]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoSplit the mlock() kernel code into two parts, mlock(), which unpacks
truckman [Thu, 26 Feb 2004 00:27:04 +0000 (00:27 +0000)]
Split the mlock() kernel code into two parts, mlock(), which unpacks
the syscall arguments and does the suser() permission check, and
kern_mlock(), which does the resource limit checking and calls
vm_map_wire().  Split munlock() in a similar way.

Enable the RLIMIT_MEMLOCK checking code in kern_mlock().

Replace calls to vslock() and vsunlock() in the sysctl code with
calls to kern_mlock() and kern_munlock() so that the sysctl code
will obey the wired memory limits.

Nuke the vslock() and vsunlock() implementations, which are no
longer used.

Add a member to struct sysctl_req to track the amount of memory
that is wired to handle the request.

Modify sysctl_wire_old_buffer() to return an error if its call to
kern_mlock() fails.  Only wire the minimum of the length specified
in the sysctl request and the length specified in its argument list.
It is recommended that sysctl handlers that use sysctl_wire_old_buffer()
should specify reasonable estimates for the amount of data they
want to return so that only the minimum amount of memory is wired
no matter what length has been specified by the request.

Modify the callers of sysctl_wire_old_buffer() to look for the
error return.

Modify sysctl_old_user to obey the wired buffer length and clean up
its implementation.

Reviewed by: bms

20 years agoAssert pipe mutex in pipeselwakeup(), as we manipulate pipe_state
rwatson [Thu, 26 Feb 2004 00:18:22 +0000 (00:18 +0000)]
Assert pipe mutex in pipeselwakeup(), as we manipulate pipe_state
in a non-atomic manner.  It appears to always be called with the
mutex (good).

20 years agoAdd a note about the routed update breaking compatibility (for MD5
bms [Wed, 25 Feb 2004 23:56:30 +0000 (23:56 +0000)]
Add a note about the routed update breaking compatibility (for MD5
authentication only) with older versions of FreeBSD's routed.

20 years agoSync HEAD sources to vendor branch import of routed v2.27 from rhyolite.com.
bms [Wed, 25 Feb 2004 23:45:57 +0000 (23:45 +0000)]
Sync HEAD sources to vendor branch import of routed v2.27 from rhyolite.com.
Update <protocols/routed.h> for the MD5 changes requested in bin/35843.
Preserve local changes.

Education by: obrien, markm, pointy-stick
PR: bin/35843 (and doubtless others)

20 years agoUpdate comment regarding MAC labels: we no longer pass endpoints
rwatson [Wed, 25 Feb 2004 23:30:56 +0000 (23:30 +0000)]
Update comment regarding MAC labels: we no longer pass endpoints
into the MAC Framework, just the pipe pair.

GC 'hadpeer' used in pipedestroy(), which is no longer needed as
we check pipe_present flags on the pair.

20 years agoSince we don't use PG_NX yet, don't turn on EFER_NXE quite yet. This needs
peter [Wed, 25 Feb 2004 23:12:39 +0000 (23:12 +0000)]
Since we don't use PG_NX yet, don't turn on EFER_NXE quite yet.  This needs
to be done based on the cpuid bits.  AMD says that we should test the cpuid
features bits for certain things, such as this.

20 years agoFix endianes bug
emax [Wed, 25 Feb 2004 22:43:43 +0000 (22:43 +0000)]
Fix endianes bug

20 years agoUse the -H option instead of the deprecated -follow predicate.
des [Wed, 25 Feb 2004 21:45:41 +0000 (21:45 +0000)]
Use the -H option instead of the deprecated -follow predicate.

20 years agoMake the resolver(3) and many associated interfaces much more reentrant.
green [Wed, 25 Feb 2004 21:03:46 +0000 (21:03 +0000)]
Make the resolver(3) and many associated interfaces much more reentrant.
The getaddrinfo(3), getipnodebyname(3) and resolver(3) can coincide now
with what should be totally reentrant, and h_errno values will now
be preserved correctly, but this does not affect interfaces such as
gethostbyname(3) which are still mostly non-reentrant.

In all of these relevant functions, the thread-safety has been pushed
down as far as it seems possible right now.  This means that operations
that are selected via nsdispatch(3) (i.e. files, yp, dns) are protected
still under global locks that getaddrinfo(3) defines, but where possible
the locking is greatly reduced.  The most noticeable improvement is
that multiple DNS lookups can now be run at the same time, and this
shows major improvement in performance of DNS-lookup threaded programs,
and solves the "Mozilla tab serialization" problem.

No single-threaded applications need to be recompiled.  Multi-threaded
applications that reference "_res" to change resolver(3) options will
need to be recompiled, and ones which reference "h_errno" will also
if they desire the correct h_errno values.  If the applications already
understood that _res and h_errno were not thread-safe and had their own
locking, they will see no performance improvement but will not
actually break in any way.

Please note that when NSS modules are used, or when nsdispatch(3)
defaults to adding any lookups of its own to the individual libc
_nsdispatch() calls, those MUST be reentrant as well.

20 years agoAccording to source code, under certain conditions, logging goes to the
charnier [Wed, 25 Feb 2004 20:31:00 +0000 (20:31 +0000)]
According to source code, under certain conditions, logging goes to the
"auth" facility not "daemon".
Submitted by: "Bill Richter (7X22KEY)" <richterb@binkley.foothill.net>

20 years agoRe-remove MT_TAGs. The problems with dummynet have been fixed now.
mlaier [Wed, 25 Feb 2004 19:55:29 +0000 (19:55 +0000)]
Re-remove MT_TAGs. The problems with dummynet have been fixed now.

Tested by: -current, bms(mentor), me
Approved by: bms(mentor), sam

20 years agoRemove freebsd-update -- it breaks the Alpha, sparc64, ia64, and amd64
obrien [Wed, 25 Feb 2004 19:05:47 +0000 (19:05 +0000)]
Remove freebsd-update -- it breaks the Alpha, sparc64, ia64, and amd64
release bulds.

20 years agoDo not depend on existence of _end symbol in obj_from_addr, use
kan [Wed, 25 Feb 2004 17:06:16 +0000 (17:06 +0000)]
Do not depend on existence of _end symbol in obj_from_addr, use
obj->mapbase and obj->mapsize instead.

Prompted by:  OpenOffice debugging session at last BSDCon.

20 years agoAdd FreeBSD 5.2.1
wosch [Wed, 25 Feb 2004 15:58:14 +0000 (15:58 +0000)]
Add FreeBSD 5.2.1

20 years agoUpdate manual page.
phk [Wed, 25 Feb 2004 13:55:56 +0000 (13:55 +0000)]
Update manual page.

Give 8" example for the heck of it.

20 years agoAdd SiI3512.
brueffer [Wed, 25 Feb 2004 13:48:48 +0000 (13:48 +0000)]
Add SiI3512.

20 years agoMake mode setting with fdcontrol(8) stick.
phk [Wed, 25 Feb 2004 13:44:58 +0000 (13:44 +0000)]
Make mode setting with fdcontrol(8) stick.

Recognize when configured for "auto".

20 years agoRecognize "auto" format.
phk [Wed, 25 Feb 2004 13:43:39 +0000 (13:43 +0000)]
Recognize "auto" format.

Be more verbose when asked to.

20 years agoSet size field correctly, it is number of sectors on the device, not
phk [Wed, 25 Feb 2004 13:43:17 +0000 (13:43 +0000)]
Set size field correctly, it is number of sectors on the device, not
number of 512 bytes sectors.

Recognize size == -1 as meaning "auto".

20 years agoDemangled vendor ids. Fixed misplaced FreeBSD id.
bde [Wed, 25 Feb 2004 13:26:53 +0000 (13:26 +0000)]
Demangled vendor ids.  Fixed misplaced FreeBSD id.

20 years agoFix style bug in last commit,
johan [Wed, 25 Feb 2004 13:12:51 +0000 (13:12 +0000)]
Fix style bug in last commit,
add a tab after WARNS?=.

While I'm here fix other style bugs.

Submitted by: bde (libbdf/Makefile)

20 years agoBacked out rev.1.6. A bogus include was added to work around breakage of
bde [Wed, 25 Feb 2004 13:11:48 +0000 (13:11 +0000)]
Backed out rev.1.6.  A bogus include was added to work around breakage of
<netinet/tcp_var.h>'s prerequisites.  Prerequistes should not grow for
userland headers, and <netinet/tcp_var.h> is unfortunately still needed
in userland.

20 years agoFixed namespace pollution in rev.1.74. Implementation of the syncache
bde [Wed, 25 Feb 2004 13:03:01 +0000 (13:03 +0000)]
Fixed namespace pollution in rev.1.74.  Implementation of the syncache
increased <netinet/tcp_var>'s already large set of prerequisites, and
this was handled badly.  Just don't declare the complete syncache struct
unless <netinet/pcb.h> is included before <netinet/tcp_var.h>.

Approved by: jlemon (years ago, for a more invasive fix)

20 years agoDon't use the negatively-opaque type uma_zone_t or be chummy with
bde [Wed, 25 Feb 2004 11:53:19 +0000 (11:53 +0000)]
Don't use the negatively-opaque type uma_zone_t or be chummy with
<vm/uma.h>'s idempotency indentifier or its misspelling.

20 years agoChange warning hint to be more useful
ache [Wed, 25 Feb 2004 10:56:54 +0000 (10:56 +0000)]
Change warning hint to be more useful

20 years agoAdd support for the sii3512 SATA chip.
sos [Wed, 25 Feb 2004 09:55:49 +0000 (09:55 +0000)]
Add support for the sii3512 SATA chip.

20 years agoFixed some insertion sort errors.
bde [Wed, 25 Feb 2004 09:35:35 +0000 (09:35 +0000)]
Fixed some insertion sort errors.

20 years agoFixed some style bugs in previous commit (unsorting of the DDB_* options,
bde [Wed, 25 Feb 2004 08:57:33 +0000 (08:57 +0000)]
Fixed some style bugs in previous commit (unsorting of the DDB_* options,
misofrmatting, and English usage errors).

20 years agoRelax a KASSERT condition to allow for a valid corner case where
hsu [Wed, 25 Feb 2004 08:53:17 +0000 (08:53 +0000)]
Relax a KASSERT condition to allow for a valid corner case where
the FIN on the last segment consumes an extra sequence number.

Spurious panic reported by Mike Silbersack <silby@silby.com>.

20 years agoRevert the last commit. I don't know what I was thinking, but this change
scottl [Wed, 25 Feb 2004 05:41:44 +0000 (05:41 +0000)]
Revert the last commit.  I don't know what I was thinking, but this change
definitely doesn't help any thing.

20 years agoAdd an 'add' command to ugidfw(8), which permits specifying a new
rwatson [Wed, 25 Feb 2004 03:59:56 +0000 (03:59 +0000)]
Add an 'add' command to ugidfw(8), which permits specifying a new
rule without explicitly specifying a new rule number.

Update copyrights, remove license clause three.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, McAfee Research

20 years agoAdd bsde_add_rule(), which is similar to bsde_set_rule() except that
rwatson [Wed, 25 Feb 2004 03:24:39 +0000 (03:24 +0000)]
Add bsde_add_rule(), which is similar to bsde_set_rule() except that
the caller does not specify the rule number -- instead, the kernel
module is probed for the next available rule, which is then used.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, McAfee Research

20 years agoBack out getopt* adding here per maintainer request
ache [Wed, 25 Feb 2004 03:10:14 +0000 (03:10 +0000)]
Back out getopt* adding here per maintainer request

20 years agoAdd upgrade instructions
ache [Wed, 25 Feb 2004 02:59:26 +0000 (02:59 +0000)]
Add upgrade instructions

20 years agoAdd "posix" subdir to "gnu"
ache [Wed, 25 Feb 2004 02:51:37 +0000 (02:51 +0000)]
Add "posix" subdir to "gnu"

20 years agoUpgrade to regex from grep 2.4.2
ache [Wed, 25 Feb 2004 02:41:15 +0000 (02:41 +0000)]
Upgrade to regex from grep 2.4.2

20 years agoMerge conflicts
ache [Wed, 25 Feb 2004 02:39:10 +0000 (02:39 +0000)]
Merge conflicts

20 years agoThis commit was generated by cvs2svn to compensate for changes in r126209,
ache [Wed, 25 Feb 2004 02:27:50 +0000 (02:27 +0000)]
This commit was generated by cvs2svn to compensate for changes in r126209,
which included commits to RCS files with non-trunk default branches.

20 years agoVirgin import (trimmed) of GNU regex, from GNU grep 2.4.2
ache [Wed, 25 Feb 2004 02:27:50 +0000 (02:27 +0000)]
Virgin import (trimmed) of GNU regex, from GNU grep 2.4.2

20 years agoBump FreeBSD_version for getopt_long_only() adding
ache [Wed, 25 Feb 2004 01:27:32 +0000 (01:27 +0000)]
Bump FreeBSD_version for getopt_long_only() adding

20 years agoNo reason to disable this test on powerpc.
grehan [Wed, 25 Feb 2004 00:53:06 +0000 (00:53 +0000)]
No reason to disable this test on powerpc.

20 years agoUse signed char cast to avoid out-of-range error on PowerPC (which has
grehan [Wed, 25 Feb 2004 00:52:14 +0000 (00:52 +0000)]
Use signed char cast to avoid out-of-range error on PowerPC (which has
unsigned char by default). This is a no-op on all other current arches.

Tested by: md5 sum before/after same on i386

20 years agoo Use `I-O DATA' instead of `IO DATA' as the vendor uses the former
non [Wed, 25 Feb 2004 00:13:43 +0000 (00:13 +0000)]
o Use `I-O DATA' instead of `IO DATA' as the vendor uses the former
  on its own site (http://www.iodata.jp/).
o Add a missing cross reference to cardbus(4).

Submitted by: rushani

20 years agoAdd DDB_NUMSYM option which in addition to the symbolic representation
phk [Tue, 24 Feb 2004 22:51:42 +0000 (22:51 +0000)]
Add DDB_NUMSYM option which in addition to the symbolic representation
also prints the actual numerical value of the symbol in question.

Users of addr2line(1) will be less proficient in hex arithmetic as a
consequence.

This amongst other things means that traceback lines change from:
   siointr1(c4016800,c073bda0,0,c06b699c,69f) at siointr1+0xc5
to
   siointr1(c4016800,c073bda0,0,c06b699c,69f) at 0xc062b0bd = siointr1+0xc5

I made this an option to avoid bikesheds.
~
~
~

20 years agoUnconst. Somebody didn't compile LINT.
phk [Tue, 24 Feb 2004 22:16:40 +0000 (22:16 +0000)]
Unconst.  Somebody didn't compile LINT.

20 years ago- Clarify ambiguous statement about not being able to both slock and xlock
jhb [Tue, 24 Feb 2004 21:36:39 +0000 (21:36 +0000)]
- Clarify ambiguous statement about not being able to both slock and xlock
  a single lock at the same time.
- Avoid using "own" to refer to holding either a shared or exclusive lock
  as it is only really correct for exclusive locks.
- Reword the sentence about sleep-ability to read easier.

Requested by: truckman (1)

20 years agoHave edquota honour MAXLOGNAME rather than hardcoding the maximum
ceri [Tue, 24 Feb 2004 20:58:16 +0000 (20:58 +0000)]
Have edquota honour MAXLOGNAME rather than hardcoding the maximum
username length to 29 characters.

PR: misc/62944
Submitted by: David Hill <david@wmol.com>
Approved by: ru
MFC after: 1 week

20 years agostyle.Makefile(5):
johan [Tue, 24 Feb 2004 20:52:58 +0000 (20:52 +0000)]
style.Makefile(5):
Use WARNS?= instead of WARNS=.

20 years agostyle.Makefile(5):
johan [Tue, 24 Feb 2004 20:51:20 +0000 (20:51 +0000)]
style.Makefile(5):
Use WARNS?= instead of WARNS=.

While I'm here,
use INTERNALPROG, instead if overriding install
remove emty lines

20 years agoWhitespace cleanup
des [Tue, 24 Feb 2004 19:31:30 +0000 (19:31 +0000)]
Whitespace cleanup

20 years agostyle.Makefile(5):
johan [Tue, 24 Feb 2004 19:23:33 +0000 (19:23 +0000)]
style.Makefile(5):
Use WARNS?= instead of WARNS=

For this to work properly for all part is the subdirectories
the WARNS assignments in Makefile.inc0 are moved to the correspondning
Makefile.inc.

Approved by: obrien (binutils maintainer)
Tested by: make universe

20 years agoCheck that a pointer is non-NULL before dereferencing it, not after.
cperciva [Tue, 24 Feb 2004 18:01:43 +0000 (18:01 +0000)]
Check that a pointer is non-NULL before dereferencing it, not after.

Reported by: "Ted Unangst" <tedu@coverity.com>
Approved by: rwatson (mentor)

20 years agoNote change of sizeof(struct tcpcb) and libkvm recompile.
andre [Tue, 24 Feb 2004 16:12:09 +0000 (16:12 +0000)]
Note change of sizeof(struct tcpcb) and libkvm recompile.

20 years agoin icmp6_mtudisc_update(), use ND link mtu to detect if the path MTU
ume [Tue, 24 Feb 2004 15:40:55 +0000 (15:40 +0000)]
in icmp6_mtudisc_update(), use ND link mtu to detect if the path MTU
should be updated.

Helped by: andre

20 years agoConvert the tcp segment reassembly queue to UMA and limit the maximum
andre [Tue, 24 Feb 2004 15:27:41 +0000 (15:27 +0000)]
Convert the tcp segment reassembly queue to UMA and limit the maximum
amount of segments it will hold.

The following tuneables and sysctls control the behaviour of the tcp
segment reassembly queue:

 net.inet.tcp.reass.maxsegments (loader tuneable)
  specifies the maximum number of segments all tcp reassemly queues can
  hold (defaults to 1/16 of nmbclusters).

 net.inet.tcp.reass.maxqlen
  specifies the maximum number of segments any individual tcp session queue
  can hold (defaults to 48).

 net.inet.tcp.reass.cursegments (readonly)
  counts the number of segments currently in all reassembly queues.

 net.inet.tcp.reass.overflows (readonly)
  counts how often either the global or local queue limit has been reached.

Tested by: bms, silby
Reviewed by: bms, silby

20 years agoWordsmithing
brueffer [Tue, 24 Feb 2004 15:00:03 +0000 (15:00 +0000)]
Wordsmithing

20 years agoDo not attempt to open NODEV
phk [Tue, 24 Feb 2004 09:59:35 +0000 (09:59 +0000)]
Do not attempt to open NODEV

20 years agoAdd getopt_long_only()
ache [Tue, 24 Feb 2004 08:09:20 +0000 (08:09 +0000)]
Add getopt_long_only()

20 years agoAdd getopt_long_only() from OpenBSD and other OpenBSD cleanups
ache [Tue, 24 Feb 2004 08:07:26 +0000 (08:07 +0000)]
Add getopt_long_only() from OpenBSD and other OpenBSD cleanups

PR:             63173
Submitted by:   Marius Strobl <marius@alchemy.franken.de>

20 years agoDon't set d_flags twice. The second setting clobbered D_NOGIANT.
bde [Tue, 24 Feb 2004 04:35:44 +0000 (04:35 +0000)]
Don't set d_flags twice.  The second setting clobbered D_NOGIANT.

20 years agoFixed the latest unsortings of CPU_ENABLE_*.
bde [Tue, 24 Feb 2004 04:18:07 +0000 (04:18 +0000)]
Fixed the latest unsortings of CPU_ENABLE_*.

20 years agoFixed missing libraries. This unbreaks the -DNO_DYNAMIC_ROOT and
bde [Tue, 24 Feb 2004 03:46:12 +0000 (03:46 +0000)]
Fixed missing libraries.  This unbreaks the -DNO_DYNAMIC_ROOT and
-DNOSHARED=yes cases.

20 years agoSend birthday presents to Wadham College, Oxford. :)
cperciva [Tue, 24 Feb 2004 01:44:47 +0000 (01:44 +0000)]
Send birthday presents to Wadham College, Oxford. :)

Approved by: rwatson (mentor)

20 years agoFix array overflow: If len=128, don't access [16] of a 16-byte IPv6
cperciva [Tue, 24 Feb 2004 01:20:51 +0000 (01:20 +0000)]
Fix array overflow: If len=128, don't access [16] of a 16-byte IPv6
address, even if we subsequently ignore its value by applying a >>8
to it.

Reported by: "Ted Unangst" <tedu@coverity.com>
Approved by: rwatson (mentor), {ume, suz} (KAME)

20 years agoPAO's work has been merged into FreeBSD not removed.
non [Mon, 23 Feb 2004 23:47:50 +0000 (23:47 +0000)]
PAO's work has been merged into FreeBSD not removed.
So remove the line.

Reviewed by: trhodes

20 years agoFix typo.
phk [Mon, 23 Feb 2004 23:39:42 +0000 (23:39 +0000)]
Fix typo.

Found by: imp

20 years agoFix typo
brueffer [Mon, 23 Feb 2004 22:00:44 +0000 (22:00 +0000)]
Fix typo

20 years agoApply style(9).
green [Mon, 23 Feb 2004 20:42:03 +0000 (20:42 +0000)]
Apply style(9).

Submitted by: "Christian S.J. Peron" <maneo@bsdpro.com>
PR: bin/63283

20 years agoAdd my birthday.
markus [Mon, 23 Feb 2004 20:39:23 +0000 (20:39 +0000)]
Add my birthday.

Approved by: arved (mentor)

20 years agostyle.Makefile(5):
johan [Mon, 23 Feb 2004 20:25:27 +0000 (20:25 +0000)]
style.Makefile(5):
Use WARNS?= instead of WARNS=.

20 years agoAdd support for Cronyx-Tau. For now I added only Tau-ISA files, system files
rik [Mon, 23 Feb 2004 20:19:00 +0000 (20:19 +0000)]
Add support for Cronyx-Tau. For now I added only Tau-ISA files, system files
would be changed in next patches, after extra verifications.

Approved by: imp (mentor)

20 years agostyle.Makefile:
johan [Mon, 23 Feb 2004 20:07:47 +0000 (20:07 +0000)]
style.Makefile:
Use WARNS?= instead of WARNS=.

20 years agoConnect gconcat(8) to the build.
pjd [Mon, 23 Feb 2004 20:06:56 +0000 (20:06 +0000)]
Connect gconcat(8) to the build.

Approved by: scottl (mentor)

20 years ago- Connect manual page for gconcat utility to the build.
pjd [Mon, 23 Feb 2004 20:06:21 +0000 (20:06 +0000)]
- Connect manual page for gconcat utility to the build.
- Fix usage.

Approved by: scottl (mentor) (implicity)

20 years agostyle.Makefile:
johan [Mon, 23 Feb 2004 20:05:14 +0000 (20:05 +0000)]
style.Makefile:
Use WARNS?= instead of WARNS=

20 years agoConnect geom_concat.ko module to the build.
pjd [Mon, 23 Feb 2004 20:03:53 +0000 (20:03 +0000)]
Connect geom_concat.ko module to the build.

Approved by: scottl (mentor)

20 years agoAdd ports-net-mgmt to the relevant supfiles.
marcus [Mon, 23 Feb 2004 19:07:47 +0000 (19:07 +0000)]
Add ports-net-mgmt to the relevant supfiles.

MFC after: 4 days

20 years agoA more accurate test in the new ufs_lock than that in 1.235.
mckusick [Mon, 23 Feb 2004 19:05:05 +0000 (19:05 +0000)]
A more accurate test in the new ufs_lock than that in 1.235.

20 years agoTeach rfcomm_pppd(8) how to register Bluetooth LAN service with local sdpd(8)
emax [Mon, 23 Feb 2004 17:30:59 +0000 (17:30 +0000)]
Teach rfcomm_pppd(8) how to register Bluetooth LAN service with local sdpd(8)
Add -u(unit) option

Reviewed by: imp (mentor), ru

20 years agoFix a problem with the USB keyboard driver not properly handling key
jhb [Mon, 23 Feb 2004 15:36:40 +0000 (15:36 +0000)]
Fix a problem with the USB keyboard driver not properly handling key
rollover resulting in duplicate keypress events.

PR: 57273
PR: 63171
Submitted by: plasma <plasma at freebsd.sinica.edu.tw>
Submitted by: Brian Candler <B.Candler at pobox.com>
MFC after: 1 week

20 years agoAdd missed CPU_ENABLE_LONGRON.
sobomax [Mon, 23 Feb 2004 14:49:15 +0000 (14:49 +0000)]
Add missed CPU_ENABLE_LONGRON.

Submitted by: Jon Noack <noackjr@alumni.rice.edu>

20 years agoAdd signal numbers to signal(3).
josef [Mon, 23 Feb 2004 13:20:03 +0000 (13:20 +0000)]
Add signal numbers to signal(3).

Approved by:    simon(mentor)
Reviewed by:    ru
Submitted by:   Stefan Farfeleder <stefan@fafoe.narf.at>
PR:             docs/60044

20 years agoAdded manual page for gconcat utility.
pjd [Mon, 23 Feb 2004 12:44:08 +0000 (12:44 +0000)]
Added manual page for gconcat utility.

Approved by: scottl (mentor) (implicity)
Reviewed by: simon

20 years agoTypo.
des [Mon, 23 Feb 2004 09:08:27 +0000 (09:08 +0000)]
Typo.

20 years agoFix vinums cdevsw{} to initialize d_version.
phk [Mon, 23 Feb 2004 08:55:13 +0000 (08:55 +0000)]
Fix vinums cdevsw{} to initialize d_version.

The nonstandard formatting made my mega-patch scripts miss it.

Retire the static major number while we're here anyway.

Reported by: Niels Chr. Bank-Pedersen <ncbp@bank-pedersen.dk>

20 years agoFix two oversights here: don't trash the freelist, and properly cleanup
phk [Mon, 23 Feb 2004 08:42:55 +0000 (08:42 +0000)]
Fix two oversights here:  don't trash the freelist, and properly cleanup
the cdevsw{}.

Submitted by: tegge

20 years agoIn the comment tell about optreset extension for NetBSD ghetopt_long()
ache [Mon, 23 Feb 2004 08:14:18 +0000 (08:14 +0000)]
In the comment tell about optreset extension for NetBSD ghetopt_long()

20 years agoIn the function clear_inodedeps(), a FREE_LOCK() should be called
mckusick [Mon, 23 Feb 2004 06:56:31 +0000 (06:56 +0000)]
In the function clear_inodedeps(), a FREE_LOCK() should be called
AFTER the call to vn_start_write(), not before it. Otherwise, it is
possible to unlock it multiple times if the vn_start_write() fails.

Submitted by: Juergen Hannken-Illjes <hannken@eis.cs.tu-bs.de>

20 years agoChange UFS from using vop_stdlock to using its own ufs_lock.
mckusick [Mon, 23 Feb 2004 06:40:17 +0000 (06:40 +0000)]
Change UFS from using vop_stdlock to using its own ufs_lock.
In ufs_lock, check for attempts to acquire shared locks on
snapshot files and change them to be exclusive locks. This
change eliminates deadlocks and machine lockups reported in
-current since most read requests started using shared lock
requests.

Submitted by: Jun Kuriyama <kuriyama@imgsrc.co.jp>

20 years agoAdd optarg, optind, opterr, optopt, optreset to SYNOPSYS
ache [Mon, 23 Feb 2004 05:07:11 +0000 (05:07 +0000)]
Add optarg, optind, opterr, optopt, optreset to SYNOPSYS

20 years agoAdd optreset to getopt.h too since NetBSD getopt_long() (but not GNU one)
ache [Mon, 23 Feb 2004 04:51:07 +0000 (04:51 +0000)]
Add optreset to getopt.h too since NetBSD getopt_long() (but not GNU one)
use it too.

20 years agoTry to better mimic GNU getopt.h which does not assume to make visible
ache [Mon, 23 Feb 2004 04:17:59 +0000 (04:17 +0000)]
Try to better mimic GNU getopt.h which does not assume to make visible
all unistd.h functions, use _GETOPT_DECLARE define for that.

20 years agoFix comment: GNU getopt.h to which this header tries to mimic does not declare
ache [Mon, 23 Feb 2004 03:47:28 +0000 (03:47 +0000)]
Fix comment: GNU getopt.h to which this header tries to mimic does not declare
getsubopt/optreset

20 years agoPOSIX clearly states that getsubopt() should be declared in <stdlib.h>
ache [Mon, 23 Feb 2004 03:32:10 +0000 (03:32 +0000)]
POSIX clearly states that getsubopt() should be declared in <stdlib.h>
not in <unistd.h>

20 years agoRemove unneccessary <unistd.h>
ache [Mon, 23 Feb 2004 03:30:02 +0000 (03:30 +0000)]
Remove unneccessary <unistd.h>

20 years agoLimit the amount of memory userspace processes can cause the kernel to
rwatson [Mon, 23 Feb 2004 03:18:18 +0000 (03:18 +0000)]
Limit the amount of memory userspace processes can cause the kernel to
allocate via DRI on r128 devices.

Obtained from: Thomas Biege <thomas@suse.de>
Reviewed by: scottl

20 years agoPOSIX clearly states that getsubopt() should be declared in <stdlib.h>,
ache [Mon, 23 Feb 2004 03:16:59 +0000 (03:16 +0000)]
POSIX clearly states that getsubopt() should be declared in <stdlib.h>,
not in <unistd.h>

20 years ago - Substitute bdone() and bwait() from vfs_bio.c for
alc [Mon, 23 Feb 2004 03:15:13 +0000 (03:15 +0000)]
 - Substitute bdone() and bwait() from vfs_bio.c for
   swap_pager_putpages()'s buffer completion code.  Note: the only
   difference between swp_pager_sync_iodone() and bdone(), aside from
   the locking in the latter, was the unnecessary clearing of B_ASYNC.
 - Remove an unnecessary pmap_page_protect() from
   swp_pager_async_iodone().

Reviewed by: tegge

20 years agoFix comment containing vop_readdir_args contents: a_cookies is really
tjr [Mon, 23 Feb 2004 01:43:00 +0000 (01:43 +0000)]
Fix comment containing vop_readdir_args contents: a_cookies is really
u_long ** not u_long *.

20 years agocookies is an array of u_long, not u_int, so MALLOC() it accordingly.
tjr [Mon, 23 Feb 2004 01:40:46 +0000 (01:40 +0000)]
cookies is an array of u_long, not u_int, so MALLOC() it accordingly.
Allocating it with the wrong size could have caused corruption on
64-bit architectures.