]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoAdd a f_vnode field to struct file.
Poul-Henning Kamp [Sun, 22 Jun 2003 08:41:43 +0000 (08:41 +0000)]
Add a f_vnode field to struct file.

Several of the subtypes have an associated vnode which is used for
stuff like the f*() functions.

By giving the vnode a speparate field, a number of checks for the specific
subtype can be replaced simply with a check for f_vnode != NULL, and
we can later free f_data up to subtype specific use.

At this point in time, f_data still points to the vnode, so any code I
might have overlooked will still work.

21 years agoHook locale(1) to build
Alexey Zelkin [Sun, 22 Jun 2003 08:41:03 +0000 (08:41 +0000)]
Hook locale(1) to build

21 years agodocument 'list' keyword & fix few typos
Alexey Zelkin [Sun, 22 Jun 2003 08:39:29 +0000 (08:39 +0000)]
document 'list' keyword & fix few typos

21 years agoAdd FreeBSD specific keyword 'list' implementation. 'locale -k list' can
Alexey Zelkin [Sun, 22 Jun 2003 08:34:27 +0000 (08:34 +0000)]
Add FreeBSD specific keyword 'list' implementation.  'locale -k list' can
be used to retrieve list of all available keywords now.

21 years agoAdd WARNS
Alexey Zelkin [Sun, 22 Jun 2003 08:24:53 +0000 (08:24 +0000)]
Add WARNS

21 years agosupport saving both user/group and permissions on symlinks (from PR)
John-Mark Gurney [Sun, 22 Jun 2003 07:02:17 +0000 (07:02 +0000)]
support saving both user/group and permissions on symlinks (from PR)

also fix a slight bogon that assumed an fd of 0 was not valid.  Changed
it to be -1.

PR: bin/25017
Submitted by: Martin Kammerhofer

21 years agoAdd ID for VT8233A.
Matthew N. Dodd [Sun, 22 Jun 2003 06:50:02 +0000 (06:50 +0000)]
Add ID for VT8233A.

PR:  i386/38299
Submitted by:  Rob Schulhof <rrs@there.net>

21 years agoRe-enabled PCI irq routing on pc98.
Yoshihiro Takahashi [Sun, 22 Jun 2003 06:09:14 +0000 (06:09 +0000)]
Re-enabled PCI irq routing on pc98.

21 years ago- Set close on exec flag for device file descriptors.
Matthew N. Dodd [Sun, 22 Jun 2003 05:57:34 +0000 (05:57 +0000)]
- Set close on exec flag for device file descriptors.
- Reset signal handlers in event_cmd_exec_act().

PR:  i386/35182
Submitted by:  Daniel O'Connor <darius@dons.net.au>

21 years agoAs vm_fault() descends the chain of backing objects, set paging-in-
Alan Cox [Sun, 22 Jun 2003 05:36:53 +0000 (05:36 +0000)]
As vm_fault() descends the chain of backing objects, set paging-in-
progress on the next object before clearing it on the current object.

21 years ago- Don't ignore SIGTERM.
Matthew N. Dodd [Sun, 22 Jun 2003 05:34:45 +0000 (05:34 +0000)]
- Don't ignore SIGTERM.
- Add a command line switch to trigger POWERSTATECHANGE actions on
  un-reported power state changes.

PR:  i386/32251
Submitted by:  Walter C. Pelissero <walter@pelissero.org>

21 years agoImplement a loader tunable/sysctl to allow the user to request that
Matthew N. Dodd [Sun, 22 Jun 2003 05:08:10 +0000 (05:08 +0000)]
Implement a loader tunable/sysctl to allow the user to request that
the APM driver byte-swap battery time values.  (For broken laptops.)

PR: i386/42439
Submitted by: Bruce M Simpson <bms@spc.org>

21 years agoWhen DDB is active, always send printf() output directly to the
Ian Dowse [Sun, 22 Jun 2003 03:20:24 +0000 (03:20 +0000)]
When DDB is active, always send printf() output directly to the
console, even if there is a TIOCCONS console tty. We were already
doing this after a panic, but it's also useful when entering DDB
for some other reason too.

21 years agoUse a new message buffer `consmsgbuf' to forward messages to a
Ian Dowse [Sun, 22 Jun 2003 02:54:33 +0000 (02:54 +0000)]
Use a new message buffer `consmsgbuf' to forward messages to a
TIOCCONS console (e.g. xconsole) via a timeout routine instead of
calling into the tty code directly from printf(). This fixes a
number of cases where calling printf() at the wrong time (such as
with locks held) would cause a panic if xconsole is running.

The TIOCCONS message buffer is 8k in size by default, but this can
be changed with the kern.consmsgbuf_size sysctl. By default, messages
are checked for 5 times per second. The timer runs and the buffer
memory remains allocated only at times when a TIOCCONS console is
active.

Discussed on: freebsd-arch

21 years agoComplete the vm object locking in vm_object_backing_scan(); specifically,
Alan Cox [Sun, 22 Jun 2003 02:35:06 +0000 (02:35 +0000)]
Complete the vm object locking in vm_object_backing_scan(); specifically,
deal with the case where we need to sleep on a busy page with two vm object
locks held.

21 years agouse a REG macro that was already defined.
John-Mark Gurney [Sun, 22 Jun 2003 02:26:17 +0000 (02:26 +0000)]
use a REG macro that was already defined.

Reorder how the pci probing in handled.  before adding devices, check to
see if the slot is a multi-function device to see if we should probe all
the functions.

Original idea by: imp

21 years agoReplace the code for reading and writing the kernel message buffer
Ian Dowse [Sun, 22 Jun 2003 02:18:31 +0000 (02:18 +0000)]
Replace the code for reading and writing the kernel message buffer
with a new implementation that has a mostly reentrant "addchar"
routine, supports multiple message buffers in the kernel, and hides
the implementation details from callers.

The new code uses a kind of sequence number to represend the current
read and write positions in the buffer. This approach (suggested
mainly by bde) permits the read and write pointers to be maintained
separately, which reduces the number of atomic operations that are
required. The "mostly reentrant" above refers to the way that while
it is now always safe to have any number of concurrent writers,
readers could see the message buffer after a writer has advanced
the pointers but before it has witten the new character.

Discussed on: freebsd-arch

21 years agoadd support for peeking at pci busses on UltraSparc systems. This prevents
John-Mark Gurney [Sun, 22 Jun 2003 01:26:08 +0000 (01:26 +0000)]
add support for peeking at pci busses on UltraSparc systems.  This prevents
data access errors when trying to read/write to non-existant PCI devices.

fix the psycho bridge to use peek for probing devices.  This no longer
fakes it if the OFW node doesn't exist (and the reg == 0).

Reviewed by: jake, tmm

21 years agoMake some style and white-space changes to the copy-on-write path through
Alan Cox [Sun, 22 Jun 2003 00:00:11 +0000 (00:00 +0000)]
Make some style and white-space changes to the copy-on-write path through
vm_fault(); remove a pointless assignment statement from that path.

21 years agoIncrease the size of the NFS server hash table to improve performance
Kirk McKusick [Sat, 21 Jun 2003 21:01:44 +0000 (21:01 +0000)]
Increase the size of the NFS server hash table to improve performance
when serving up more than about 32 active files. For details see
section 6.3 (pg 111) of Daniel Ellard and Margo Seltzer, ``NFS
Tricks and Benchmarking Traps'' in the Proceedings of the Usenix
2003 Freenix Track, June 9-14, 2003 pg 101-114.

Obtained from: Daniel Ellard <ellard@eecs.harvard.edu>
Sponsored by:   DARPA & NAI Labs.

21 years agoAdd clarifications about the information that ntp_gettime returns.
Warner Losh [Sat, 21 Jun 2003 20:27:54 +0000 (20:27 +0000)]
Add clarifications about the information that ntp_gettime returns.
TAI is a timescale, just like UTC.  The tai field returns the offset
between the two, and isn't really used for precision time keeping.
Explain in brief what a positive and a negative leap seconds are.  Add
some URLs to very useful web pages about time and time keeping for
more information on using this API.

Reviewed by: phk

21 years agoDon't declare unneeded extern variables,
Yaroslav Tykhiy [Sat, 21 Jun 2003 10:45:38 +0000 (10:45 +0000)]
Don't declare unneeded extern variables,
leave alone specifying a wrong type for one of them.

21 years agoThough manpage says that 0.0.0.0 can be used as HISADDR for gw
Hajimu UMEMOTO [Sat, 21 Jun 2003 10:14:52 +0000 (10:14 +0000)]
Though manpage says that 0.0.0.0 can be used as HISADDR for gw
in Framed-Route, it didn't work.  Since ncprange_aton() treats
0.0.0.0 and :: as prefixlen=0, we need to care the case.

MFC after: 1 week

21 years agoUse a do {...} while (0); and a couple of breaks to reduce the level
Poul-Henning Kamp [Sat, 21 Jun 2003 08:27:06 +0000 (08:27 +0000)]
Use a do {...} while (0); and a couple of breaks to reduce the level
of indentation a bit.

21 years agoUserland spinlocks bad. Sleep locks good.
David Schultz [Sat, 21 Jun 2003 08:20:14 +0000 (08:20 +0000)]
Userland spinlocks bad.  Sleep locks good.
Use the latter for gdtoa.

Requested by: deischen (far too long ago)

21 years agottyname(3) must return absolute pathnames.
Poul-Henning Kamp [Sat, 21 Jun 2003 08:16:12 +0000 (08:16 +0000)]
ttyname(3) must return absolute pathnames.

Noticed by: kris

21 years agoLock one of the vm objects involved in an optimized copy-on-write fault.
Alan Cox [Sat, 21 Jun 2003 06:31:42 +0000 (06:31 +0000)]
Lock one of the vm objects involved in an optimized copy-on-write fault.

21 years agoWe don't need two $FreeBSD$'s. Remove the older one.
Garance A Drosehn [Sat, 21 Jun 2003 05:59:59 +0000 (05:59 +0000)]
We don't need two $FreeBSD$'s.  Remove the older one.

21 years ago - Increase the scope of the vm object lock in vm_object_collapse().
Alan Cox [Sat, 21 Jun 2003 04:14:48 +0000 (04:14 +0000)]
 - Increase the scope of the vm object lock in vm_object_collapse().
 - Assert that the vm object and its backing vm object are both locked in
   vm_object_qcollapse().

21 years ago - lticks was erroneously being updated in sched_pctcpu(). This was causing
Jeff Roberson [Sat, 21 Jun 2003 02:31:49 +0000 (02:31 +0000)]
 - lticks was erroneously being updated in sched_pctcpu().  This was causing
   us to skip the pctcpu_update() call which lead to inaccurate cpu usage
   statistics for processes that didn't run often.

21 years ago - Don't allow nice to have such a large effect on priority. This was
Jeff Roberson [Sat, 21 Jun 2003 02:22:47 +0000 (02:22 +0000)]
 - Don't allow nice to have such a large effect on priority.  This was
   causing poor interactive performance while unnice processes were running.
   The new scheme still allows nice to have an effect on priority but it is
   not as dramatic as the effect of the interactivity score.

21 years agoNew release note: texinfo 4.6.
Bruce A. Mah [Sat, 21 Jun 2003 00:04:11 +0000 (00:04 +0000)]
New release note:  texinfo 4.6.

21 years agomake pciconf understand it's own output as stated in the manpage.
John-Mark Gurney [Fri, 20 Jun 2003 23:59:25 +0000 (23:59 +0000)]
make pciconf understand it's own output as stated in the manpage.

pciconf -r none8@pci1:12:2: 0x0
now works.

PR: bin/10312
Submitted by: Castor Fu

21 years agofix grammar in comment
John-Mark Gurney [Fri, 20 Jun 2003 23:29:04 +0000 (23:29 +0000)]
fix grammar in comment

21 years agoNew release notes: PAE panic fixed, bge(4) at 10 Mbps fixed,
Bruce A. Mah [Fri, 20 Jun 2003 23:13:23 +0000 (23:13 +0000)]
New release notes:  PAE panic fixed, bge(4) at 10 Mbps fixed,
GEOM_FOX, dev_db gone, local(1) reimplementation, ps(1) -H.

21 years agoRemove _PATH_DEVDB
Poul-Henning Kamp [Fri, 20 Jun 2003 22:50:33 +0000 (22:50 +0000)]
Remove _PATH_DEVDB

21 years agoUse devname(3) to do the task.
Poul-Henning Kamp [Fri, 20 Jun 2003 22:45:53 +0000 (22:45 +0000)]
Use devname(3) to do the task.

Reviewed by: imp

21 years agoExplain the relationship to POSIX.
Garrett Wollman [Fri, 20 Jun 2003 22:41:00 +0000 (22:41 +0000)]
Explain the relationship to POSIX.

21 years agoFix a divide-by-zero on kern.log_wakeups_per_second tunable.
Bosko Milekic [Fri, 20 Jun 2003 22:18:38 +0000 (22:18 +0000)]
Fix a divide-by-zero on kern.log_wakeups_per_second tunable.

Submitted by: Christian S.J. Peron <maneo@bsdpro.com>
PR: kern/53557

21 years agontp_adjtime returns the current state of the clock (TIME_*) on success
Warner Losh [Fri, 20 Jun 2003 21:14:59 +0000 (21:14 +0000)]
ntp_adjtime returns the current state of the clock (TIME_*) on success
or -1 on failure.  The manual used to say it returned 0 or -1.  Both
examination of the kernel sources, and ntpd show that this is the case.

MFC After: 3 days

21 years agoNew errata: PAE kernel panic.
Bruce A. Mah [Fri, 20 Jun 2003 21:07:05 +0000 (21:07 +0000)]
New errata:  PAE kernel panic.

Mostly submitted by scottl.  Any errors are mine.

21 years agoMake swap_pager_haspages() static; remove unused function prototypes.
Alan Cox [Fri, 20 Jun 2003 20:20:06 +0000 (20:20 +0000)]
Make swap_pager_haspages() static; remove unused function prototypes.

21 years agos/warn/info/
Mike Makonnen [Fri, 20 Jun 2003 19:36:56 +0000 (19:36 +0000)]
s/warn/info/

21 years agoAdd support for the special shell nologin.
Mike Makonnen [Fri, 20 Jun 2003 19:23:55 +0000 (19:23 +0000)]
Add support for the special shell nologin.

Prodded by: mikeh

21 years agoMention the use of Framed-IPv6-Prefix.
Hajimu UMEMOTO [Fri, 20 Jun 2003 18:14:57 +0000 (18:14 +0000)]
Mention the use of Framed-IPv6-Prefix.

MFC after: 1 week

21 years agoamd64 is actually stable enough for the regular -CURRENT tinderbox now.
Dag-Erling Smørgrav [Fri, 20 Jun 2003 16:57:37 +0000 (16:57 +0000)]
amd64 is actually stable enough for the regular -CURRENT tinderbox now.

21 years agoUpdate my email address.
Mike Makonnen [Fri, 20 Jun 2003 16:53:36 +0000 (16:53 +0000)]
Update my email address.

21 years agoDon't forget the -r on the second password prompt.
Mike Makonnen [Fri, 20 Jun 2003 16:52:06 +0000 (16:52 +0000)]
Don't forget the -r on the second password prompt.

PR: bin/53550

21 years agoIPV6PREFIX is set when Framed-IPv6-Prefix is defined, You may
Hajimu UMEMOTO [Fri, 20 Jun 2003 16:15:59 +0000 (16:15 +0000)]
IPV6PREFIX is set when Framed-IPv6-Prefix is defined,  You may
want to pass the value to upper layer protocol such as DHCPv6
for prefix delegation.

MFC after: 1 week

21 years agoMention the use of Framed-IPv6-Route.
Hajimu UMEMOTO [Fri, 20 Jun 2003 15:44:52 +0000 (15:44 +0000)]
Mention the use of Framed-IPv6-Route.

MFC after: 1 week

21 years agoMerge from NetBSD src/sys/ntfs/ntfs_subr.c 1.5 & 1.30 (jdolecek):
Tim J. Robbins [Fri, 20 Jun 2003 14:52:52 +0000 (14:52 +0000)]
Merge from NetBSD src/sys/ntfs/ntfs_subr.c 1.5 & 1.30 (jdolecek):
- Avoid calling bread() with different sizes on the same blkno.
  Although the buffer cache is designed to handle differing size
  buffers, it erroneously tries to write the incorrectly-sized buffer
  buffer back to disk before reading the correctly-sized one, even
  when it's not dirty. This behaviour caused a panic for read-only
  NTFS mounts when INVARIANTS was enabled ("bundirty: buffer x still
  on queue y"), reported by NAKAJI Hiroyuki.
- Fix a bug in the code handling holes: a variable was incremented
  instead of decremented, which could cause an infinite loop.

21 years agoUse direct map in pmap_map().
Hidetoshi Shimokawa [Fri, 20 Jun 2003 14:09:33 +0000 (14:09 +0000)]
Use direct map in pmap_map().

This saves much KVA for vm_pages and you don't need to increase NKPT
for large physical memory anymore.

Suggested by: dfr

21 years agoFix few typos and remove two unneeded checks
Alexey Zelkin [Fri, 20 Jun 2003 13:29:43 +0000 (13:29 +0000)]
Fix few typos and remove two unneeded checks

Noticed by: ache

21 years agoWe have manpage now, so remove NOMAN
Alexey Zelkin [Fri, 20 Jun 2003 13:24:46 +0000 (13:24 +0000)]
We have manpage now, so remove NOMAN

21 years agoFix most warnings, sort prototypes, kill spaces at EOLs
Alexey Zelkin [Fri, 20 Jun 2003 13:23:51 +0000 (13:23 +0000)]
Fix most warnings, sort prototypes, kill spaces at EOLs

21 years agoAdd comment about **vpp being special-cased in vnode_if.awk (1.38)
Stefan Eßer [Fri, 20 Jun 2003 12:24:06 +0000 (12:24 +0000)]
Add comment about **vpp being special-cased in vnode_if.awk (1.38)

21 years agoThe assertions generated to test VFS locking never included checks
Stefan Eßer [Fri, 20 Jun 2003 12:15:37 +0000 (12:15 +0000)]
The assertions generated to test VFS locking never included checks
for vnodes reached through double indirection (i.e. **vpp). This
is worked-around by special-casing the identifier "vpp" (adding one
level of indirection).

The alternative fix mentioned in the PR had required substantial
changes to this script.

In case there are locking violations that had been hidden without
this patch, they may suddenly show up, now ...

This change does not affect code compiled without DEBUG_VFS_LOCKS.

PR: kern/46652

21 years agoAdd minimalistic manual page for locale(1)
Alexey Zelkin [Fri, 20 Jun 2003 11:52:05 +0000 (11:52 +0000)]
Add minimalistic manual page for locale(1)

21 years agoComplete rewrite of locale(1) in order to become POSIX complaint utilitty.
Alexey Zelkin [Fri, 20 Jun 2003 11:45:43 +0000 (11:45 +0000)]
Complete rewrite of locale(1) in order to become POSIX complaint utilitty.

It's possibly not completely complaint with POSIX requirements, but
very close to it now.

21 years agoAdd an amd64 category now that amd64 has its own mailing list and all.
Dag-Erling Smørgrav [Fri, 20 Jun 2003 10:48:26 +0000 (10:48 +0000)]
Add an amd64 category now that amd64 has its own mailing list and all.

21 years agoAdd devname_r(3) which takes a buffer as argument.
Poul-Henning Kamp [Fri, 20 Jun 2003 09:52:27 +0000 (09:52 +0000)]
Add devname_r(3) which takes a buffer as argument.

21 years agoRemove dev_mkdb(8). The kernel answers these questions.
Poul-Henning Kamp [Fri, 20 Jun 2003 09:48:11 +0000 (09:48 +0000)]
Remove dev_mkdb(8).  The kernel answers these questions.

21 years agoRemove the devdb stuff, the kernel answers these questions just fine.
Poul-Henning Kamp [Fri, 20 Jun 2003 09:47:09 +0000 (09:47 +0000)]
Remove the devdb stuff, the kernel answers these questions just fine.

21 years agocpu_set_upcall_kse needs to access userspace, release schedule lock
David Xu [Fri, 20 Jun 2003 09:12:12 +0000 (09:12 +0000)]
cpu_set_upcall_kse needs to access userspace, release schedule lock
before calling it for bound thread. To avoid this problem, change
thread_schedule_upcall to not put new thread on run queue, let caller
do it, so we can tweak the new thread before setting it to run.

Reported by: pho

21 years agoDon't put callout_lock under #ifdef DIAGNOSTIC despite the fact that it
Poul-Henning Kamp [Fri, 20 Jun 2003 08:39:04 +0000 (08:39 +0000)]
Don't put callout_lock under #ifdef DIAGNOSTIC despite the fact that it
works anyway.

21 years agoInitialize b_saveaddr when we hand out pbufs
Poul-Henning Kamp [Fri, 20 Jun 2003 08:35:28 +0000 (08:35 +0000)]
Initialize b_saveaddr when we hand out pbufs

21 years agoInitialize b_saveaddr when we hand out buffers
Poul-Henning Kamp [Fri, 20 Jun 2003 08:26:38 +0000 (08:26 +0000)]
Initialize b_saveaddr when we hand out buffers

21 years agoCrude but efficient:
Poul-Henning Kamp [Fri, 20 Jun 2003 08:07:15 +0000 (08:07 +0000)]
Crude but efficient:

#ifdef DIAGNOSTIC hold a mutex while calling callout's so that we hear
about it if they sleep.

21 years agoDon't (re)initialize f_gcflag to zero.
Poul-Henning Kamp [Fri, 20 Jun 2003 08:02:30 +0000 (08:02 +0000)]
Don't (re)initialize f_gcflag to zero.

Move initialization of DTYPE_VNODE specific field f_seqcount into
the DTYPE_VNODE specific code.

21 years agoMove FMARK and FDEFER til sys/file.h where they belong.
Poul-Henning Kamp [Fri, 20 Jun 2003 07:59:59 +0000 (07:59 +0000)]
Move FMARK and FDEFER til sys/file.h where they belong.

Order the fields in struct file in sections after their scope.

21 years agofix another LP64 problem. READ_IVAR takes a pointer to an uintptr_t, not
John-Mark Gurney [Fri, 20 Jun 2003 07:22:54 +0000 (07:22 +0000)]
fix another LP64 problem.  READ_IVAR takes a pointer to an uintptr_t, not
an int.

21 years agofix problem with uninitalized ptr (from PR)
John-Mark Gurney [Fri, 20 Jun 2003 04:54:27 +0000 (04:54 +0000)]
fix problem with uninitalized ptr (from PR)

other minor clean up.

PR: bin/2785
Submitted by: mpp

21 years agoUpdate the description of the Netgear MA401 to say Prism-II/2.5;
Mike Silbersack [Fri, 20 Jun 2003 04:27:55 +0000 (04:27 +0000)]
Update the description of the Netgear MA401 to say Prism-II/2.5;
my 401 is a 2.5, so presumably many of them are.

MFC after: 3 days

21 years agoThe so-called "optimized copy-on-write fault" case should not require
Alan Cox [Fri, 20 Jun 2003 04:20:36 +0000 (04:20 +0000)]
The so-called "optimized copy-on-write fault" case should not require
the vm map lock.  What's really needed is vm object locking, which
is (for the moment) provided Giant.

Reviewed by: tegge

21 years agoWhen a STOP signal is being sent to a process, it is possible all
David Xu [Fri, 20 Jun 2003 03:36:45 +0000 (03:36 +0000)]
When a STOP signal is being sent to a process, it is possible all
threads in the process have already masked the signal, so job control
is delayed. But later a thread unmasking the STOP signal should enable
job control, so in issignal(), scanning all threads in process to see
if we can direct suspend some of them, not just suspend current thread.

21 years agoFix typo. td should be td0.
David Xu [Fri, 20 Jun 2003 01:56:28 +0000 (01:56 +0000)]
Fix typo. td should be td0.

21 years agoRemove near-duplicate entry, keeping the better attributed one.
Gregory Sutter [Thu, 19 Jun 2003 22:52:50 +0000 (22:52 +0000)]
Remove near-duplicate entry, keeping the better attributed one.

21 years agoCorrect an apparent typo.
Matthew N. Dodd [Thu, 19 Jun 2003 22:12:54 +0000 (22:12 +0000)]
Correct an apparent typo.

PR:  bin/53515
Submitted by:  Lukas Ertl <l.ertl@univie.ac.at>
Approved by:  jmallett

21 years agoFix a coredump that would occur when fdopen was unable to
Mike Silbersack [Thu, 19 Jun 2003 20:04:51 +0000 (20:04 +0000)]
Fix a coredump that would occur when fdopen was unable to
return a valid fd.

PR: 49096
Submitted by: demon
MFC after: 3 days

21 years agoAvoid using v8 opcodes; use ba instead of b for unconditional branches.
Jake Burkholder [Thu, 19 Jun 2003 19:11:21 +0000 (19:11 +0000)]
Avoid using v8 opcodes; use ba instead of b for unconditional branches.

21 years agoDo RADIUS accounting on IPV6CP.
Hajimu UMEMOTO [Thu, 19 Jun 2003 18:55:49 +0000 (18:55 +0000)]
Do RADIUS accounting on IPV6CP.

MFC after: 1 week

21 years agoOops, I put unused variable in my previous commit.
Hajimu UMEMOTO [Thu, 19 Jun 2003 18:28:37 +0000 (18:28 +0000)]
Oops, I put unused variable in my previous commit.

MFC after: 1 week

21 years agoInstall routes specified by Framed-IPv6-Route. Since the format
Hajimu UMEMOTO [Thu, 19 Jun 2003 18:19:31 +0000 (18:19 +0000)]
Install routes specified by Framed-IPv6-Route.  Since the format
of Framed-IPv6-Route is user defined, it follows Framed-IP-route.

MFC after: 1 week

21 years agoUnlock the struct file lock before aquiring Giant, otherwise
Alfred Perlstein [Thu, 19 Jun 2003 18:13:07 +0000 (18:13 +0000)]
Unlock the struct file lock before aquiring Giant, otherwise
we can deadlock because of lock order reversals.  This was not
caught because Witness ignores pool mutexes right now.

Diagnosis and help: truckman
Noticed by: pho

21 years agoHook openpromio up to the build.
Jake Burkholder [Thu, 19 Jun 2003 18:06:58 +0000 (18:06 +0000)]
Hook openpromio up to the build.

21 years agoLock the vm object when freeing a vm page.
Alan Cox [Thu, 19 Jun 2003 17:56:12 +0000 (17:56 +0000)]
Lock the vm object when freeing a vm page.

21 years agoFix warnings on 64 bit platforms.
Matthew N. Dodd [Thu, 19 Jun 2003 16:09:18 +0000 (16:09 +0000)]
Fix warnings on 64 bit platforms.

Noticed by:  jake

21 years agoAdd support for the HighPoint HPT302 & HPT371
Søren Schmidt [Thu, 19 Jun 2003 15:11:04 +0000 (15:11 +0000)]
Add support for the HighPoint HPT302 & HPT371

HW sponsored by: Martin Blapp <mbr@FreeBSD.ORG>

21 years agoFix direct map page table for 2GB+ physical memory.
Hidetoshi Shimokawa [Thu, 19 Jun 2003 12:14:37 +0000 (12:14 +0000)]
Fix direct map page table for 2GB+ physical memory.

You may still need to increase NKPT for larger memory.
I have successfully booted 8GB system with NKPT=256.

21 years agoFix broken -a functionality.
Martin Blapp [Thu, 19 Jun 2003 09:19:51 +0000 (09:19 +0000)]
Fix broken -a functionality.

Reviewed by: phk
Submitted by: Martin Kammerhofer <mkamm@gmx.net>
PR: 53451

MFC: 1 week

21 years agoGerman keymap with dead keys.
Murray Stokely [Thu, 19 Jun 2003 08:34:38 +0000 (08:34 +0000)]
German keymap with dead keys.

PR: misc/28456
Submitted by: Ulrich Luttner <luttner@web.de>
MFC After: 1 week

21 years agoAdd US syscons keymap w/ accents.
Murray Stokely [Thu, 19 Jun 2003 08:08:17 +0000 (08:08 +0000)]
Add US syscons keymap w/ accents.

PR: misc/18459

21 years agoRemove bogus non-reentrant "temporary" implementation of gethostbyaddr_r()
Kris Kennaway [Thu, 19 Jun 2003 07:57:11 +0000 (07:57 +0000)]
Remove bogus non-reentrant "temporary" implementation of gethostbyaddr_r()
that has been here for 6 years and 9 months.

Reviewed by: deischen
MFC After: 1 week

21 years agophk added the -x option in May 2000, but only for regular files; for
Murray Stokely [Thu, 19 Jun 2003 07:24:26 +0000 (07:24 +0000)]
phk added the -x option in May 2000, but only for regular files; for
special files it was treated like -l.  This commit adds the -x option
in for special files as well.

PR: bin/46249
Submitted by: Colin Percival <cperciva@sfu.ca>

21 years agoIn the build process, fortune files have been randomized since at
Murray Stokely [Thu, 19 Jun 2003 07:02:00 +0000 (07:02 +0000)]
In the build process, fortune files have been randomized since at
least the 4.4-lite days.  This is pointless, since fortune(6) performs
its own randomization.  It was also problematic for binary update
systems such as ports/security/freebsd-update.  This commit simply
removes the -r option to strfile so that the datfiles are constant.

Submitted by: Colin Percival <cperciva@sfu.ca>

21 years agoAdd TLS related relocation.
Marcel Moolenaar [Thu, 19 Jun 2003 06:51:43 +0000 (06:51 +0000)]
Add TLS related relocation.

21 years agoAdd a ratelimited message of the form
Mike Silbersack [Thu, 19 Jun 2003 05:57:25 +0000 (05:57 +0000)]
Add a ratelimited message of the form
"maxproc limit exceeded by uid %i, please see tuning(7) and login.conf(5)."

Which will be triggered whenever a user hits his/her maxproc limit or
the systemwide maxproc limit is reached.

MFC after: 1 week

21 years agoAdd function prototypes.
Matthew N. Dodd [Thu, 19 Jun 2003 05:28:26 +0000 (05:28 +0000)]
Add function prototypes.

21 years ago- Rename the IPI_WAIT macro to IPI_DONE.
Jake Burkholder [Thu, 19 Jun 2003 05:27:04 +0000 (05:27 +0000)]
- Rename the IPI_WAIT macro to IPI_DONE.
- Don't require all receivers of ipis to wait for all other receivers,
  only that the sender wait for all receivers.  This should reduce the
  amount of time spent with interrupts disabled, which may be a cause
  of ipi timeouts.

Discussed with: tmm

21 years agoLD_DUMP_REL_PRE and LD_DUMP_REL_POST don't output to stderr; don't
Matthew N. Dodd [Thu, 19 Jun 2003 04:34:09 +0000 (04:34 +0000)]
LD_DUMP_REL_PRE and LD_DUMP_REL_POST don't output to stderr; don't
claim that they do.