]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
23 years agoAdd groff to bootstrap-tools.
Ruslan Ermilov [Wed, 18 Apr 2001 12:20:28 +0000 (12:20 +0000)]
Add groff to bootstrap-tools.

23 years agoAdd missing -mtty-char for MANBUILDCAT.
Ruslan Ermilov [Wed, 18 Apr 2001 12:13:08 +0000 (12:13 +0000)]
Add missing -mtty-char for MANBUILDCAT.

23 years agoReclaim directory vnodes held in namecache if few free vnodes are
Seigo Tanimura [Wed, 18 Apr 2001 11:19:50 +0000 (11:19 +0000)]
Reclaim directory vnodes held in namecache if few free vnodes are
available.

Only directory vnodes holding no child directory vnodes held in
v_cache_src are recycled, so that directory vnodes near the root of
the filesystem hierarchy remain in namecache and directory vnodes are
not reclaimed in cascade.

The period of vnode reclaiming attempt and the number of vnodes
attempted to reclaim can be tuned via sysctl(2).

Suggested by: tegge
Approved by: phk

23 years agoSynch with devascii/R.proto, RFC 1489, and groff_char(7).
Ruslan Ermilov [Wed, 18 Apr 2001 09:22:51 +0000 (09:22 +0000)]
Synch with devascii/R.proto, RFC 1489, and groff_char(7).

23 years agoNFS module now requires nfs_lock.c
Alfred Perlstein [Wed, 18 Apr 2001 08:33:52 +0000 (08:33 +0000)]
NFS module now requires nfs_lock.c

23 years agomdoc(7) police: uppercase document title.
Ruslan Ermilov [Wed, 18 Apr 2001 08:25:26 +0000 (08:25 +0000)]
mdoc(7) police: uppercase document title.

23 years agoList the explanations of the command line arguments in the same order
Dima Dorfman [Wed, 18 Apr 2001 07:21:58 +0000 (07:21 +0000)]
List the explanations of the command line arguments in the same order
as they're listed in the synopsis (alphabetical).

23 years agobread() is a special case of breadn(), so don't replicate code.
Poul-Henning Kamp [Wed, 18 Apr 2001 07:16:07 +0000 (07:16 +0000)]
bread() is a special case of breadn(), so don't replicate code.

23 years agoAdd some more O2Micro devices I gleened from the web site/linux pcmcia cs.
Warner Losh [Wed, 18 Apr 2001 05:45:15 +0000 (05:45 +0000)]
Add some more O2Micro devices I gleened from the web site/linux pcmcia cs.
Add TI4451 as well.

These are untested since I don't have the hardware to test against.

Also, some O2Micro devices are #define w/o numbers as place holders so that
I can encourage people to submit them when they appear in the channels.

23 years agoFix the botched rev 1.59 where I made it such that without INVARIANTS
Alfred Perlstein [Wed, 18 Apr 2001 05:30:24 +0000 (05:30 +0000)]
Fix the botched rev 1.59 where I made it such that without INVARIANTS
the map is never locked.

Submitted by: tegge

23 years agoAdd support for the OZ6832.
Warner Losh [Wed, 18 Apr 2001 04:42:11 +0000 (04:42 +0000)]
Add support for the OZ6832.

Submitted by: Martin Nilsson <martin@svenskabutiker.se>

23 years agoA few more mountd cleanups:
Ian Dowse [Wed, 18 Apr 2001 00:28:37 +0000 (00:28 +0000)]
A few more mountd cleanups:
- Remove some horrible code that faked a "struct addrinfo" to be
  later passed to freeaddrinfo(). Instead, add a new group type
  "GT_DEFAULT" used to denote that the filesystem is exported to the
  world, and treat this case separately.
- Don't clear the AI_CANONNAME flag in a struct addrinfo returned
  by getaddrinfo. There's still a bit more struct addrinfo abuse
  left in here.
- Simplify do_mount() slightly by using an addrinfo pointer to keep
  track of the current address.

23 years agoMove setting of TI113X_PCI_CARD_CONTROL register sooner
Warner Losh [Tue, 17 Apr 2001 23:56:12 +0000 (23:56 +0000)]
Move setting of TI113X_PCI_CARD_CONTROL register sooner

23 years agoMinor comment that missed the last change
Warner Losh [Tue, 17 Apr 2001 23:50:24 +0000 (23:50 +0000)]
Minor comment that missed the last change

23 years agoWhen booting, turn on the 3E0 compatibility address for ricoh cardbus
Warner Losh [Tue, 17 Apr 2001 23:15:00 +0000 (23:15 +0000)]
When booting, turn on the 3E0 compatibility address for ricoh cardbus
parts.  This is based on the newcard code that turns it off :-).  We
can now reboot after NEWCARD or Windows and have OLDCARD work.  Add
support for the RL5C466 while I'm at it.

Treat TI1031 the same as the CLPD6832.  It doesn't work yet, but sucks
less than it did before.

Also add a few #defines for other changes in the pipe.

23 years agoVarious bugfixes and cleanups, mainly from Martin Blapp:
Ian Dowse [Tue, 17 Apr 2001 22:25:48 +0000 (22:25 +0000)]
Various bugfixes and cleanups, mainly from Martin Blapp:
- Revert del_mlist() to its pre-tirpc prototype. Unlike NetBSD's version,
  ours lets the caller generate any syslog() messages, so that it
  can include the service name in the message.
- Initialise a few local variables to clarify the logic and avoid some
  compiler warnings.
- Remove a few unused functions and local variables, and fix some
  whitespace issues.
- Reinstate the logic for avoiding duplicate host entries that got
  removed accidentally in revision 1.41 (added in r1.5). This bit
  was submitted in a slightly different form by Thomas Quinot.

Submitted by: Martin Blapp <mb@imp.ch>,
Thomas Quinot <quinot@inf.enst.fr>
PR: bin/26148

23 years agoMake this driver play ball with devfs(5).
Dima Dorfman [Tue, 17 Apr 2001 20:53:11 +0000 (20:53 +0000)]
Make this driver play ball with devfs(5).

Reviewed by: brian

23 years agoAdd a sanity check on ucred refcount.
Alfred Perlstein [Tue, 17 Apr 2001 20:50:43 +0000 (20:50 +0000)]
Add a sanity check on ucred refcount.

Submitted by: Terry Lambert <terry@lambert.org>

23 years agoImplement client side NFS locks.
Alfred Perlstein [Tue, 17 Apr 2001 20:45:23 +0000 (20:45 +0000)]
Implement client side NFS locks.

Obtained from: BSD/os
Import Ok'd by: mckusick, jkh, motd on builder.freebsd.org

23 years agoWrite a switch statement as less obscure if statements.
Poul-Henning Kamp [Tue, 17 Apr 2001 20:22:07 +0000 (20:22 +0000)]
Write a switch statement as less obscure if statements.

23 years agoUpdate for security advisories: 01:32 (ipfilter), 01:33 (glob/ftp).
Bruce A. Mah [Tue, 17 Apr 2001 19:22:07 +0000 (19:22 +0000)]
Update for security advisories:  01:32 (ipfilter), 01:33 (glob/ftp).

New release notes:  TCP ISS randomization, groff-1.17.

23 years agoFixed typos.
Jens Schweikhardt [Tue, 17 Apr 2001 19:08:41 +0000 (19:08 +0000)]
Fixed typos.
Approved by: jkh

23 years agoOptimize make invocations by using multiple targets.
David E. O'Brien [Tue, 17 Apr 2001 18:32:35 +0000 (18:32 +0000)]
Optimize make invocations by using multiple targets.

23 years agoSave are floating point state in cpu_switch() if needed instead of relying
John Baldwin [Tue, 17 Apr 2001 18:27:55 +0000 (18:27 +0000)]
Save are floating point state in cpu_switch() if needed instead of relying
completely on lazy floating point state saving.  This is needed for the
SMP case since processes can migrate to other CPUs.

Submitted by: dfr

23 years agoI think this was supposed to be __const like in NetBSD: I have no idea
Kris Kennaway [Tue, 17 Apr 2001 18:16:46 +0000 (18:16 +0000)]
I think this was supposed to be __const like in NetBSD: I have no idea
why I changed it to const.

Noticed by: David Wolfskill <david@catwhisker.org>

23 years agoNote that the previous commit also restored some historical behaviour
Kris Kennaway [Tue, 17 Apr 2001 18:14:37 +0000 (18:14 +0000)]
Note that the previous commit also restored some historical behaviour
in the TCP_COMPAT_42 case (e.g. choosing '1' as the initial sequence
number at boot-time, instead of randomizing it).  TCP_COMPAT_42 is the
repository for old security holes, too :-)

23 years agoRandomize the TCP initial sequence numbers more thoroughly.
Kris Kennaway [Tue, 17 Apr 2001 18:08:01 +0000 (18:08 +0000)]
Randomize the TCP initial sequence numbers more thoroughly.

Obtained from: OpenBSD
Reviewed by: jesper, peter, -developers

23 years agoFix an old bug related to BETTER_CLOCK. Call forward_*clock if SMP
John Baldwin [Tue, 17 Apr 2001 17:53:36 +0000 (17:53 +0000)]
Fix an old bug related to BETTER_CLOCK.  Call forward_*clock if SMP
and __i386__ are defined rather than if SMP and BETTER_CLOCK are defined.
The removal of BETTER_CLOCK would have broken this except that kern_clock.c
doesn't include <machine/smptests.h>, so it doesn't see the definition of
BETTER_CLOCK, and forward_*clock aren't called, even on 4.x.  This seems to
fix the problem where a n-way SMP system would see 100 * n clk interrupts
and 128 * n rtc interrupts.

23 years agoTypos fixed.
Jens Schweikhardt [Tue, 17 Apr 2001 17:42:46 +0000 (17:42 +0000)]
Typos fixed.
Approved by: jkh

23 years agoRegen.
Nick Hibma [Tue, 17 Apr 2001 15:53:35 +0000 (15:53 +0000)]
Regen.

23 years agoAdd the HP scanner 3400C
Nick Hibma [Tue, 17 Apr 2001 15:53:20 +0000 (15:53 +0000)]
Add the HP scanner 3400C

23 years agoRegen.
Nick Hibma [Tue, 17 Apr 2001 15:34:44 +0000 (15:34 +0000)]
Regen.

23 years agoLexar jumpSHOT Id.
Nick Hibma [Tue, 17 Apr 2001 15:34:22 +0000 (15:34 +0000)]
Lexar jumpSHOT Id.

Submitted by: gshapiro

23 years agoRemove a bogus %s from a printf.
Nick Hibma [Tue, 17 Apr 2001 15:00:03 +0000 (15:00 +0000)]
Remove a bogus %s from a printf.

Submitted by: kris

23 years agoset gd_other_cpus later, when we have a better idea what the other cpus
Andrew Gallatin [Tue, 17 Apr 2001 14:59:05 +0000 (14:59 +0000)]
set gd_other_cpus later, when we have a better idea what the other cpus
are.

submitted by: jhb

23 years agoImproved support for alpha SMP. The following commit gets dual AS2100s
Andrew Gallatin [Tue, 17 Apr 2001 14:55:09 +0000 (14:55 +0000)]
Improved support for alpha SMP.  The following commit gets dual AS2100s
and AS4100s into single user mode. This work was done jointly by jhb and
myself, and builds on dfr's earlier work.

smp_init_secondary() / smp_start_secondary()
- use the uniq val to pass the globalp (me)
- fancy footwork to take any pending machine checks (me)
- doing things the FreeBSD way and getting the per-cpu idleproc created
  correctly, and synchronizing the startup of secondaries (jhb)

mp_start()
- better recognition of available cpus (jhb)

smp_rendezvous()
- if smp hasn't started, only run the rendezvous function on the current
cpu.  Sleuthing and (prior) incorrect fix by me, correct fix by jhb

smp_handle_ipi()
- more verbose handling of console messages (jhb)
- grab sched lock around setting PS_ASTPENDING (jhb)

forward_*clock()
- commented out.  Joint decision by dfr, jhb and myself

 General synchronization improvements (more mb()s, etc) (jhb)

 Printf cleanups (joint)

 Whitespace cleanups (jhb)

23 years agoUpdate upgrade instructions.
Ruslan Ermilov [Tue, 17 Apr 2001 14:31:56 +0000 (14:31 +0000)]
Update upgrade instructions.

23 years agoRemove share/tmac/locale/ and share/tmac/mdoc/locale/.
Ruslan Ermilov [Tue, 17 Apr 2001 14:22:50 +0000 (14:22 +0000)]
Remove share/tmac/locale/ and share/tmac/mdoc/locale/.

23 years agoChanges to support SMP:
Andrew Gallatin [Tue, 17 Apr 2001 14:20:33 +0000 (14:20 +0000)]
Changes to support SMP:

- don't do the stack overflow sanity check on MP systems -- p->p_addr
will be malloc'ed memory (not K0SEG) and the check will fail.

- don't ignore clock interrupts on secondaries.  Alphas apparently
roundrobin clock interrupts to all cpus, so we're going to take clock
interrupts on all CPUS and not forward them.

23 years agochanges to smp_init_secondary_glue():
Andrew Gallatin [Tue, 17 Apr 2001 14:15:12 +0000 (14:15 +0000)]
changes to smp_init_secondary_glue():

- use the unique value to save the per-cpu globalp struct like the
  comment says

- don't lower the ipl to ALPHA_PSL_IPL_HIGH: we may have a pending machine
  check to take and we're not prepared for that yet, as we haven't setup
  our interrupt entry points. (this may only happen on sable/lynx)

- indicate the fact that the working version of smp_init_secondary() doesn't
  return (this is tied up in other changes and hasn't yet been committed).

23 years agoFreeBSD provides its own version of this OS-specific file.
Ruslan Ermilov [Tue, 17 Apr 2001 13:32:10 +0000 (13:32 +0000)]
FreeBSD provides its own version of this OS-specific file.

23 years agoAdd Ukrainian language support for FreeBSD console.
Maxim Sobolev [Tue, 17 Apr 2001 12:51:27 +0000 (12:51 +0000)]
Add Ukrainian language support for FreeBSD console.

Submitted by: Olexander Kunytsa <kunia@istc.kiev.ua>
Reviewed by: ache, phantom, ru, sobomax

23 years agoSynch template with one documented in the new mdoc(7) manpage.
Ruslan Ermilov [Tue, 17 Apr 2001 12:51:21 +0000 (12:51 +0000)]
Synch template with one documented in the new mdoc(7) manpage.

23 years agomdoc.samples(7) is superseded by mdoc(7).
Ruslan Ermilov [Tue, 17 Apr 2001 12:49:42 +0000 (12:49 +0000)]
mdoc.samples(7) is superseded by mdoc(7).

23 years agoUse Groff versions of man(7) and mdoc(7) manpages.
Ruslan Ermilov [Tue, 17 Apr 2001 12:48:14 +0000 (12:48 +0000)]
Use Groff versions of man(7) and mdoc(7) manpages.
mdoc.samples(7) is superseded by mdoc(7).

23 years agoSynch with new contrib/groff/ directory layout.
Ruslan Ermilov [Tue, 17 Apr 2001 12:40:46 +0000 (12:40 +0000)]
Synch with new contrib/groff/ directory layout.

23 years agoUpgrade to Groff 1.17.
Ruslan Ermilov [Tue, 17 Apr 2001 12:37:18 +0000 (12:37 +0000)]
Upgrade to Groff 1.17.

23 years agoRemoved yacc(1) output files.
Ruslan Ermilov [Tue, 17 Apr 2001 12:29:43 +0000 (12:29 +0000)]
Removed yacc(1) output files.

23 years agoMerge Groff 1.17 changes.
Ruslan Ermilov [Tue, 17 Apr 2001 12:28:31 +0000 (12:28 +0000)]
Merge Groff 1.17 changes.

23 years agoUse stock (FSF) version of this file.
Ruslan Ermilov [Tue, 17 Apr 2001 12:28:00 +0000 (12:28 +0000)]
Use stock (FSF) version of this file.

23 years agoThis commit was generated by cvs2svn to compensate for changes in r75587,
Ruslan Ermilov [Tue, 17 Apr 2001 12:23:50 +0000 (12:23 +0000)]
This commit was generated by cvs2svn to compensate for changes in r75587,
which included commits to RCS files with non-trunk default branches.

23 years agoRemoved files not present in v1.17 import.
Ruslan Ermilov [Tue, 17 Apr 2001 12:23:50 +0000 (12:23 +0000)]
Removed files not present in v1.17 import.

23 years agoThis commit was generated by cvs2svn to compensate for changes in r75584,
Ruslan Ermilov [Tue, 17 Apr 2001 12:12:05 +0000 (12:12 +0000)]
This commit was generated by cvs2svn to compensate for changes in r75584,
which included commits to RCS files with non-trunk default branches.

23 years agoVirgin import of FSF groff v1.17
Ruslan Ermilov [Tue, 17 Apr 2001 12:12:05 +0000 (12:12 +0000)]
Virgin import of FSF groff v1.17

23 years agoUpdate/improve Ukrainian translation.
Maxim Sobolev [Tue, 17 Apr 2001 11:00:58 +0000 (11:00 +0000)]
Update/improve Ukrainian translation.

Submitted by: Olexander Kunytsa <kunia@istc.kiev.ua> (author of translation)

23 years agoDont (ab)use drv2 to know if (si_flags & SI_NAMED) (pointed out by dd)
Brian Somers [Tue, 17 Apr 2001 09:59:34 +0000 (09:59 +0000)]
Dont (ab)use drv2 to know if (si_flags & SI_NAMED) (pointed out by dd)
Call cdevsw_remove when we unload.

23 years agoBy popular demand, have adduser preserve comments at the top of the
Doug Barton [Tue, 17 Apr 2001 09:42:07 +0000 (09:42 +0000)]
By popular demand, have adduser preserve comments at the top of the
group file. Because of the way the group sorting works while printing
out the new file it's not possible at this time to restore comments
in other locations, but at least they won't just disappear altogether.

23 years agoThis patch removes the VOP_BWRITE() vector.
Poul-Henning Kamp [Tue, 17 Apr 2001 08:56:39 +0000 (08:56 +0000)]
This patch removes the VOP_BWRITE() vector.

VOP_BWRITE() was a hack which made it possible for NFS client
side to use struct buf with non-bio backing.

This patch takes a more general approach and adds a bp->b_op
vector where more methods can be added.

The success of this patch depends on bp->b_op being initialized
all relevant places for some value of "relevant" which is not
easy to determine.  For now the buffers have grown a b_magic
element which will make such issues a tiny bit easier to debug.

23 years agoCorrect a typo; prefered -> preferred.
Kris Kennaway [Tue, 17 Apr 2001 08:01:54 +0000 (08:01 +0000)]
Correct a typo; prefered -> preferred.

23 years agoAdd fmtcheck(), a function for checking consistency of format string
Kris Kennaway [Tue, 17 Apr 2001 07:59:52 +0000 (07:59 +0000)]
Add fmtcheck(), a function for checking consistency of format string
arguments where the format string is obtained from user data, or
otherwise difficult to verify statically.

Example usage:

printf(fmtcheck(user_format, standard_format), arg1, arg2);

checks the format string user_format for consistency (same number/order/
type of format operators) with standard_format.  If they differ,
standard_format is used instead to avoid potential crashes or security
violations.

Obtained from:  NetBSD
Reviewed by:    -arch

23 years ago-Wnon-const-format sweep: make format strings const char *'s, add
Kris Kennaway [Tue, 17 Apr 2001 07:46:38 +0000 (07:46 +0000)]
-Wnon-const-format sweep: make format strings const char *'s, add
__printflike()/__printf0like() to function prototypes, as appropriate.

Reviewed by: bde, -audit

23 years ago'%' is not a valid format string :-)
Kris Kennaway [Tue, 17 Apr 2001 07:35:41 +0000 (07:35 +0000)]
'%' is not a valid format string :-)

23 years agoRemove spurious extra format string argument.
Kris Kennaway [Tue, 17 Apr 2001 07:32:37 +0000 (07:32 +0000)]
Remove spurious extra format string argument.

23 years agoAdd a missing argument to an error message format string.
Kris Kennaway [Tue, 17 Apr 2001 07:21:48 +0000 (07:21 +0000)]
Add a missing argument to an error message format string.

23 years agoAdd debugging option to always read/write cylinder groups as full
Kirk McKusick [Tue, 17 Apr 2001 05:37:51 +0000 (05:37 +0000)]
Add debugging option to always read/write cylinder groups as full
sized blocks. To enable this option, use: `sysctl -w debug.bigcgs=1'.
Add debugging option to disable background writes of cylinder
groups. To enable this option, use: `sysctl -w debug.dobkgrdwrite=0'.
These debugging options should be tried on systems that are panicing
with corrupted cylinder group maps to see if it makes the problem
go away. The set of panics in question are:

ffs_clusteralloc: map mismatch
ffs_nodealloccg: map corrupted
ffs_nodealloccg: block not in map
ffs_alloccg: map corrupted
ffs_alloccg: block not in map
ffs_alloccgblk: cyl groups corrupted
ffs_alloccgblk: can't find blk in cyl
ffs_checkblk: partially free fragment

The following panics are less likely to be related to this problem,
but might be helped by these debugging options:

ffs_valloc: dup alloc
ffs_blkfree: freeing free block
ffs_blkfree: freeing free frag
ffs_vfree: freeing free inode

If you try these options, please report whether they helped reduce your
bitmap corruption panics to Kirk McKusick at <mckusick@mckusick.com>
and to Matt Dillon <dillon@earth.backplane.com>.

23 years agoBackground fsck sysctl operations must use vn_start_write and
Kirk McKusick [Tue, 17 Apr 2001 05:06:37 +0000 (05:06 +0000)]
Background fsck sysctl operations must use vn_start_write and
vn_finished_write so that they do not attempt to modify a
suspended filesystem.

23 years agoIn my first reading of POSIX.1e, I misinterpreted handling of the
Robert Watson [Tue, 17 Apr 2001 04:33:34 +0000 (04:33 +0000)]
In my first reading of POSIX.1e, I misinterpreted handling of the
ACL_USER_OBJ and ACL_GROUP_OBJ fields, believing that modification of the
access ACL could be used by privileged processes to change file/directory
ownership.  In fact, this is incorrect; ACL_*_OBJ (+ ACL_MASK and
ACL_OTHER) should have undefined ae_id fields; this commit attempts
to correct that misunderstanding.

o Modify arguments to vaccess_acl_posix1e() to accept the uid and gid
  associated with the vnode, as those can no longer be extracted from
  the ACL passed as an argument.  Perform all comparisons against
  the passed arguments.  This actually has the effect of simplifying
  a number of components of this call, as well as reducing the indent
  level, but now seperates handling of ACL_GROUP_OBJ from ACL_GROUP.

o Modify acl_posix1e_check() to return EINVAL if the ae_id field of
  any of the ACL_{USER_OBJ,GROUP_OBJ,MASK,OTHER} entries is a value
  other than ACL_UNDEFINED_ID.  As a temporary work-around to allow
  clean upgrades, set the ae_id field to ACL_UNDEFINED_ID before
  each check so that this cannot cause a failure in the short term
  (this work-around will be removed when the userland libraries and
  utilities are updated to take this change into account).

o Modify ufs_sync_acl_from_inode() so that it forces
  ACL_{USER_OBJ,GROUP_OBJ,MASK,OTHER} ae_id fields to ACL_UNDEFINED_ID
  when synchronizing the ACL from the inode.

o Modify ufs_sync_inode_from_acl to not propagate uid and gid
  information to the inode from the ACL during ACL update.  Also
  modify the masking of permission bits that may be set from
  ALLPERMS to (S_IRWXU|S_IRWXG|S_IRWXO), as ACLs currently do not
  carry none-ACCESSPERMS (S_ISUID, S_ISGID, S_ISTXT).

o Modify ufs_getacl() so that when it emulates an access ACL from
  the inode, it initializes the ae_id fields to ACL_UNDEFINED_ID.

o Clean up ufs_setacl() substantially since it is no longer possible
  to perform chown/chgrp operations using vop_setacl(), so all the
  access control for that can be eliminated.

o Modify ufs_access() so that it passes owner uid and gid information
  into vaccess_acl_posix1e().

Pointed out by: jedger
Obtained from: TrustedBSD Project

23 years agoBlow away the panic mutex in favor of using a single atomic_cmpset() on a
John Baldwin [Tue, 17 Apr 2001 04:18:08 +0000 (04:18 +0000)]
Blow away the panic mutex in favor of using a single atomic_cmpset() on a
panic_cpu shared variable.  I used a simple atomic operation here instead
of a spin lock as it seemed to be excessive overhead.  Also, this can avoid
recursive panics if, for example, witness is broken.

23 years agoCheck to see if enroll() returns NULL in the witness initialization. This
John Baldwin [Tue, 17 Apr 2001 03:35:38 +0000 (03:35 +0000)]
Check to see if enroll() returns NULL in the witness initialization.  This
can happen if witness runs out of resources during initialization or if
witness_skipspin is enabled.

Sleuthing by: Peter Jeremy <peter.jeremy@alcatel.com.au>

23 years agoExit and re-enter the critical section while spinning for a spinlock so
John Baldwin [Tue, 17 Apr 2001 03:34:52 +0000 (03:34 +0000)]
Exit and re-enter the critical section while spinning for a spinlock so
that interrupts can come in while we are waiting for a lock.

23 years agoPrevious clobbered a work-in-progress. Here is the merged result:
Peter Wemm [Tue, 17 Apr 2001 03:03:45 +0000 (03:03 +0000)]
Previous clobbered a work-in-progress.  Here is the merged result:

Limit the "pathname" glob to one item, as that is what all users of it
are expecting, except for LIST.

Always glob, instead of when the first character is a ~.  For example,
if you had directories ~/x1, and ~/x2, then "cwd x[1]" would fail, but
"cwd ~/x[1]" would work since it was globbed due to the ~ character.
Also, "cwd ~/x[12]" used to arbitarily work as it used the first
expansion (ie: x1) without an error.  Make it return '550 ambiguous'
instead of '550 not found' so that the user can see the difference.

For LIST, just use the user supplied string as the popen does the glob.

Problem noticed by:  Ajay Mittal <amittal@iprg.nokia.com>

23 years ago- Add appropriate #ifndef/#define/#endif to protect against multiple
John Baldwin [Tue, 17 Apr 2001 02:51:28 +0000 (02:51 +0000)]
- Add appropriate #ifndef/#define/#endif to protect against multiple
  inclusions.
- Blow away all evidence of a static curpcb as curpcb is a per-CPU variable
  and this definition is now bogus.

23 years ago- Fix memory barriers in atomic operations so that the barriers are always
John Baldwin [Tue, 17 Apr 2001 02:50:05 +0000 (02:50 +0000)]
- Fix memory barriers in atomic operations so that the barriers are always
  "inside" of locked regions.  That is, an acquire atomic operation will
  always enforce a memory barrier after the atomic operation and a release
  operation will always enforce a memory barrier before the atomic
  operation.
- Explicitly use 'mb' instead of 'wmb' in release atomic operations.  The
  'wmb' memory barrier is not strong enough to guarantee coherence with
  other processors.  This is effectively a nop since alpha_wmb() actually
  performs a 'mb' and not a 'wmb', but I wanted the code to be more
  correct since at some point in the future alpha_wmb()'s implementation
  may switch to being a real 'wmb'.

23 years agoAdd a cpu_throw() function that secondary CPU's can use for their first
John Baldwin [Tue, 17 Apr 2001 02:46:13 +0000 (02:46 +0000)]
Add a cpu_throw() function that secondary CPU's can use for their first
context switch.

23 years agoIn exception_return(), test for usermode before testing the IPL to see if
John Baldwin [Tue, 17 Apr 2001 02:44:35 +0000 (02:44 +0000)]
In exception_return(), test for usermode before testing the IPL to see if
we should call ast().  This allows us to branch to a separate Lkernelret
label so we can fixup the saved t7 register in the trapframe.  Otherwise
we can run into a problem on SMP systems where a process is interrupted by
a trap or interrupt on one CPU, migrates to another CPU, and then returns
with the t7 in the stack clobbering the CPU's t7.  As a result, two CPU's
would both point to the same per-CPU data and things would go downhill from
there.

Sleuthing help by: gallatin

23 years ago- Stop other CPU's in the SMP case when we enter ddb.
John Baldwin [Tue, 17 Apr 2001 02:41:41 +0000 (02:41 +0000)]
- Stop other CPU's in the SMP case when we enter ddb.
- Add a new ddb command: 'show pcpu' similar to the i386 command added
  recently.  By default it displays the current CPU's info, but an optional
  argument can specify the logical ID of a specific CPU to examine.

23 years agoLimit number of paths returned via glob() for authorized users
Chris D. Faulhaber [Tue, 17 Apr 2001 02:33:20 +0000 (02:33 +0000)]
Limit number of paths returned via glob() for authorized users
using tilde expansion.

23 years agos/Over 4500 ports/Over 5000 ports/
Bruce A. Mah [Tue, 17 Apr 2001 02:22:16 +0000 (02:22 +0000)]
s/Over 4500 ports/Over 5000 ports/

23 years agoMinor background cleanups:
Kirk McKusick [Mon, 16 Apr 2001 22:22:21 +0000 (22:22 +0000)]
Minor background cleanups:
1) Set the FS_NEEDSFSCK flag when unexpected problems are encountered.
2) Clear the FS_NEEDSFSCK flag after a successful foreground cleanup.
3) Refuse to run in background when the FS_NEEDSFSCK flag is set.
4) Avoid taking and removing a snapshot when the filesystem is already clean.
5) Properly implement the force cleaning (-f) flag when in preen mode.

Note that you need to have revision 1.21 (date: 2001/04/14 05:26:28) of
fs.h installed in <ufs/ffs/fs.h> defining FS_NEEDSFSCK for this to compile.

23 years agoSupport the empty "PASS\r\n" command.
Brian Feldman [Mon, 16 Apr 2001 22:20:26 +0000 (22:20 +0000)]
Support the empty "PASS\r\n" command.

23 years agoAdd support the the Intel ICH2 mobile
Søren Schmidt [Mon, 16 Apr 2001 21:22:34 +0000 (21:22 +0000)]
Add support the the Intel ICH2 mobile

Tidy chip name printing a bit.

23 years agoFix typo s/thissystem/this system/
Jesper Skriver [Mon, 16 Apr 2001 21:14:13 +0000 (21:14 +0000)]
Fix typo s/thissystem/this system/

Submitted by: Stephen <sdk@shell.yuck.net>

23 years agoClean up temporary file(s) and directory when m4 exits without falling
Gregory Neil Shapiro [Mon, 16 Apr 2001 18:36:35 +0000 (18:36 +0000)]
Clean up temporary file(s) and directory when m4 exits without falling
through main() (e.g., signals or calls to errx()).

PR: conf/25715

23 years agoAdd missing sound module.
Mark Murray [Mon, 16 Apr 2001 17:04:41 +0000 (17:04 +0000)]
Add missing sound module.

OK'ed by: cg

23 years agoFix an off-by-2 error in periphdriver_register(). The read side of the
Kenneth D. Merry [Mon, 16 Apr 2001 15:53:54 +0000 (15:53 +0000)]
Fix an off-by-2 error in periphdriver_register().  The read side of the
bcopy would go off the end of the array by two elements, which sometimes
causes a panic if it happens to cross into a page that isn't mapped.

Submitted by: gibbs
Reviewed by: peter

23 years agomdoc(7) police: Prepare for mdocNG.
Ruslan Ermilov [Mon, 16 Apr 2001 15:18:16 +0000 (15:18 +0000)]
mdoc(7) police: Prepare for mdocNG.

The .Cd macro does not cause a line break until in the SYNOPSIS section.

23 years agoGrrr, get the right change from the the machine I tested this on.
David E. O'Brien [Mon, 16 Apr 2001 15:17:27 +0000 (15:17 +0000)]
Grrr, get the right change from the the machine I tested this on.

23 years agomdoc(7) police: cosmetics.
Ruslan Ermilov [Mon, 16 Apr 2001 15:14:07 +0000 (15:14 +0000)]
mdoc(7) police: cosmetics.

23 years agomdoc(7) police: prepare for mdocNG.
Ruslan Ermilov [Mon, 16 Apr 2001 15:12:58 +0000 (15:12 +0000)]
mdoc(7) police: prepare for mdocNG.

23 years agomdoc(7) police: add missing .El call.
Ruslan Ermilov [Mon, 16 Apr 2001 15:06:57 +0000 (15:06 +0000)]
mdoc(7) police: add missing .El call.

23 years agomdoc(7) police: fix markup.
Ruslan Ermilov [Mon, 16 Apr 2001 15:06:19 +0000 (15:06 +0000)]
mdoc(7) police: fix markup.

23 years agoDocument that SITE extensions are disabled for anonymous logins.
Ruslan Ermilov [Mon, 16 Apr 2001 14:51:11 +0000 (14:51 +0000)]
Document that SITE extensions are disabled for anonymous logins.

Obtained from: logdaemon package by Wietse Venema

23 years agoMove the isa parts to a separate file.
John Hay [Mon, 16 Apr 2001 13:20:39 +0000 (13:20 +0000)]
Move the isa parts to a separate file.

23 years agoUpdate to the 2001-04-02 version of the nanokernel code from Dave Mills.
John Hay [Mon, 16 Apr 2001 13:05:05 +0000 (13:05 +0000)]
Update to the 2001-04-02 version of the nanokernel code from Dave Mills.

23 years agoMerged from sys/i386/isa/npx.c revision 1.95.
KATO Takenori [Mon, 16 Apr 2001 09:15:40 +0000 (09:15 +0000)]
Merged from sys/i386/isa/npx.c revision 1.95.

23 years agoMerged from sys/conf/options.i386 revision 1.148.
KATO Takenori [Mon, 16 Apr 2001 09:13:51 +0000 (09:13 +0000)]
Merged from sys/conf/options.i386 revision 1.148.

23 years agoNew script to help creation of shared readonly diskless partition.
Luigi Rizzo [Mon, 16 Apr 2001 06:37:03 +0000 (06:37 +0000)]
New script to help creation of shared readonly diskless partition.
It also has some instructions on how to setup the client and
the server. I have been using this code for over 2 years
on RELENG_3 and later RELENG_4. Have not tried on CURRENT, but
in case there are any issues these are in /etc/rc and
/etc/rc.diskless{12}

23 years agoAdd the "SITE MD5 filename" facility.
Poul-Henning Kamp [Sun, 15 Apr 2001 20:59:29 +0000 (20:59 +0000)]
Add the "SITE MD5 filename" facility.

This allows you to determine if the file on the other side is the same
as the one you have without transferring the entire file to compare.

Needless to say, if the server end lies to you this check doesn't work,
but on the other hand, if it lies to you about the files checksum,
what can you trust from it ?

23 years agoAdd a more useful solution to the problem of password files with more than
Doug Barton [Sun, 15 Apr 2001 20:43:49 +0000 (20:43 +0000)]
Add a more useful solution to the problem of password files with more than
one user who differs only by case. The other perl tools assume (or enforce)
the all lowercase requirement, therefore making the search through
master.passwd case insensitive seemed a reasonable optimization, IMO.

I understand, although I do not sympathize with, the argument that someone
might want to do this on purpose, and might subsequently want to use the
wrong tool for the job. So, this fix should hopefully satisfy both camps.

23 years agoForced commit to point out that PR 25961 was actually originated by
Doug Barton [Sun, 15 Apr 2001 20:38:01 +0000 (20:38 +0000)]
Forced commit to point out that PR 25961 was actually originated by
Dan Langille <dan@freebsddiary.org>.