]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years ago- Fix uma_zcreate parameters -- don't pass M_* flags here, doh.
Andrew R. Reiter [Thu, 23 May 2002 23:21:00 +0000 (23:21 +0000)]
- Fix uma_zcreate parameters -- don't pass M_* flags here, doh.

22 years agoUpdate comments to better match reality.
Maxime Henrion [Thu, 23 May 2002 23:18:25 +0000 (23:18 +0000)]
Update comments to better match reality.

22 years agoMake mount_nullfs(8) use nmount(2) rather than mount(2) now
Maxime Henrion [Thu, 23 May 2002 23:08:27 +0000 (23:08 +0000)]
Make mount_nullfs(8) use nmount(2) rather than mount(2) now
that nullfs has been converted to nmount.

22 years agoConvert nullfs to nmount.
Maxime Henrion [Thu, 23 May 2002 23:07:27 +0000 (23:07 +0000)]
Convert nullfs to nmount.

22 years agoSlightly change the way we pass mount options to the filesystem
Maxime Henrion [Thu, 23 May 2002 23:02:19 +0000 (23:02 +0000)]
Slightly change the way we pass mount options to the filesystem
VFS_NMOUNT operations.

Reviewed by: phk

22 years agoJust to show that PAM can do almost anything from the ridiculous to the
Dag-Erling Smørgrav [Thu, 23 May 2002 22:03:06 +0000 (22:03 +0000)]
Just to show that PAM can do almost anything from the ridiculous to the
obscene, or - as they say in New York - sophisticated, add pam_echo(8) and
pam_exec(8) to our ever-lengthening roster of PAM modules.

Sponsored by: DARPA, NAI Labs.

22 years agoo Conditionalize sections for POSIX.1-2001 compatibility.
Mike Barcroft [Thu, 23 May 2002 18:48:27 +0000 (18:48 +0000)]
o Conditionalize sections for POSIX.1-2001 compatibility.
o Use POSIX spelling for types, where possible.
o Define size_t in the __BSD_VISIBLE case (this isn't really needed
  for standards conformance, but follows the tradition of not
  requiring <sys/types.h> as a prerequisite).
o Use _BYTE_ORDER and friends instead of BYTE_ORDER and friends, since
  there may not be enough pollution in order for the latter to work.
o Add an XXX note about the missing IPPROTO_IPV6 macro.

22 years agoANSIfy variable-argument macros.
Mark Murray [Thu, 23 May 2002 18:26:23 +0000 (18:26 +0000)]
ANSIfy variable-argument macros.

22 years agoSync with upstream version:
Tony Finch [Thu, 23 May 2002 16:50:41 +0000 (16:50 +0000)]
Sync with upstream version:
  * Ensure we work within the array bounds when parsing command-line options;
  * Replace h0h0getopt with getopt(3);
  * Use consistent whitespace style in the function declarations.

Revieweded by: dwmalone (mentor)

22 years agoIn m_aux_delete, no need to chase beyond victim.
Hajimu UMEMOTO [Thu, 23 May 2002 15:59:48 +0000 (15:59 +0000)]
In m_aux_delete, no need to chase beyond victim.

Submitted by: archie
Obtained from: KAME

22 years agoUn-kmemize. Portions of the tty mode code have been temporarily disabled;
Dag-Erling Smørgrav [Thu, 23 May 2002 15:09:49 +0000 (15:09 +0000)]
Un-kmemize.  Portions of the tty mode code have been temporarily disabled;
everything else, including dead kernel support, works just like before.

Sponsored by: DARPA, NAI Labs

22 years agoHandle numeric keys by checking a "default" file. Handle number-less keys
Juli Mallett [Thu, 23 May 2002 14:58:22 +0000 (14:58 +0000)]
Handle numeric keys by checking a "default" file.  Handle number-less keys
by failing, since comparing up to the length of the key (0) against the
begin line for a key in the help file will always succeed, and print what
is wholly bogus output.

22 years agoNits in previous commits.
Dag-Erling Smørgrav [Thu, 23 May 2002 14:58:07 +0000 (14:58 +0000)]
Nits in previous commits.

Sponsored by: DARPA, NAI Labs

22 years agoCull large amounts of dead code (deprecated since 1997)
Dag-Erling Smørgrav [Thu, 23 May 2002 14:53:04 +0000 (14:53 +0000)]
Cull large amounts of dead code (deprecated since 1997)

Sponsored by: DARPA, NAI Labs

22 years agoANSIfy.
Dag-Erling Smørgrav [Thu, 23 May 2002 14:48:25 +0000 (14:48 +0000)]
ANSIfy.

Sponsored by: DARPA, NAI Labs

22 years agoStaticize.
Dag-Erling Smørgrav [Thu, 23 May 2002 14:44:23 +0000 (14:44 +0000)]
Staticize.

Sponsored by: DARPA, NAI Labs

22 years agoWhitespace cleanup.
Dag-Erling Smørgrav [Thu, 23 May 2002 14:42:17 +0000 (14:42 +0000)]
Whitespace cleanup.

Sponsored by: DARPA, NAI Labs

22 years agoUnbreak tty mode (cons was renamed to constty a while ago)
Dag-Erling Smørgrav [Thu, 23 May 2002 14:41:52 +0000 (14:41 +0000)]
Unbreak tty mode (cons was renamed to constty a while ago)

Sponsored by: DARPA, NAI Labs

22 years agoAvoid wandering over the beginning of the actual buffer
Robert Drehmel [Thu, 23 May 2002 14:21:02 +0000 (14:21 +0000)]
Avoid wandering over the beginning of the actual buffer
if the passed template string contains only 'X' characters.

Submitted by: Mark Andrews <marka@isc.org> (patch modified)
PR: 38402

22 years agoRemoved the ch, sa and ses devices from 'SMALL' boot floppy image.
Yoshihiro Takahashi [Thu, 23 May 2002 14:15:48 +0000 (14:15 +0000)]
Removed the ch, sa and ses devices from 'SMALL' boot floppy image.

22 years agoload_env(), the function that attempts to parse a crontab
Ollivier Robert [Thu, 23 May 2002 13:16:30 +0000 (13:16 +0000)]
load_env(), the function that attempts to parse a crontab
line as an environment variable assignment, is broken
and not conformant to its description in the manual page.

I think it is worthwhile to have that fix in 4.6.

PR:             bin/38374
Submitted by:   Thomas Quinot <thomas@cuivre.fr.eu.org>
MFC after:      2 days

22 years agoWhitespace only; fix indentation.
Mark Murray [Thu, 23 May 2002 12:09:14 +0000 (12:09 +0000)]
Whitespace only; fix indentation.

22 years agoSet the MAKEFILE variable to the value passed to ReadMakefile(), not the full
Juli Mallett [Thu, 23 May 2002 12:01:54 +0000 (12:01 +0000)]
Set the MAKEFILE variable to the value passed to ReadMakefile(), not the full
path to it.  Use the full path only for parsing it.

22 years agoDont panic if only one disk on a channel can do tagged queueing.
Søren Schmidt [Thu, 23 May 2002 08:17:16 +0000 (08:17 +0000)]
Dont panic if only one disk on a channel can do tagged queueing.

MFC: asap

22 years agoThis manpage is for Alpha only.
Ruslan Ermilov [Thu, 23 May 2002 07:38:51 +0000 (07:38 +0000)]
This manpage is for Alpha only.

22 years agoFixed broken ``make -jX install''.
Ruslan Ermilov [Thu, 23 May 2002 07:25:01 +0000 (07:25 +0000)]
Fixed broken ``make -jX install''.

Spotted by: make release TARGET_ARCH=ia64

22 years agoMinor nit: get p pointer in msleep() from td->td_proc (where
John Baldwin [Thu, 23 May 2002 04:14:18 +0000 (04:14 +0000)]
Minor nit: get p pointer in msleep() from td->td_proc (where
td == curthread) rather than from curproc.

22 years agoWhitespace: trim a trailing tab.
John Baldwin [Thu, 23 May 2002 04:12:28 +0000 (04:12 +0000)]
Whitespace: trim a trailing tab.

22 years agoMake the counters uintmax_ts, and use %ju rather than %llu.
Dag-Erling Smørgrav [Thu, 23 May 2002 03:08:42 +0000 (03:08 +0000)]
Make the counters uintmax_ts, and use %ju rather than %llu.

22 years agoMake my style consistent.
Juli Mallett [Thu, 23 May 2002 03:03:44 +0000 (03:03 +0000)]
Make my style consistent.

Remove two includes.

Fix a typo (semicolon instead of period at EOL).

22 years agoComment out extra tokens after #else and #endif.
Dag-Erling Smørgrav [Thu, 23 May 2002 00:38:07 +0000 (00:38 +0000)]
Comment out extra tokens after #else and #endif.

22 years agoFix a panic by allocating the iface structure locally in the attach
Josef Karthauser [Thu, 23 May 2002 00:36:14 +0000 (00:36 +0000)]
Fix a panic by allocating the iface structure locally in the attach
function instead of in usb_probe_and_attach.

22 years ago- Turn two more storage pools into UMA zones and make the related memory
Andrew R. Reiter [Thu, 23 May 2002 00:34:12 +0000 (00:34 +0000)]
- Turn two more storage pools into UMA zones and make the related memory
  allocations and frees use the UMA api.

22 years agoReflect some changes in the NetBSD code path, and sligh adjustments to ours.
Josef Karthauser [Thu, 23 May 2002 00:26:06 +0000 (00:26 +0000)]
Reflect some changes in the NetBSD code path, and sligh adjustments to ours.
(Non-functional changes).

22 years agoHide a couple of unguarded error returns behind the no_fail test.
Dag-Erling Smørgrav [Thu, 23 May 2002 00:04:14 +0000 (00:04 +0000)]
Hide a couple of unguarded error returns behind the no_fail test.

Sponsored by: DARPA, NAI Labs

22 years agoHide a couple of unguarded error returns behind the no_fail test.
Dag-Erling Smørgrav [Thu, 23 May 2002 00:02:59 +0000 (00:02 +0000)]
Hide a couple of unguarded error returns behind the no_fail test.

22 years agoFree old_pwd only in the code path where it has been allocated.
Juli Mallett [Wed, 22 May 2002 23:18:25 +0000 (23:18 +0000)]
Free old_pwd only in the code path where it has been allocated.

Reviewed by: des

22 years agoTaking a leap of faith, tie the help command in to the build.
Juli Mallett [Wed, 22 May 2002 23:15:58 +0000 (23:15 +0000)]
Taking a leap of faith, tie the help command in to the build.

22 years agoAdd my PD implementation of the SCCS help command, which prints help from
Juli Mallett [Wed, 22 May 2002 23:14:17 +0000 (23:14 +0000)]
Add my PD implementation of the SCCS help command, which prints help from
files in the format used by SCCS, given a key.  It behaves exactly like the
``proper'' SCCS help command, from what testing I can do.

22 years agoNew release notes: asa(1), pathchk(1), sccs(1),
Bruce A. Mah [Wed, 22 May 2002 23:03:43 +0000 (23:03 +0000)]
New release notes:  asa(1), pathchk(1), sccs(1),
uuencode(1)/uudecode(1) base64 synonyms, ptx removal.

Modified release notes:  ncurses 5.2-20020518.

22 years agoSync with NetBSD. (Non functional changes).
Josef Karthauser [Wed, 22 May 2002 23:01:13 +0000 (23:01 +0000)]
Sync with NetBSD. (Non functional changes).

22 years agoFix a bug: Use USBD_DEFAULT_INTERVAL instead of USBD_DEFAULT_TIMEOUT
Josef Karthauser [Wed, 22 May 2002 22:58:38 +0000 (22:58 +0000)]
Fix a bug: Use USBD_DEFAULT_INTERVAL instead of USBD_DEFAULT_TIMEOUT
in a call to usbd_open_pipe_ival.

22 years agoIncrease the maximum FSM option length to 50
Brian Somers [Wed, 22 May 2002 21:17:13 +0000 (21:17 +0000)]
Increase the maximum FSM option length to 50

22 years agoAdd some missing #includes that weren't required due to namespace polution
Brian Somers [Wed, 22 May 2002 21:08:58 +0000 (21:08 +0000)]
Add some missing #includes that weren't required due to namespace polution
in our headers.

Submitted by: bde

22 years agoRename pause() to ia32_pause() so it doesn't conflict with the pause()
John Baldwin [Wed, 22 May 2002 20:32:39 +0000 (20:32 +0000)]
Rename pause() to ia32_pause() so it doesn't conflict with the pause()
function defined in <unistd.h>.  I didn't #ifdef _KERNEL it because the
mutex implementation in libpthread will probably need this.

22 years agoRestore us back to the rev 1.324 level of having an Intel gigE driver.
David E. O'Brien [Wed, 22 May 2002 19:00:48 +0000 (19:00 +0000)]
Restore us back to the rev 1.324 level of having an Intel gigE driver.

22 years agoUpdate 164SX section with newly acquired insights.
Wilko Bulte [Wed, 22 May 2002 18:09:37 +0000 (18:09 +0000)]
Update 164SX section with newly acquired insights.

Submitted by: Sten <sten@blinkenlights.nl>

22 years agoAdd a warning regarding localhost-only listening daemons inside jails.
Gregory Neil Shapiro [Wed, 22 May 2002 16:37:32 +0000 (16:37 +0000)]
Add a warning regarding localhost-only listening daemons inside jails.
Apparently binding only to 127.0.0.1 inside of a jail actually binds
to the jail IP address as well (in effect, bind to all available
interfaces in the jail).

Submitted by: Helge Oldach <test-smtp@oldach.net>
MFC after: 1 day
pending RE approval

22 years ago1/2assed reimplementation of c-common.c revs 1.2 (-fformat-extensions)
David E. O'Brien [Wed, 22 May 2002 16:37:09 +0000 (16:37 +0000)]
1/2assed reimplementation of c-common.c revs 1.2 (-fformat-extensions)
and 1.3 (printf0) for GCC 3.1.

22 years agoTie sccs(1) in to the build, as it now does one thing right: sccs what
Juli Mallett [Wed, 22 May 2002 16:19:31 +0000 (16:19 +0000)]
Tie sccs(1) in to the build, as it now does one thing right: sccs what

22 years agoRemove mention of the GNU version of ptx, it is dead.
Juli Mallett [Wed, 22 May 2002 16:18:14 +0000 (16:18 +0000)]
Remove mention of the GNU version of ptx, it is dead.

22 years agoAdd a usage().
Juli Mallett [Wed, 22 May 2002 16:09:52 +0000 (16:09 +0000)]
Add a usage().

Print usage() if right before executing the specified command, it comes to be
that *argv is NULL (i.e. a flag was specified without a command being given).

22 years agoRemove #ifndef's on V6.
Juli Mallett [Wed, 22 May 2002 16:01:50 +0000 (16:01 +0000)]
Remove #ifndef's on V6.

22 years agoPut braces around the command table properly.
Juli Mallett [Wed, 22 May 2002 15:59:27 +0000 (15:59 +0000)]
Put braces around the command table properly.

22 years agoUse what(1) here, in /usr/bin/what. It doesn't print error 26 correctly, but
Juli Mallett [Wed, 22 May 2002 15:44:29 +0000 (15:44 +0000)]
Use what(1) here, in /usr/bin/what.  It doesn't print error 26 correctly, but
neither does the pd sccs(1) implementation I have around, so there's no loss
for now.

22 years agostr_concat() doesn't really take const arguments.
Juli Mallett [Wed, 22 May 2002 15:34:00 +0000 (15:34 +0000)]
str_concat() doesn't really take const arguments.

Submitted by: bde
Pointy hat to: jmallett

22 years agoDo not run shell from /bin, run it from $PATH.
Ruslan Ermilov [Wed, 22 May 2002 14:53:18 +0000 (14:53 +0000)]
Do not run shell from /bin, run it from $PATH.
Bump MAKE_VERSION to 5200205221.

22 years agoAdded the MAKE_VERSION global that could be useful in determining
Ruslan Ermilov [Wed, 22 May 2002 14:35:47 +0000 (14:35 +0000)]
Added the MAKE_VERSION global that could be useful in determining
if a given make(1) is feature-compatible with a set of makefiles.

When merged, this will be used to replace the ugly upgrade_checks
hacks in src/Makefile.

Version has the RYYYYMMDDX format, where R is from RELENG_<R> and
X allows for 10 distinguishable changes per day.

Discussed with: bde

22 years agoRepair a bug where type AAAA answer records were not displayed.
Jacques Vidrine [Wed, 22 May 2002 14:27:35 +0000 (14:27 +0000)]
Repair a bug where type AAAA answer records were not displayed.
Reference ISC BIND issue:
``1279.   [bug]           nslookup: partial coversion to similar style outputs
                          for both -type=aaaa and -type=a.''

Reported by: ume

22 years agoDo not attempt to free static memory.
Jacques Vidrine [Wed, 22 May 2002 14:24:44 +0000 (14:24 +0000)]
Do not attempt to free static memory.
Reference ISC BIND issue:
``1278.   [bug]           free() of non-malloced memory in nslookup.''

Reported by: phkmalloc, Mike Tancsa <mike@sentex.net>

22 years agoFormat internal commands properly.
Juli Mallett [Wed, 22 May 2002 14:17:16 +0000 (14:17 +0000)]
Format internal commands properly.

Kill a bit of trailing whitespace.

Fix a path format.

Submitted by: mdoc(7) police (ru)

22 years agoClean up the manual page by leaps and bounds in terms of formatting.
Juli Mallett [Wed, 22 May 2002 13:59:52 +0000 (13:59 +0000)]
Clean up the manual page by leaps and bounds in terms of formatting.

Ruslan's version took away the '.Nm' for some commands, but not others, so
I chose to go with leaving '.Nm'.

Submitted by: ru

22 years agoMake the guarded string functions take a const "from" addr, and make the
Juli Mallett [Wed, 22 May 2002 13:41:08 +0000 (13:41 +0000)]
Make the guarded string functions take a const "from" addr, and make the
function that prints when a botched guarded string operation occurs take
two const arguments.

XXX Should we use strlcat/strlcpy instead and hope for the best?

22 years agoDebug registers aren't selectors, so use saner names for the variables in
John Baldwin [Wed, 22 May 2002 13:29:18 +0000 (13:29 +0000)]
Debug registers aren't selectors, so use saner names for the variables in
the inline functions for reading and writing the debug registers.

22 years ago- Sort the pause() inline into the appropriate location.
John Baldwin [Wed, 22 May 2002 13:27:05 +0000 (13:27 +0000)]
- Sort the pause() inline into the appropriate location.
- Add many missing prototypes to the non-GCC section.

22 years agoRename cpu_pause() to pause(). Originally I was going to make this an
John Baldwin [Wed, 22 May 2002 13:19:22 +0000 (13:19 +0000)]
Rename cpu_pause() to pause().  Originally I was going to make this an
MI API with empty cpu_pause() functions on other arch's, but this
functionality is definitely unique to IA-32, so I decided to leave it
as i386-only and wrap it in #ifdef's.  I should have dropped the cpu_
prefix when I made that decision.

Requested by: bde

22 years agoRemove register qualifier.
Juli Mallett [Wed, 22 May 2002 12:32:54 +0000 (12:32 +0000)]
Remove register qualifier.

22 years agoUse function prototypes.
Juli Mallett [Wed, 22 May 2002 12:31:40 +0000 (12:31 +0000)]
Use function prototypes.

22 years agoPSD:19 (curses) is long dead.
Ruslan Ermilov [Wed, 22 May 2002 11:58:34 +0000 (11:58 +0000)]
PSD:19 (curses) is long dead.

22 years agoUnencumber.
Ruslan Ermilov [Wed, 22 May 2002 11:57:29 +0000 (11:57 +0000)]
Unencumber.

22 years agoBuild sccs(1) docs.
Ruslan Ermilov [Wed, 22 May 2002 11:55:47 +0000 (11:55 +0000)]
Build sccs(1) docs.

22 years ago19.curses wasn't encumbered, it's just dead.
Ruslan Ermilov [Wed, 22 May 2002 11:45:23 +0000 (11:45 +0000)]
19.curses wasn't encumbered, it's just dead.

22 years agoClean up formatting.
Juli Mallett [Wed, 22 May 2002 11:29:21 +0000 (11:29 +0000)]
Clean up formatting.

22 years agoReplace the evil that is __DECONST() with (void *). This is one of the least
Juli Mallett [Wed, 22 May 2002 11:16:48 +0000 (11:16 +0000)]
Replace the evil that is __DECONST() with (void *).  This is one of the least
evil things we can do involving the const qualifier and a pointer.

Submitted by: bde, ru

22 years agoMake sccs(1) compile cleanly by fixing syntax errors such as #endif followed
Juli Mallett [Wed, 22 May 2002 11:10:03 +0000 (11:10 +0000)]
Make sccs(1) compile cleanly by fixing syntax errors such as #endif followed
by a token, and by including headers to get prototypes for many things, and
also by spelling the type of structure readdir(3) returns as "dirent".

22 years agoUnroff all forms of \f and \*, and the simplest form of \s.
Ruslan Ermilov [Wed, 22 May 2002 11:08:41 +0000 (11:08 +0000)]
Unroff all forms of \f and \*, and the simplest form of \s.

Submitted by: fenner, ru
Reviewed by: ru, fenner

22 years agoRevision 1.39 made filtering of CFLAGS unnecessary.
Ruslan Ermilov [Wed, 22 May 2002 11:03:35 +0000 (11:03 +0000)]
Revision 1.39 made filtering of CFLAGS unnecessary.

22 years agoRevive the SCCS front-end, sccs(1), back from the Attic, since it's something
Juli Mallett [Wed, 22 May 2002 10:43:50 +0000 (10:43 +0000)]
Revive the SCCS front-end, sccs(1), back from the Attic, since it's something
SUS wants (they want the entire suite of SCCS commands, we don't have them,
but we *had* the frontend, so we can have it again, and now).

Add $FreeBSD$ where appropriate, don't revive PSD.doc/spell.ok.

22 years agoFixed build of recently added legacy docs.
Ruslan Ermilov [Wed, 22 May 2002 10:41:12 +0000 (10:41 +0000)]
Fixed build of recently added legacy docs.

Reviewed by: grog

22 years agoAdded share/doc/[pu]sd entries.
Ruslan Ermilov [Wed, 22 May 2002 10:38:25 +0000 (10:38 +0000)]
Added share/doc/[pu]sd entries.

Reviewed by: grog

22 years agoEliminate free'ing already freed chunks.
Maxim Sobolev [Wed, 22 May 2002 10:33:25 +0000 (10:33 +0000)]
Eliminate free'ing already freed chunks.

MFC after: 1 week

22 years agoBuild pathchk(1).
Tim J. Robbins [Wed, 22 May 2002 10:32:24 +0000 (10:32 +0000)]
Build pathchk(1).

22 years agoAdd the pathchk utility, which checks pathnames for validity or
Tim J. Robbins [Wed, 22 May 2002 10:30:16 +0000 (10:30 +0000)]
Add the pathchk utility, which checks pathnames for validity or
portability between POSIX systems.

Submitted by: Chuck Rouillard (manpage, initial implementation)

22 years agoQuick fix for non-unique inode numbers for hard links. We use the
Bruce Evans [Wed, 22 May 2002 08:50:18 +0000 (08:50 +0000)]
Quick fix for non-unique inode numbers for hard links.  We use the
byte offset of the directory entry for the inode number for all types
of files except directories, although this breaks hard links for
non-directories even if it doesn't cause overflow.  Just ignore this
broken inode number for stat() and readdir() and return a less broken
one (the block offset of the file), so that applications normally can't
see the brokenness.

This leaves at least the following brokenness:
- extra inodes, vnodes and caching for hard links.
- various overflow bugs.  cd9660 supports 64-bit block numbers, but we
  silently ignore the top 32 bits in isonum_733() and then drop another
  10 bits for our broken inode numbers.  We may also have sign extension
  bugs from storing 32-bit extents in ints and longs even if ints are
  32-bits.  These bugs affect DVDs.  mkisofs apparently limits them
  by writing directory entries first.

Inode numbers were broken mainly in 4.4BSD-Lite2.  FreeBSD-1.1.5 seems
to have a correct implementation modulo the overflow bugs.  We need
to look up directory entries from inodes for symlinks only.  FreeBSD-1.1.5
use separate fields (iso_parent_extent, iso_parent) to point to the
directory entry.  4.4BSD-Lite doesn't have these, and abuses i_ino to
point to the directory entry.  Correct pointers are impossible for
hard links, but symlinks can't be hard links.

22 years agoInclude <sys.systm.h> for the declaration of some atomic functions -- don't
Bruce Evans [Wed, 22 May 2002 06:26:44 +0000 (06:26 +0000)]
Include <sys.systm.h> for the declaration of some atomic functions -- don't
depend on namespace pollution in <sys/mutex.h>.

22 years agoBring back the cd -L and -P options from revision 1.24, but try harder not
Tim J. Robbins [Wed, 22 May 2002 05:15:53 +0000 (05:15 +0000)]
Bring back the cd -L and -P options from revision 1.24, but try harder not
to fail when the logical current directory no longer exists. Allow changes
to absolute paths when logical cwd is invalid, fall back to physical cd
if logical cd fails.

22 years agoTemporarily back out revision 1.24; it seems to handle the case where the
Tim J. Robbins [Wed, 22 May 2002 03:29:20 +0000 (03:29 +0000)]
Temporarily back out revision 1.24; it seems to handle the case where the
current directory no longer exists incorrectly and breaks `make cleandir'.

22 years agoo Add records for PCI bus and PCI device errors.
Marcel Moolenaar [Wed, 22 May 2002 02:25:06 +0000 (02:25 +0000)]
o  Add records for PCI bus and PCI device errors.
o  Rename mem_platform_id to mem_oem_id.
o  Minor style fixes.

22 years agoDon't build doc on ia64. No groff in sight.
Marcel Moolenaar [Wed, 22 May 2002 01:04:42 +0000 (01:04 +0000)]
Don't build doc on ia64. No groff in sight.

22 years ago o Make contigmalloc1() static.
Alan Cox [Wed, 22 May 2002 01:01:37 +0000 (01:01 +0000)]
 o Make contigmalloc1() static.

22 years agoPermit alternative break sequence to break to debugger in GENERIC. Breakage
Robert Watson [Tue, 21 May 2002 23:35:51 +0000 (23:35 +0000)]
Permit alternative break sequence to break to debugger in GENERIC.  Breakage
of serial break on -CURRENT seems rampant for some reason, and I like
being able to get into ddb.

Reviewed by: peter

22 years agoAdd appropriate IA32 "pause" instructions to improve performanec on
John Baldwin [Tue, 21 May 2002 22:26:35 +0000 (22:26 +0000)]
Add appropriate IA32 "pause" instructions to improve performanec on
Pentium 4's and newer IA32 processors.  The "pause" instruction has been
verified by Intel to be a NOP on all currently existing IA32 processors
prior to the Pentium 4.

22 years ago- td will never be NULL, so the call to soalloc() in socreate() will always
Andrew R. Reiter [Tue, 21 May 2002 21:30:44 +0000 (21:30 +0000)]
- td will never be NULL, so the call to soalloc() in socreate() will always
  be passed a 1; we can, however, use M_NOWAIT to indicate this.
- Check so against NULL since it's a pointer to a structure.

22 years agoFix an old cut 'n' paste bug inherited from BSD/OS: don't increment 'i'
John Baldwin [Tue, 21 May 2002 21:27:05 +0000 (21:27 +0000)]
Fix an old cut 'n' paste bug inherited from BSD/OS: don't increment 'i'
twice once we are in the long wait stage of spinning on a spin mutex.

22 years ago- OR the flag variable with M_ZERO so that the uma_zalloc() handles the
Andrew R. Reiter [Tue, 21 May 2002 21:18:41 +0000 (21:18 +0000)]
- OR the flag variable with M_ZERO so that the uma_zalloc() handles the
  zero'ing out of the allocated memory.  Also removed the logical bzero
  that followed.

22 years agoWhitespace fixup, properly indent the body of an else clause.
John Baldwin [Tue, 21 May 2002 21:13:27 +0000 (21:13 +0000)]
Whitespace fixup, properly indent the body of an else clause.

22 years agoAdd code to make default mutexes adaptive if the ADAPTIVE_MUTEXES kernel
John Baldwin [Tue, 21 May 2002 20:47:11 +0000 (20:47 +0000)]
Add code to make default mutexes adaptive if the ADAPTIVE_MUTEXES kernel
option is used (not on by default).

- In the case of trying to lock a mutex, if the MTX_CONTESTED flag is set,
  then we can safely read the thread pointer from the mtx_lock member while
  holding sched_lock.  We then examine the thread to see if it is currently
  executing on another CPU.  If it is, then we keep looping instead of
  blocking.
- In the case of trying to unlock a mutex, it is now possible for a mutex
  to have MTX_CONTESTED set in mtx_lock but to not have any threads
  actually blocked on it, so we need to handle that case.  In that case,
  we just release the lock as if MTX_CONTESTED was not set and return.
- We do not adaptively spin on Giant as Giant is held for long times and
  it slows SMP systems down to a crawl (it was taking several minutes,
  like 5-10 or so for my test alpha and sparc64 SMP boxes to boot up when
  they adaptively spinned on Giant).
- We only compile in the code to do this for SMP kernels, it doesn't make
  sense for UP kernels.

Tested on: i386, alpha, sparc64

22 years agoUpdate to match kernel side.
Poul-Henning Kamp [Tue, 21 May 2002 20:42:51 +0000 (20:42 +0000)]
Update to match kernel side.

Sponsored by: DARPA & NAI Labs.

22 years agoOptimize spin mutexes for UP kernels without debugging to just enter and
John Baldwin [Tue, 21 May 2002 20:34:28 +0000 (20:34 +0000)]
Optimize spin mutexes for UP kernels without debugging to just enter and
exit critical sections.  We only contest on a spin mutex on an SMP kernel
running on an SMP machine.

22 years agoIntroduce the concept of "magic spaces", and implement them in most of
Poul-Henning Kamp [Tue, 21 May 2002 20:33:49 +0000 (20:33 +0000)]
Introduce the concept of "magic spaces", and implement them in most of
the relevant classes.

Some methods may implement various "magic spaces", this is reserved
or magic areas on the disk, set a side for various and sundry purposes.
A good example is the BSD disklabel and boot code on i386 which occupies
a total of four magic spaces: boot1, the disklabel, the padding behind
the disklabel and boot2.  The reason we don't simply tell people to
write the appropriate stuff on the underlying device is that (some of)
the magic spaces might be real-time modifiable.  It is for instance
possible to change a disklabel while partitions are open, provided
the open partitions do not get trampled in the process.

Sponsored by: DARPA & NAI Labs.