]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
22 years agoFix a problem in the disk related hack where device nodes for a physically
phk [Sun, 28 Oct 2001 09:39:28 +0000 (09:39 +0000)]
Fix a problem in the disk related hack where device nodes for a physically
non-existent disk in a legacy /dev on a DEVFS system would panic the system
if stat(2)'ed.

Do not whine about anonymous device nodes not having a si_devsw, they're
not supposed to.

22 years agoIntroduce [IPC|SHM]_[INFO|STAT] to shmctl to make
mr [Sun, 28 Oct 2001 09:29:10 +0000 (09:29 +0000)]
Introduce [IPC|SHM]_[INFO|STAT] to shmctl to make
`/compat/linux/usr/bin/ipcs -m` happy.

22 years agodeltat declared time_t, msg("") call used %d (assumed time_t == int).
dillon [Sun, 28 Oct 2001 06:13:47 +0000 (06:13 +0000)]
deltat declared time_t, msg("") call used %d (assumed time_t == int).
Changed deltat to be an int (result of delta time calculation).

MFC after: 1 day

22 years agofixed to support pc98
nyan [Sun, 28 Oct 2001 04:39:02 +0000 (04:39 +0000)]
fixed to support pc98

22 years agoRemove duplicate apm entry for pc98
nyan [Sun, 28 Oct 2001 04:34:24 +0000 (04:34 +0000)]
Remove duplicate apm entry for pc98

22 years agoRemoved pmc_isa_identify function. It is not needed.
nyan [Sun, 28 Oct 2001 04:16:50 +0000 (04:16 +0000)]
Removed pmc_isa_identify function. It is not needed.

Submitted by: takawata

22 years agoUse a quad int conversion routine for time_t instead of ulong, capable of
dillon [Sun, 28 Oct 2001 02:51:43 +0000 (02:51 +0000)]
Use a quad int conversion routine for time_t instead of ulong, capable of
decoding tar's full time buffer (12 digits).

22 years agoVarious string fields for certain cpio types are 11 rather then 8
dillon [Sun, 28 Oct 2001 02:45:04 +0000 (02:45 +0000)]
Various string fields for certain cpio types are 11 rather then 8
characters.  Use quad conversion functions rather then long conversion
where appropriate to handle the available range.  Mainly fixes time_t
but there was also a st_size ulong conversion in there that has to be
quad or cpio cannot be used to copy files > 2G.

MFC after: 1 day

22 years agoRemove unnecessary casts in timeval.tv_sec load from st_*time
dillon [Sun, 28 Oct 2001 02:42:26 +0000 (02:42 +0000)]
Remove unnecessary casts in timeval.tv_sec load from st_*time

22 years agoRemove assumption that time_t is an int
dillon [Sun, 28 Oct 2001 02:28:04 +0000 (02:28 +0000)]
Remove assumption that time_t is an int

MFC after: 1 day

22 years agoRemove some extraneous spaces from the usage message.
joe [Sun, 28 Oct 2001 02:10:40 +0000 (02:10 +0000)]
Remove some extraneous spaces from the usage message.

22 years agosparc64 verions of the crt initialization and finalization files required
obrien [Sun, 28 Oct 2001 00:20:34 +0000 (00:20 +0000)]
sparc64 verions of the crt initialization and finalization files required
by the ELF ABI.

22 years agoAdd missing cast for improper time_t use in computation.
dillon [Sun, 28 Oct 2001 00:01:58 +0000 (00:01 +0000)]
Add missing cast for improper time_t use in computation.

22 years agoll_time and ut_time are int32_t on disk and have to remain that way
peter [Sat, 27 Oct 2001 20:40:54 +0000 (20:40 +0000)]
ll_time and ut_time are int32_t on disk and have to remain that way
unless we dont want to run 4.x binaries any more.

22 years agosc_lasttime and sc_starttime are time_t's, not long's.
dillon [Sat, 27 Oct 2001 20:31:24 +0000 (20:31 +0000)]
sc_lasttime and sc_starttime are time_t's, not long's.

22 years agoOnly provide function information in compile environments that support
mike [Sat, 27 Oct 2001 20:11:10 +0000 (20:11 +0000)]
Only provide function information in compile environments that support
the C99 variable __func__ and never for C++.  Provide a more meaningful
example in the assert(3) manual.

Reviewed by: asmodai, bde

22 years agosyncdelay, filedelay, dirdelay, metadelay are ints, not time_t's,
dillon [Sat, 27 Oct 2001 19:58:56 +0000 (19:58 +0000)]
syncdelay, filedelay, dirdelay, metadelay are ints, not time_t's,
and can also be made static.

22 years agoRestore K&R prototype. Fix other style bugs.
mike [Sat, 27 Oct 2001 19:37:20 +0000 (19:37 +0000)]
Restore K&R prototype.  Fix other style bugs.

Reviewed by: asmodai, bde

22 years agoNudge the axe a bit closer to cdevsw[]:
phk [Sat, 27 Oct 2001 17:44:21 +0000 (17:44 +0000)]
Nudge the axe a bit closer to cdevsw[]:

   Make it a panic to repeat make_dev() or destroy_dev(), this check
   should maybe be neutered when -current goes -stable.

   Whine if devsw() is called on anon dev_t's in a devfs system.

   Make a hack to avoid our lazy-eval disk code triggering the above whine.

   Fix the multiple make_dev() in disk code by making ${disk}${unit}s${slice}
   an alias/symlink to ${disk}${unit}s${slice}c

22 years agoDiff-reduce these two.
markm [Sat, 27 Oct 2001 12:49:19 +0000 (12:49 +0000)]
Diff-reduce these two.

Really, one of them needs to disappear. I'll figure out which
later.

Reported by: bde

22 years agoEliminate the prefix parameter to linux_emul_find(), which was always
des [Sat, 27 Oct 2001 11:15:19 +0000 (11:15 +0000)]
Eliminate the prefix parameter to linux_emul_find(), which was always
linux_emul_path anyway.  Linux_emul_find() has interesting bugs in its
prefix handling (which luckily are not currently exploitable); this
commit is preliminary to an attempt at cleaning it up.

Approved by: marcel

22 years agoAdd a P_INEXEC flag that indicates that the process has called execve() and
des [Sat, 27 Oct 2001 11:11:25 +0000 (11:11 +0000)]
Add a P_INEXEC flag that indicates that the process has called execve() and
it has not yet returned.  Use this flag to deny debugging requests while
the process is execve()ing, and close once and for all any race conditions
that might occur between execve() and various debugging interfaces.

Reviewed by: jhb, rwatson

22 years agoAdd missing crti.S and crtn.S files. I have tested these with -static
peter [Sat, 27 Oct 2001 10:10:45 +0000 (10:10 +0000)]
Add missing crti.S and crtn.S files.  I have tested these with -static
linking only.  They may require some gp relative tweaks for dynamic use.

22 years agoSync with other platforms.
obrien [Sat, 27 Oct 2001 08:32:07 +0000 (08:32 +0000)]
Sync with other platforms.

22 years agominor style updating
obrien [Sat, 27 Oct 2001 08:30:36 +0000 (08:30 +0000)]
minor style updating

22 years agoUpdate for reality and syncing with other FreeBSD platforms.
obrien [Sat, 27 Oct 2001 08:29:03 +0000 (08:29 +0000)]
Update for reality and syncing with other FreeBSD platforms.

22 years agoFeh, it helps to include a file other than yourself... wonder how
obrien [Sat, 27 Oct 2001 08:25:32 +0000 (08:25 +0000)]
Feh, it helps to include a file other than yourself... wonder how
GCC cpp liked that.

Properly include the branding info.

22 years agoTell how the import was done.
obrien [Sat, 27 Oct 2001 08:11:44 +0000 (08:11 +0000)]
Tell how the import was done.

22 years agoThis commit was generated by cvs2svn to compensate for changes in r85587,
obrien [Sat, 27 Oct 2001 08:07:37 +0000 (08:07 +0000)]
This commit was generated by cvs2svn to compensate for changes in r85587,
which included commits to RCS files with non-trunk default branches.

22 years agoImport the One True AWK, in the form of bwk's AWK Nov 15, 2000 release.
obrien [Sat, 27 Oct 2001 08:07:37 +0000 (08:07 +0000)]
Import the One True AWK, in the form of bwk's AWK Nov 15, 2000 release.

22 years agoImplement elf_reloc. This makes klds work.
jake [Sat, 27 Oct 2001 07:09:40 +0000 (07:09 +0000)]
Implement elf_reloc.  This makes klds work.

Obtained from: netbsd

22 years agoHandle instruction access mmu miss faults in kernel mode. These can only
jake [Sat, 27 Oct 2001 07:06:15 +0000 (07:06 +0000)]
Handle instruction access mmu miss faults in kernel mode.  These can only
be generated by non-preloaded klds.

22 years agoo Modify copyright date format for improved consistency: use '-' for
rwatson [Sat, 27 Oct 2001 05:47:47 +0000 (05:47 +0000)]
o Modify copyright date format for improved consistency: use '-' for
  year ranges, rather than a comma-delimited list of years.

22 years agoo Update copyright dates.
rwatson [Sat, 27 Oct 2001 05:46:43 +0000 (05:46 +0000)]
o Update copyright dates.

Obtained from: TrustedBSD Project

22 years agoo Improve style(9) compliance following KSE modifications. In particular,
rwatson [Sat, 27 Oct 2001 05:45:42 +0000 (05:45 +0000)]
o Improve style(9) compliance following KSE modifications.  In particular,
  strip the space from '( struct thread *...', wrap long lines.
o Remove an unneeded comment on the topic of no lock being required as
  part of the NDINIT() in __acl_get_file(), as it's really not required
  there.

Obtained from: TrustedBSD Project

22 years agoo Althought this is not specified in POSIX.1e, the UFS ACL implementation
rwatson [Sat, 27 Oct 2001 05:39:17 +0000 (05:39 +0000)]
o Althought this is not specified in POSIX.1e, the UFS ACL implementation
  coerces the deletion of a default ACL on a directory when no default
  ACL EA is present to success.  Because the UFS EA implementation doesn't
  disinguish the EA failure modes "that EA name has not been
  administratively enabled" from "that EA name has no defined data",
  there's a potential conflict in error return values.  Normally, the
  lack of administratively configured EA support is coerced to
  EOPNOTSUPP to indicate that ACLs are not available; in this case,
  it is possible to get a successful return, even if ACLs are not
  available because EA support for them has not been enabled.

  Expand the comment in ufs_setacl() to identify this case.

Obtained from: TrustedBSD Project

22 years agoo Clarify a comment about the locking condition of the vnode upon exit
rwatson [Sat, 27 Oct 2001 05:19:14 +0000 (05:19 +0000)]
o Clarify a comment about the locking condition of the vnode upon exit
  from ufs_extattr_enable_with_open().
o Print auto-start notifications if (bootverbose).  This was previously
  commented out since it didn't know how to check for bootverbose.
o Drop in comments throughout indicating where ENOENT should be replaced
  with ENOATTR once that is available.

Obtained from: TrustedBSD Project

22 years agoo The comment about ordering the destruction of the lock and the removal of
rwatson [Sat, 27 Oct 2001 05:05:39 +0000 (05:05 +0000)]
o The comment about ordering the destruction of the lock and the removal of
  the flag indicating that the structure was initialized didn't need
  an XXX, since it didn't need fixing.

Obtained from: TrustedBSD Project

22 years agoo Wrap a number of long lines of code, many of which were introduced
rwatson [Sat, 27 Oct 2001 05:03:05 +0000 (05:03 +0000)]
o Wrap a number of long lines of code, many of which were introduced
  due to KSE-related (p) expansions.

Obtained from: TrustedBSD Project

22 years agoSince namespace support was added to the UFS extended attribute
rwatson [Sat, 27 Oct 2001 04:58:28 +0000 (04:58 +0000)]
Since namespace support was added to the UFS extended attribute
implementation to replace single-character namespace prefixes, '$' is no
longer an invalid attribute name, and the namespace is relevant to
validity determination.

o Remove '$' case from ufs_extattr_valid_attrname()
o Add attrnamespace argument to ufs_extattr_valid_attrname(), and
  fill out appropriately.

Currently no decisions are made based on the namespace argument, but
may be in the future.

Obtained from: TrustedBSD Project

22 years agoo Spelling error s/suffient/sufficient/
rwatson [Sat, 27 Oct 2001 03:14:37 +0000 (03:14 +0000)]
o Spelling error s/suffient/sufficient/

PR: 31524
Submitted by: Dan Langille <dan@freebsddiary.org>

22 years agoCorrect James Cook's year of birth. I wonder how anybody could expect
grog [Sat, 27 Oct 2001 02:51:59 +0000 (02:51 +0000)]
Correct James Cook's year of birth.  I wonder how anybody could expect
him to have been born in the 15th century.

22 years agoTurn on table-of-contents generation for the release notes and
bmah [Sat, 27 Oct 2001 02:41:29 +0000 (02:41 +0000)]
Turn on table-of-contents generation for the release notes and
hardware notes.

Requested by: many

22 years agoExplicitly use int32_t for on-disk records for pw_change and pw_expire,
peter [Sat, 27 Oct 2001 02:13:41 +0000 (02:13 +0000)]
Explicitly use int32_t for on-disk records for pw_change and pw_expire,
since that is what we use now and this insulates us from any time_t
tweaks here.  We can define a record format that uses 64 bit times if/when
we need to.

22 years agoEnable round-robin arbitration between transmit and receive unit
luigi [Sat, 27 Oct 2001 00:59:17 +0000 (00:59 +0000)]
Enable round-robin arbitration between transmit and receive unit
in the 21143, instead of giving priority to the receive unit.
This gives a 10-15% performance improvement in the forwarding rate
under heavy load.

Reviewed-by: Bill Paul
22 years agoSupport the "install.debug" and "reinstall.debug" targets for kernel modules.
des [Sat, 27 Oct 2001 00:52:50 +0000 (00:52 +0000)]
Support the "install.debug" and "reinstall.debug" targets for kernel modules.
Small tweaks to kldxref may be necessary to avoid the surprising (but harm-
less) behaviour of 'kldload foo' loading foo.ko.debug instead of foo.ko if
it is present in the kernel directory.

Approved by: a week of silence on -arch
MFC after: 2 weeks

22 years agoForce the length of the sockaddr to be correct for AF_INET and AF_INET6
fenner [Fri, 26 Oct 2001 23:10:08 +0000 (23:10 +0000)]
Force the length of the sockaddr to be correct for AF_INET and AF_INET6
in bind() and connect().  Linux doesn't care if the length of the
sockaddr matches its address family; FreeBSD does.  This fixes the
known issues with the resolver in linux_base-7.

22 years agoRemove the -I../libc/${MACHINE_ARCH} that was there solely for
peter [Fri, 26 Oct 2001 21:20:23 +0000 (21:20 +0000)]
Remove the -I../libc/${MACHINE_ARCH} that was there solely for
the #include "DEFS.h" that was only used on i386 (which has been fixed).

22 years agoMake libc_r check the kern.usrstack sysctl instead of using internal
peter [Fri, 26 Oct 2001 21:19:22 +0000 (21:19 +0000)]
Make libc_r check the kern.usrstack sysctl instead of using internal
kernel #defines to figure out where the stack is located.  This stops
libc_r from exploding when the kernel is compiled with a different
KVM size.  IMHO this is all kinda bogus, it would be better to just
check %esp and work from that.

22 years agoAdd 'ciss'.
msmith [Fri, 26 Oct 2001 21:17:26 +0000 (21:17 +0000)]
Add 'ciss'.

22 years agoAdd mtx_lock_giant() and mtx_unlock_giant() wrappers for sysctl management
dillon [Fri, 26 Oct 2001 20:48:04 +0000 (20:48 +0000)]
Add mtx_lock_giant() and mtx_unlock_giant() wrappers for sysctl management
of Giant during the Giant unwinding phase, and start work on instrumenting
Giant for the file and proc mutexes.

These wrappers allow developers to turn on and off Giant around various
subsystems.  DEVELOPERS SHOULD NEVER TURN OFF GIANT AROUND A SUBSYSTEM JUST
BECAUSE THE SYSCTL EXISTS!  General developers should only considering
turning on Giant for a subsystem whos default is off (to help track down
bugs).  Only developers working on particular subsystems who know what
they are doing should consider turning off Giant.

These wrappers will greatly improve our ability to unwind Giant and test
the kernel on a (mostly) subsystem by subsystem basis.   They allow Giant
unwinding developers (GUDs) to emplace appropriate subsystem and structural
mutexes in the main tree and then request that the larger community test
the work by turning off Giant around the subsystem(s), without the larger
community having to mess around with patches.  These wrappers also allow
GUDs to boot into a (more likely to be) working system in the midst of
their unwinding work and to test that work under more controlled
circumstances.

There is a master sysctl, kern.giant.all, which defaults to 0 (off).  If
turned on it overrides *ALL* other kern.giant sysctls and forces Giant to
be turned on for all wrapped subsystems.  If turned off then Giant around
individual subsystems are controlled by various other kern.giant.XXX sysctls.

Code which overlaps multiple subsystems must have all related subsystem Giant
sysctls turned off in order to run without Giant.

22 years agominor commenting based on syscall environment
dillon [Fri, 26 Oct 2001 20:35:52 +0000 (20:35 +0000)]
minor commenting based on syscall environment

22 years agoIf an article in the release documentation has a table of contents,
bmah [Fri, 26 Oct 2001 20:26:59 +0000 (20:26 +0000)]
If an article in the release documentation has a table of contents,
make it go down to 3 section levels (the default depth for FreeBSD
articles is 1 level, but few articles have tables-of-contents
enabled).

22 years agoAdd VOP_IOCTL support, and fix a bug that would cause a panic if a file or
des [Fri, 26 Oct 2001 18:52:47 +0000 (18:52 +0000)]
Add VOP_IOCTL support, and fix a bug that would cause a panic if a file or
symlink lacked a filler function.

22 years ago- Change the taskqueue locking to protect the necessary parts of a task
jhb [Fri, 26 Oct 2001 18:46:48 +0000 (18:46 +0000)]
- Change the taskqueue locking to protect the necessary parts of a task
  while it is on a queue with the queue lock and remove the per-task locks.
- Remove TASK_DESTROY now that it is no longer needed.
- Go back to inlining TASK_INIT now that it is short again.

Inspired by: dfr

22 years agoStyle: sort __sys_foo() prototypes, tabs -> spaces, etc.
ru [Fri, 26 Oct 2001 18:45:02 +0000 (18:45 +0000)]
Style: sort __sys_foo() prototypes, tabs -> spaces, etc.

22 years agoRemoved:
ru [Fri, 26 Oct 2001 17:46:36 +0000 (17:46 +0000)]
Removed:

- uthread_signal.c; libc_r does not wrap signal() since 1998/04/29.

- uthread_attr_setprio.c; it was never connected to the build, and
  pthread_attr_setprio() does not exist in POSIX.

- uthread_sigblock.c and uthread_sigsetmask.c; these were no-ops
  bloating libc_r's space.

pthread_private.h:

- Removed prototypes of non-syscalls: send().

- Removed prototypes of unused syscalls: sigpending(), sigsuspend(),
  and select().

- Fixed prototype of fork().

- MFS: Fixed prototypes of <sys/socket.h> syscalls.

Reviewed by: deischen
Approved by: deischen, jasone

22 years agoAdd APM compatibility feature to ACPI.
iwasaki [Fri, 26 Oct 2001 17:43:05 +0000 (17:43 +0000)]
Add APM compatibility feature to ACPI.
This emulates APM device node interface APIs (mainly ioctl) and
provides APM services for the applications.  The goal is to support
most of APM applications without any changes.
Implemented ioctls in this commit are:
 - APMIO_SUSPEND (mapped ACPI S3 as default but changable by sysctl)
 - APMIO_STANDBY (mapped ACPI S1 as default but changable by sysctl)
 - APMIO_GETINFO and APMIO_GETINFO_OLD
 - APMIO_GETPWSTATUS

With above, many APM applications which get batteries, ac-line
info. and transition the system into suspend/standby mode (such as
wmapm, xbatt) should work with ACPI enabled kernel (if ACPI works well :-)

Reviewed by: arch@, audit@ and some guys

22 years agoRemove the internal implementation details of wrapping syscalls,
ru [Fri, 26 Oct 2001 17:38:20 +0000 (17:38 +0000)]
Remove the internal implementation details of wrapping syscalls,
which do not match the reality anyway.

Approved by: deischen, bde

22 years agoThis commit was generated by cvs2svn to compensate for changes in r85552,
ru [Fri, 26 Oct 2001 17:22:12 +0000 (17:22 +0000)]
This commit was generated by cvs2svn to compensate for changes in r85552,
which included commits to RCS files with non-trunk default branches.

22 years agoFrom ChangeLog:
ru [Fri, 26 Oct 2001 17:22:12 +0000 (17:22 +0000)]
From ChangeLog:

: 2001-10-19  Ruslan Ermilov  <ru@FreeBSD.org>
:
: * tmac/doc.tmac (doc-flag-recursion): Protect arguments against
: being handled as end-of-sentence characters,

This fixes ".Fl \&?".  Previously, it produced an additional whitespace.
There were no precedents in FreeBSD manpages.  That's why it was

Noticed by: Thomas Klausner <wiz@netbsd.org>

22 years agomdoc(7) police:
ru [Fri, 26 Oct 2001 17:11:50 +0000 (17:11 +0000)]
mdoc(7) police:

- Minor markup nits.
- Arrange for appearance in 4.5.

22 years agoThere are users of FreeBSD 4.5 already.
ru [Fri, 26 Oct 2001 17:01:15 +0000 (17:01 +0000)]
There are users of FreeBSD 4.5 already.

22 years agoDocument HPT372 support.
asmodai [Fri, 26 Oct 2001 16:46:10 +0000 (16:46 +0000)]
Document HPT372 support.

22 years agoFix title to conform to FreeBSD/port naming convention.
bmah [Fri, 26 Oct 2001 16:43:20 +0000 (16:43 +0000)]
Fix title to conform to FreeBSD/port naming convention.

22 years agoWe support the Intel ICH3 [ATA-100] chipset as well.
asmodai [Fri, 26 Oct 2001 16:34:26 +0000 (16:34 +0000)]
We support the Intel ICH3 [ATA-100] chipset as well.

22 years agoNew release notes: isic(4) support for Compaq Microcom 610 ISDN, fbtab(5)
bmah [Fri, 26 Oct 2001 16:28:36 +0000 (16:28 +0000)]
New release notes:  isic(4) support for Compaq Microcom 610 ISDN, fbtab(5)
globbing, rcmd(3) uses ${RSH}.

MFCs noted:  sysinstall(8) can load KLDs at install-time.

Fix typo for isdnphone(8).

22 years agoMove recently added procedure which was incorrectly placed within an
dillon [Fri, 26 Oct 2001 16:27:54 +0000 (16:27 +0000)]
Move recently added procedure which was incorrectly placed within an
#ifdef DDB block.

22 years agoDetach the prom console when platform.cons_init is called. This seems
mjacob [Fri, 26 Oct 2001 16:13:47 +0000 (16:13 +0000)]
Detach the prom console when platform.cons_init is called. This seems
to avoid most of the double character kernel goop we've been having by having
both a prom console && a normal console.

Was not able to test with graphics head. Hope this doesn't break anything.

Reviewed by: silence on alpha

22 years agoMake cdevsw[] static.
phk [Fri, 26 Oct 2001 15:31:22 +0000 (15:31 +0000)]
Make cdevsw[] static.

22 years agoReporting device drivers by traversing cdevsw[] is at best a hack
phk [Fri, 26 Oct 2001 15:30:44 +0000 (15:30 +0000)]
Reporting device drivers by traversing cdevsw[] is at best a hack
which may or may not return something which is partially right.

Disable the "devices" file until we find out what this is needed for,
and what exactly those apps need.

This will allow cdevsw to become static again.

Approved by: DES

22 years agomdoc(7) police:
ru [Fri, 26 Oct 2001 11:16:18 +0000 (11:16 +0000)]
mdoc(7) police:

- Bump document date.
- Remove hard sentence breaks.
- Fix markup.

22 years agoFinish the asbestos suited move of $mach/conf/*.$mach to conf/*.$mach.
ru [Fri, 26 Oct 2001 10:33:45 +0000 (10:33 +0000)]
Finish the asbestos suited move of $mach/conf/*.$mach to conf/*.$mach.
Fix some more typos.

22 years agoRemove /var/spool/uucp subtree, not needed for 'cu'
ache [Fri, 26 Oct 2001 09:40:28 +0000 (09:40 +0000)]
Remove /var/spool/uucp subtree, not needed for 'cu'

22 years agoCompensate for the way that _setjmp aligns the start of jmp_buf.
dfr [Fri, 26 Oct 2001 08:26:45 +0000 (08:26 +0000)]
Compensate for the way that _setjmp aligns the start of jmp_buf.

22 years agoClose buffer overflow noted in security advisory DSA-085.
jkh [Fri, 26 Oct 2001 08:25:32 +0000 (08:25 +0000)]
Close buffer overflow noted in security advisory DSA-085.
Submitted by: Koga Youichirou <y-koga@jp.FreeBSD.org>
Obtained from: Debian

22 years agoAdd a per-thread ucred reference for syscalls and synchronous traps from
jhb [Fri, 26 Oct 2001 08:12:54 +0000 (08:12 +0000)]
Add a per-thread ucred reference for syscalls and synchronous traps from
userland.  The per thread ucred reference is immutable and thus needs no
locks to be read.  However, until all the proc locking associated with
writes to p_ucred are completed, it is still not safe to use the per-thread
reference.

Tested on: x86 (SMP), alpha, sparc64

22 years agoFix nit in copyright.
obrien [Fri, 26 Oct 2001 06:45:10 +0000 (06:45 +0000)]
Fix nit in copyright.

22 years agoUse the new machine-independent versions of crtbegin and crtend
obrien [Fri, 26 Oct 2001 06:39:54 +0000 (06:39 +0000)]
Use the new machine-independent versions of crtbegin and crtend
from the "common" directory.

22 years agoAdd locking to taskqueues. There is one mutex per task, one mutex per
jhb [Fri, 26 Oct 2001 06:32:21 +0000 (06:32 +0000)]
Add locking to taskqueues.  There is one mutex per task, one mutex per
queue, and a mutex to protect the global list of taskqueues.  The only
visible change is that a TASK_DESTROY() macro has been added to mirror
the TASK_INIT() macro to destroy a task before it is free'd.

Submitted by: Andrew Reiter <awr@watson.org>

22 years agoUse msleep() to avoid lost wakeup's instead of doing an ineffective
jhb [Fri, 26 Oct 2001 06:09:01 +0000 (06:09 +0000)]
Use msleep() to avoid lost wakeup's instead of doing an ineffective
splhigh() before the mtx_unlock and tsleep().  The splhigh() was probably
correct in the original code using simplelocks but is not correct in
5.0-current.

Noticed by: Andrew Reiter <awr@FreeBSD.org>

22 years agoMinimal libc for sparc64.
jake [Fri, 26 Oct 2001 05:40:07 +0000 (05:40 +0000)]
Minimal libc for sparc64.

Reviewed by: obrien

22 years agoImplement kern.maxvnodes. adjusting kern.maxvnodes now actually has a
dillon [Fri, 26 Oct 2001 00:08:05 +0000 (00:08 +0000)]
Implement kern.maxvnodes.  adjusting kern.maxvnodes now actually has a
real effect.

Optimize vfs_msync().  Avoid having to continually drop and re-obtain
mutexes when scanning the vnode list.  Improves looping case by 500%.

Optimize ffs_sync().  Avoid having to continually drop and re-obtain
mutexes when scanning the vnode list.  This makes a couple of assumptions,
which I believe are ok, in regards to vnode stability when the mount list
mutex is held.  Improves looping case by 500%.

(more optimization work is needed on top of these fixes)

MFC after: 1 week

22 years agoThe same unbreakage (0755 -> 0775) for /var/games and subdirs
ache [Thu, 25 Oct 2001 23:18:41 +0000 (23:18 +0000)]
The same unbreakage (0755 -> 0775) for /var/games and subdirs

22 years agoAdd missing TAILQ_INSERT_TAIL's which somehow didn't get comitted with
dillon [Thu, 25 Oct 2001 23:13:56 +0000 (23:13 +0000)]
Add missing TAILQ_INSERT_TAIL's which somehow didn't get comitted with
the recent vnode cleanup.

22 years agoFix /var/mail, /var/rwho and /var/spool/lock back to 0775
ache [Thu, 25 Oct 2001 23:13:11 +0000 (23:13 +0000)]
Fix /var/mail, /var/rwho and /var/spool/lock back to 0775
Not sure about other dirs with the same damage (0755) by recent commit.

22 years agoIn > LONG_MAX test use sseek return value and not _offset which can be not
ache [Thu, 25 Oct 2001 22:56:04 +0000 (22:56 +0000)]
In > LONG_MAX test use sseek return value and not _offset which can be not
active.

22 years agoDefault to not performing ufs_dirhash's extensive directory-block
iedowse [Thu, 25 Oct 2001 22:55:59 +0000 (22:55 +0000)]
Default to not performing ufs_dirhash's extensive directory-block
sanity check after every directory modification. This check can be
re-enabled at any time by setting the sysctl "vfs.ufs.dirhash_docheck"
to 1.

This group of sanity tests was there to ensure that any UFS_DIRHASH
bugs could be caught by a panic before a potentially corrupted
directory block would be written to disk. It has served its main
purpose now, so disable it in the interest of performance.

MFC after: 1 week

22 years agoIn cluster_rbuild(), 'size' had better match buf->b_bcount and buf->b_bufsize
dillon [Thu, 25 Oct 2001 22:49:48 +0000 (22:49 +0000)]
In cluster_rbuild(), 'size' had better match buf->b_bcount and buf->b_bufsize
or the cluster will not be properly merged.  Dup the code from
cluster_wbuild() and add some printf()s to see if bad cases are present.

MFC after: 2 weeks

22 years agoMove abstract inside articleinfo and re-indent, no content changes.
bmah [Thu, 25 Oct 2001 21:06:37 +0000 (21:06 +0000)]
Move abstract inside articleinfo and re-indent, no content changes.

22 years agoNew release notes: Multiple low-level consoles, fxp(4) bundling.
bmah [Thu, 25 Oct 2001 20:47:12 +0000 (20:47 +0000)]
New release notes:  Multiple low-level consoles, fxp(4) bundling.

22 years agoo Modify format of /etc/fbtab to accept glob matching patterns for
rwatson [Thu, 25 Oct 2001 20:45:47 +0000 (20:45 +0000)]
o Modify format of /etc/fbtab to accept glob matching patterns for
  target devices, not just individual devices and directories.  This
  permits activities such as:

ttyv0 0600 /dev/dsp*

  Whereas previously that was not supported.  This change is
  backwards-compatible, except where device names included globbing
  characters, which is not the case for any devices listed in MAKEDEV.

Submitted by: Maxime Henrion <mux@qualys.com>
MFC after: 3 weeks

22 years agoUse TASK_INIT to initialize taskqueue task instead of violating the
jhb [Thu, 25 Oct 2001 19:56:02 +0000 (19:56 +0000)]
Use TASK_INIT to initialize taskqueue task instead of violating the
abstraction.

Submitted by: Andrew Reiter <arr@watson.org>

22 years agoRemove wx driver, which got recently removed.
asmodai [Thu, 25 Oct 2001 19:45:39 +0000 (19:45 +0000)]
Remove wx driver, which got recently removed.

Submitted by: revamped kernincludes.sh

22 years agoUnbreak NEWCARD by removing options NFS and replacing it with the
asmodai [Thu, 25 Oct 2001 19:41:13 +0000 (19:41 +0000)]
Unbreak NEWCARD by removing options NFS and replacing it with the
new NFSCLIENT and NFSSERVER options.

Submitted by: revamped kernincludes.sh

22 years agoAargh. I really shouldn't do late night commits. Remove a floating point
jlemon [Thu, 25 Oct 2001 19:36:44 +0000 (19:36 +0000)]
Aargh. I really shouldn't do late night commits.  Remove a floating point
multiply, and replace it with a close equivalent.  1.488 =~ 1.5

22 years agoNow that nfsm_reply() does not usually set 'error' to 0, we need
iedowse [Thu, 25 Oct 2001 19:07:56 +0000 (19:07 +0000)]
Now that nfsm_reply() does not usually set 'error' to 0, we need
to do it explicitly in nfsrv_noop so that the reply gets sent back
to the client. This fixes the generation of a selection of RPC
error replies (RPC_PROGMISMATCH, RPC_PROGUNAVAIL, RPC_PROCUNAVAIL
etc.) that are used by some clients to detect support for optional
protocols and features.

Reviewed by: peter
Reported by: Thomas Quinot <quinot@inf.enst.fr>
PR: kern/31479

22 years agoDefs for three (unused so far) bits in PCI command/status register
luigi [Thu, 25 Oct 2001 17:43:26 +0000 (17:43 +0000)]
Defs for three (unused so far) bits in PCI command/status register
were off by one bit.

22 years agoBetter code to scan the ethertable.
luigi [Thu, 25 Oct 2001 17:40:03 +0000 (17:40 +0000)]
Better code to scan the ethertable.

22 years agoFix an inverted test csae. Success of getenv() is determined by a return
jhb [Thu, 25 Oct 2001 17:22:31 +0000 (17:22 +0000)]
Fix an inverted test csae.  Success of getenv() is determined by a return
value of !NUL rather than NUL.

Submitted by: luigi
Pointy hat to: jhb