
exist in -stable.

Reviewed by: freebsd-audit

22 years agoDon't try to cross-build sysinstall.
Dag-Erling Smørgrav [Tue, 12 Mar 2002 18:55:53 +0000 (18:55 +0000)]
Don't try to cross-build sysinstall.

22 years agoGerman Translation of the Release Notes for FreeBSD-current
Udo Erdelhoff [Tue, 12 Mar 2002 18:35:54 +0000 (18:35 +0000)]
German Translation of the Release Notes for FreeBSD-current

22 years agoHandle B460800 and B921600 being defined in <termios.h>
Brian Somers [Tue, 12 Mar 2002 16:05:26 +0000 (16:05 +0000)]
Handle B460800 and B921600 being defined in <termios.h>

22 years agoDocument the existence of machdep.guessed_bootdev, as suggested
Luigi Rizzo [Tue, 12 Mar 2002 13:55:19 +0000 (13:55 +0000)]
Document the existence of machdep.guessed_bootdev, as suggested
by Sheldon.
For a detailed description look at the commit log for sysctl.c
rev.1.42 -- i do not think it is appropriate to put the full
description in this manpage, and the "boot" and "loader" manpages
where this description might go are also missing a description of
a number of similar variables, so i think this it is ok to limit
documentation to this now, and update it later when I (or someone
else) have a chance to revise "boot" and "loader".

22 years agoTurn back on (x)lint. It is working just fine.
Mark Murray [Tue, 12 Mar 2002 13:43:02 +0000 (13:43 +0000)]
Turn back on (x)lint. It is working just fine.

22 years agoThis driver first appeared in FreeBSD 4.5, not 5.0.
Murray Stokely [Tue, 12 Mar 2002 12:26:05 +0000 (12:26 +0000)]
This driver first appeared in FreeBSD 4.5, not 5.0.

22 years agoPoint readers to the Porter's Handbook instead of the non-existant
Murray Stokely [Tue, 12 Mar 2002 09:45:29 +0000 (09:45 +0000)]
Point readers to the Porter's Handbook instead of the non-existant
"Porting Applications" section of the FreeBSD Handbook.

MFC After: 2 days

22 years agoOnly i386 has wbinvd().
Doug Rabson [Tue, 12 Mar 2002 09:45:17 +0000 (09:45 +0000)]
Only i386 has wbinvd().

22 years ago* Add print/teTeX to the list of important packages for disc #3. This
Murray Stokely [Tue, 12 Mar 2002 08:37:50 +0000 (08:37 +0000)]
* Add print/teTeX to the list of important packages for disc #3.  This
  very large package is exactly the kind of thing that users buy the
  CDs for, so leaving it off of a 4 CD set is very bad.

* Properly quote the last security/sudo entry.

* Add a comment in the CDROM #2 section explaining that packages don't
  typically go on this disc.

22 years agoThis patch adds the "LOCKSHARED" option to namei which causes it to only acquire...
Jeff Roberson [Tue, 12 Mar 2002 04:00:11 +0000 (04:00 +0000)]
This patch adds the "LOCKSHARED" option to namei which causes it to only acquire shared locks on leafs.
The stat() and open() calls have been changed to make use of this new functionality.  Using shared locks in
these cases is sufficient and can significantly reduce their latency if IO is pending to these vnodes.  Also,
this reduces the number of exclusive locks that are floating around in the system, which helps reduce the
number of deadlocks that occur.

A new kernel option "LOOKUP_SHARED" has been added.  It defaults to off so this patch can be turned on for
testing, and should eventually go away once it is proven to be stable.  I have personally been running this
patch for over a year now, so it is believed to be fully stable.

Reviewed by: jake, obrien
Approved by: jake

22 years agoConsistently use $PICFLAG.
David E. O'Brien [Tue, 12 Mar 2002 01:18:07 +0000 (01:18 +0000)]
Consistently use $PICFLAG.

22 years agoSpell pic "PIC" for sparc64.
David E. O'Brien [Tue, 12 Mar 2002 01:14:19 +0000 (01:14 +0000)]
Spell pic "PIC" for sparc64.

22 years agoBackground the startup of `Amd', it often blocks on startup.
David E. O'Brien [Tue, 12 Mar 2002 01:04:35 +0000 (01:04 +0000)]
Background the startup of `Amd', it often blocks on startup.

22 years agoWhy shouldn't amd always write its PID to a file?
David E. O'Brien [Tue, 12 Mar 2002 01:01:53 +0000 (01:01 +0000)]
Why shouldn't amd always write its PID to a file?
Since I cannot answer that question, make it.

22 years agoFix some -Wunused warnings by "using" a macro argument
Peter Wemm [Tue, 12 Mar 2002 00:19:14 +0000 (00:19 +0000)]
Fix some -Wunused warnings by "using" a macro argument

22 years agoFix a warning (make ucontext_t *ucp a const)
Peter Wemm [Tue, 12 Mar 2002 00:18:20 +0000 (00:18 +0000)]
Fix a warning (make ucontext_t *ucp a const)

22 years agoStop concatenating __func__ with strings
Peter Wemm [Tue, 12 Mar 2002 00:17:45 +0000 (00:17 +0000)]
Stop concatenating __func__ with strings

22 years agoDeal with a structure member rename in a recent acpica import
Peter Wemm [Tue, 12 Mar 2002 00:15:57 +0000 (00:15 +0000)]
Deal with a structure member rename in a recent acpica import

22 years agoDo not concatenate __func__ with strings, because it is not a string.
Peter Wemm [Tue, 12 Mar 2002 00:15:06 +0000 (00:15 +0000)]
Do not concatenate __func__ with strings, because it is not a string.
Later gcc's blow up on this.

22 years agoDo not do string concatenation with __func__ (which is not a string)
Peter Wemm [Tue, 12 Mar 2002 00:12:59 +0000 (00:12 +0000)]
Do not do string concatenation with __func__ (which is not a string)

22 years agoRecent acpica imports have changed the lengths from UINT32 to ACPI_SIZE,
Peter Wemm [Tue, 12 Mar 2002 00:10:40 +0000 (00:10 +0000)]
Recent acpica imports have changed the lengths from UINT32 to ACPI_SIZE,
which is 64 bit on ia64.  Fix it.

22 years agoAdd support for 3Com Airconnect PCI cards.
Brooks Davis [Mon, 11 Mar 2002 23:29:59 +0000 (23:29 +0000)]
Add support for 3Com Airconnect PCI cards.

MFC after: 3 days

22 years agoRemove some files we don't want.
Brian Feldman [Mon, 11 Mar 2002 22:53:27 +0000 (22:53 +0000)]
Remove some files we don't want.

22 years agos/nilename/filename/
Brian Feldman [Mon, 11 Mar 2002 22:36:45 +0000 (22:36 +0000)]
s/nilename/filename/

22 years agoFix conflicts.
Brian Feldman [Mon, 11 Mar 2002 22:36:26 +0000 (22:36 +0000)]
Fix conflicts.

22 years agoThis commit was generated by cvs2svn to compensate for changes in r92111,
Brian Feldman [Mon, 11 Mar 2002 22:30:00 +0000 (22:30 +0000)]
This commit was generated by cvs2svn to compensate for changes in r92111,
which included commits to RCS files with non-trunk default branches.

22 years agoImport zlib 1.1.4 onto the vendor branch.
Brian Feldman [Mon, 11 Mar 2002 22:30:00 +0000 (22:30 +0000)]
Import zlib 1.1.4 onto the vendor branch.

22 years agoThis commit was generated by cvs2svn to compensate for changes in r92109,
Brian Feldman [Mon, 11 Mar 2002 22:29:59 +0000 (22:29 +0000)]
This commit was generated by cvs2svn to compensate for changes in r92109,
which included commits to RCS files with non-trunk default branches.

22 years agoImport zlib 1.1.4 onto the vendor branch.
Brian Feldman [Mon, 11 Mar 2002 22:29:59 +0000 (22:29 +0000)]
Import zlib 1.1.4 onto the vendor branch.

22 years agoFirst commit of the GEOM subsystem to make it easier for people to
Poul-Henning Kamp [Mon, 11 Mar 2002 21:42:35 +0000 (21:42 +0000)]
First commit of the GEOM subsystem to make it easier for people to
test and play with this.

This is not yet production quality and should be run only on dedicated
test boxes.

For people who want to develop transformations for GEOM there exist a
set of shims to run geom in userland (ask phk@freebsd.org).

Reports of all kinds to: phk@freebsd.org
Please include in report:
        dmesg
        sysctl debug.geomdot
        sysctl debug.geomconf

Known significant limitations:
        no kernel dump facility.
        ioctls severely restricted.

Sponsored by:   DARPA, NAI Labs

22 years agoAdd new support for locking an ATA channel and use that throughout
Søren Schmidt [Mon, 11 Mar 2002 21:04:32 +0000 (21:04 +0000)]
Add new support for locking an ATA channel and use that throughout
the ATA/ATAPI driver. This solves the concurrency problem with
the new GEOM code, and also cuts a good deal of the patch size
in the upcoming MFC.