]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoCorrect spelling of "accommodate", "guarantee", "guaranteed", "guard",
Colin Percival [Thu, 28 Dec 2006 05:21:48 +0000 (05:21 +0000)]
Correct spelling of "accommodate", "guarantee", "guaranteed", "guard",
"return", "ridiculous", and "success".

MFC after: 3 days

17 years agoFix markup.
Ruslan Ermilov [Thu, 28 Dec 2006 04:58:38 +0000 (04:58 +0000)]
Fix markup.

17 years agoo unbreak rate set defaulting
Sam Leffler [Thu, 28 Dec 2006 01:31:26 +0000 (01:31 +0000)]
o unbreak rate set defaulting
o mark 11g mode support on finding 11g or pure 11g (OFDM-only)
  channels; was requiring pure 11g which caused some contortions
  in drivers that manually setup their channel lists

17 years agoFix shared authentication mode.
Jung-uk Kim [Thu, 28 Dec 2006 00:05:40 +0000 (00:05 +0000)]
Fix shared authentication mode.

17 years agoAvoid an instruction in atomic_cmpset_{int_long)() in most cases.
Bruce Evans [Wed, 27 Dec 2006 20:26:00 +0000 (20:26 +0000)]
Avoid an instruction in atomic_cmpset_{int_long)() in most cases.
These functions are used a lot for mutexes, so this reduces the text
size of an average kernel by about 0.75%.  This wasn't intended to
be a significant optimization, but it somehow increased the maximum
number of packets per second that can be transmitted by my bge hardware
from 320000 to 460000 (this benchmark is CPU-bound and remarkably
sensitive to changes in the text section).

Details: we would prefer to leave the result of the cmpxchg in %al,
but cannot tell gcc that it is there, so we have to convert it to an
integer register.  We converted  to %al, then to %[re]ax, but the
latter step is usually wasted since gcc usually only wants the condition
code and can recover it from %al just as easily as from %[re]ax.  Let
gcc promote %al in the few cases where this is needed.

Nearby style fixes;
- let gcc manage the load of `res', and don't abuse `res' for a copy of `exp'
- don't echo `res's name in comments
- consistently spell the condition code as 'e' after comparison for equality
- don't hard-code %al anywhere except in constraints
- for the version that doesn't use cmpxchg, there is no requirement to use
  %al anywhere, so don't hard-code it in the constraints either.

Style non-fix:
- for the versions that use cmpxchg, keep using "a" (was %[re]ax, now %al)
  for the main output operand, although this is not required.  The input
  and output operands that use the "a" constraint are now decoupled, and
  this makes things clearer except for the reason that the output register
  is hard-coded.  It is now just a hack to tell gcc that the input "a" has
  been clobbered without increasing the number of operands.

17 years agoAdd half/quarter rate 11a channel support:
Sam Leffler [Wed, 27 Dec 2006 19:07:09 +0000 (19:07 +0000)]
Add half/quarter rate 11a channel support:
o change handling of regdomain-related mib knobs so they can be set
  post-attach: regdomain, countrycode, outdoor, and xchanmode; the
  hal will not permit changing the regdomain but we expose it for now
o on regdomain/countrycode change recalculate the channel list and
  push it to the net80211 layer (NB: looks to need more tweaking)
o setup rate tables for half/quarter rate channels
o honor half/quarter rate channel configs when changing channels
o honor half/quarter rate channel configs when setting the slot time
o use hack/nonstandard channel numbering scheme for the public safety
  band to avoid overlapping 2.4G channels on dual-band cards
o remove setup of ic_sup_rates; the net80211 layer can do this for us
  and it simplifies handling of half/quarter rate channels

Tested only in Public Safety Band with cards that have RF5112.

17 years agoAdd half/quarter rate 11a channel support:
Sam Leffler [Wed, 27 Dec 2006 18:48:50 +0000 (18:48 +0000)]
Add half/quarter rate 11a channel support:
o add hack/nonstandard channel mapping for public safety band channels to
  mirror kernel (temporary until we have proper 802.11 state)
o change ieee80211_mhz2ieee to take channel flags (unused right now)

While here do some minor fixups like using IEEE80211_IS_CHAN_ANYG.

17 years agoFirst cut at half/quarter-rate 11a channel support (e.g. for use
Sam Leffler [Wed, 27 Dec 2006 18:46:18 +0000 (18:46 +0000)]
First cut at half/quarter-rate 11a channel support (e.g. for use
in the Public Safety Band):
o add channel flags to identify half/quarter-rate operation
o add rate sets (need to check spec on 4Mb/s in 1/4 rate)
o add if_media definitions for new rates
o split net80211 channel setup out into ieee80211_chan_init
o fixup ieee80211_mhz2ieee and ieee80211_ieee2mhz to understand half/quarter
  rate channels: note we temporarily use a nonstandard/hack numbering that
  avoids overlap with 2.4G channels because we don't (yet) have enough
  state to identify and/or map overlapping channel sets
o fixup ieee80211_ifmedia_init so it can be called post attach and will
  recalculate the channel list and associated state; this enables changing
  channel-related state like the regulatory domain after attach (will be
  needed for 802.11d support too)
o add ieee80211_get_suprates to return a reference to the supported rate
  set for a given channel
o add 3, 4.5, and 27 MB/s tx rates to rate <-> media conversion routines
o const-poison channel arg to ieee80211_chan2mode

17 years ago- Remove the ambiguity in the input format description.
Ruslan Ermilov [Wed, 27 Dec 2006 14:56:18 +0000 (14:56 +0000)]
- Remove the ambiguity in the input format description.
- Fix markup while here.

17 years agoFix markup and change some layout; no content changes.
Ruslan Ermilov [Wed, 27 Dec 2006 13:52:57 +0000 (13:52 +0000)]
Fix markup and change some layout; no content changes.

17 years agoMFp4:
Yaroslav Tykhiy [Wed, 27 Dec 2006 13:22:41 +0000 (13:22 +0000)]
MFp4:

Document my recent changes to rc.subr(8):
- there is $required_modules now;
- $required_* are checked before invoking a custom start method, too.

MFC after: 1 month

17 years agoMFp4:
Yaroslav Tykhiy [Wed, 27 Dec 2006 13:15:33 +0000 (13:15 +0000)]
MFp4:

Implement the checks for required_* objects as two functions, one
to be run before precmd and the other after it.  They get the current
rc command as an argument so they can choose what requirement tests
to perform.  As of now, only "start" needs such tests.

Implement a new requirement variable, required_modules.  It can
list kernel modules that need to be loaded after start_precmd
indicated success.  Each name in the list can be just "file", or
"file:module", or "file~regex".  This will allow us to remove a lot
of duplicated code from rc.d scripts.

Perform the checks not only for the default start method, but for
any method.  This allows for more flexibility and fixes a few rc.d
scripts (namely newsyslog, pf, sendmail) that rely on a required_*
variable while providing a non-default start method.

To be able to call the new check_required* functions naturally,
remove lots of crufty duplicated code pieces from run_rc_command
and replace each of them by a call to the helper function providing
a single corrected instance of the respective code snippet.  Now
run_rc_command isn't as scary as it used to be, and it even appears
to have quite a nice logic that was obscured by the old crufty code.

In the default handler for restart, run start from a subshell to
protect global varibles, e.g., _postcmd, from modification by the
start handler.  This enables using restart_postcmd. [x]

PR: conf/98734 [x]
Submitted by: Rick van der Zwet <rick@wzoeterwoude.net> [x]
Reviewed by: freebsd-rc (silence for an older version)
MFC after: 1 month

17 years agoFix markup.
Ruslan Ermilov [Wed, 27 Dec 2006 12:14:56 +0000 (12:14 +0000)]
Fix markup.

17 years agoMore markup fixes.
Ruslan Ermilov [Wed, 27 Dec 2006 11:28:00 +0000 (11:28 +0000)]
More markup fixes.

17 years agoFix markup.
Ruslan Ermilov [Wed, 27 Dec 2006 11:19:06 +0000 (11:19 +0000)]
Fix markup.

17 years agoFix markup.
Ruslan Ermilov [Tue, 26 Dec 2006 22:25:58 +0000 (22:25 +0000)]
Fix markup.

17 years agoAfter rev.1.169, the "interrupt" coalescing parameters are not used in
Bruce Evans [Tue, 26 Dec 2006 18:33:55 +0000 (18:33 +0000)]
After rev.1.169, the "interrupt" coalescing parameters are not used in
bge_intr().  Some of them are used in bge_poll().  Simplify by only
initializing these for polling mode and not toggling them when switching
modes.  This also fixes missing synchronization with the coalescing
engine in the toggling.

17 years agoBack out revision 1.33. usb/98983 was misfiled and the patch had no effect.
Jung-uk Kim [Tue, 26 Dec 2006 17:43:46 +0000 (17:43 +0000)]
Back out revision 1.33.  usb/98983 was misfiled and the patch had no effect.
The originator confirmed the adapter works fine without the patch.

Tested by: Massimo Lusetti (mlusetti at gmail dot com)

17 years agoGive a hint to the reader as to what the "whiteout" actually means.
Ruslan Ermilov [Tue, 26 Dec 2006 13:57:40 +0000 (13:57 +0000)]
Give a hint to the reader as to what the "whiteout" actually means.

17 years agoFix markup, add the EXIT STATUS section.
Ruslan Ermilov [Tue, 26 Dec 2006 13:44:41 +0000 (13:44 +0000)]
Fix markup, add the EXIT STATUS section.

17 years agoSimplify.
Ruslan Ermilov [Tue, 26 Dec 2006 12:43:12 +0000 (12:43 +0000)]
Simplify.

17 years agoAdd my birthday to calendar.freebsd
Dryice Dong Liu [Tue, 26 Dec 2006 11:33:26 +0000 (11:33 +0000)]
Add my birthday to calendar.freebsd

Approved by: itetcu (mentor)

17 years agoSay "utility", not "function".
Ruslan Ermilov [Tue, 26 Dec 2006 10:37:37 +0000 (10:37 +0000)]
Say "utility", not "function".

17 years agoAdd a sentence to the DESCRIPTION noting that updates aren't available
Colin Percival [Tue, 26 Dec 2006 05:49:06 +0000 (05:49 +0000)]
Add a sentence to the DESCRIPTION noting that updates aren't available
for -STABLE or -CURRENT.

Inspired by submission from: Scott Robbins
MFC after: 3 days

17 years agoturn non-INVARIANT KASSERT into an empty but real C
Kip Macy [Tue, 26 Dec 2006 02:48:42 +0000 (02:48 +0000)]
turn non-INVARIANT KASSERT into an empty but real C
statement so that syntax errors will still be caught
even if INVARIANTS aren't enabled in the config file

17 years agoFix markup.
Ruslan Ermilov [Mon, 25 Dec 2006 20:30:10 +0000 (20:30 +0000)]
Fix markup.

17 years agoSimplify and improve the markup.
Ruslan Ermilov [Mon, 25 Dec 2006 18:36:53 +0000 (18:36 +0000)]
Simplify and improve the markup.

17 years agoadd entry points required by newer broadcom wireless driver
Sam Leffler [Mon, 25 Dec 2006 17:04:41 +0000 (17:04 +0000)]
add entry points required by newer broadcom wireless driver

PR: kern/106131
Submitted by: Scot Hetzel
MFC after: 2 weeks

17 years agoFix build
Xin LI [Mon, 25 Dec 2006 17:03:04 +0000 (17:03 +0000)]
Fix build

17 years agoMarkup revision.
Ruslan Ermilov [Mon, 25 Dec 2006 15:09:06 +0000 (15:09 +0000)]
Markup revision.

17 years agoFreeBSD 5.5 is part of official -mdoc.
Ruslan Ermilov [Mon, 25 Dec 2006 09:09:47 +0000 (09:09 +0000)]
FreeBSD 5.5 is part of official -mdoc.

17 years agoMFV: Sync with vendor branch.
Ruslan Ermilov [Mon, 25 Dec 2006 09:03:10 +0000 (09:03 +0000)]
MFV: Sync with vendor branch.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r165538,
Ruslan Ermilov [Mon, 25 Dec 2006 08:59:09 +0000 (08:59 +0000)]
This commit was generated by cvs2svn to compensate for changes in r165538,
which included commits to RCS files with non-trunk default branches.

17 years agoPull up current -mdoc version from vendor.
Ruslan Ermilov [Mon, 25 Dec 2006 08:59:09 +0000 (08:59 +0000)]
Pull up current -mdoc version from vendor.

- New Darwin, FreeBSD, and NetBSD versions.
- DragonFly support including the new .Dx macro.
- New .St strings: -isoC-amd1, -isoC-tcor1, -isoC-tcor2, and -ieee1275-94.

17 years agobreak loop early if we know that there are at least two signals.
David Xu [Mon, 25 Dec 2006 03:00:15 +0000 (03:00 +0000)]
break loop early if we know that there are at least two signals.

17 years agoDocument the additional error returns possible when handling ELF objects
Joseph Koshy [Mon, 25 Dec 2006 02:24:39 +0000 (02:24 +0000)]
Document the additional error returns possible when handling ELF objects
that require extended numbering.

17 years agoKeep shadow copies of the `e_shnum', `e_phnum' and `e_shstrndx'
Joseph Koshy [Mon, 25 Dec 2006 02:22:22 +0000 (02:22 +0000)]
Keep shadow copies of the `e_shnum', `e_phnum' and `e_shstrndx'
members of the ELF Executable Header inside the library-private
`struct _Elf' descriptor and only update the underlying Elf{32,64}_Ehdr
structure on an elf_update(3) call.  These fields of the Ehdr
structure are technically `out of bounds' for an application program
per the ELF(3) API, but we've seen applications that initialize
a new Ehdr structure using memcpy(), messing up the library's
invariants. [1]

Implement elf_getphnum() and handle ELF objects with more than
64K program header table entries.

Reported by: jb [1]

17 years agoUse strncpy() instead of strlcpy() when copying members of
Joseph Koshy [Mon, 25 Dec 2006 02:06:32 +0000 (02:06 +0000)]
Use strncpy() instead of strlcpy() when copying members of
a `struct ar_hdr'.  These members do not use NUL-termination
while strlcpy() expects its source buffer to be NUL-terminated.

17 years ago- add ranged shootdowns when fewer than 64 mappings are being invalidated
Kip Macy [Mon, 25 Dec 2006 02:05:52 +0000 (02:05 +0000)]
- add ranged shootdowns when fewer than 64 mappings are being invalidated

17 years agoFix markup, remove a dummy BUGS section.
Ruslan Ermilov [Sun, 24 Dec 2006 23:07:11 +0000 (23:07 +0000)]
Fix markup, remove a dummy BUGS section.

17 years agoMFC noted: arcmsr 1.20.00.13.
Bruce A. Mah [Sun, 24 Dec 2006 22:52:27 +0000 (22:52 +0000)]
MFC noted:  arcmsr 1.20.00.13.

Updated release note:  libc resolver from BIND 9.3.3.

17 years agoFix a typo in __fpu_ftox() that caused long double to long (and long long)
Marius Strobl [Sun, 24 Dec 2006 22:24:39 +0000 (22:24 +0000)]
Fix a typo in __fpu_ftox() that caused long double to long (and long long)
conversion of negative numbers to always result in -1.
While at it, rearrange the nearby comment so it fits in 80 chars per line,
like the rest of this file does.

PR: 107130
MFC after: 1 day

17 years agoMarkup nits.
Ruslan Ermilov [Sun, 24 Dec 2006 19:55:38 +0000 (19:55 +0000)]
Markup nits.

17 years agoMarkup revision.
Ruslan Ermilov [Sun, 24 Dec 2006 19:18:52 +0000 (19:18 +0000)]
Markup revision.

17 years agoDocument the -d option.
Ruslan Ermilov [Sun, 24 Dec 2006 14:23:35 +0000 (14:23 +0000)]
Document the -d option.

17 years agoFix markup.
Ruslan Ermilov [Sun, 24 Dec 2006 13:58:17 +0000 (13:58 +0000)]
Fix markup.

17 years agoNitpicking: <sysexits.h> defines exit codes.
Ruslan Ermilov [Sun, 24 Dec 2006 10:25:23 +0000 (10:25 +0000)]
Nitpicking: <sysexits.h> defines exit codes.

17 years agoCorrect a logic error.
Joseph Koshy [Sun, 24 Dec 2006 09:45:10 +0000 (09:45 +0000)]
Correct a logic error.

17 years agoNote that rev. 1.221 introduced a local workaround for a general problem.
Yaroslav Tykhiy [Sun, 24 Dec 2006 08:52:13 +0000 (08:52 +0000)]
Note that rev. 1.221 introduced a local workaround for a general problem.
Add a pointer to the relevant PR for future reference.  The whole comment
will be OK to remove as soon as the general solution is applied.

PR: kern/105943

17 years ago- remove all calls to sched_pin and sched_unpin as they are only useful to
Kip Macy [Sun, 24 Dec 2006 08:03:27 +0000 (08:03 +0000)]
- remove all calls to sched_pin and sched_unpin as they are only useful to
  pmap on i386
- check for change in executable status in pmap_enter
- pmap_qenter and pmap_qremove only need to invalidate the range if one
  of the pages has been referenced
- remove pmap_kenter/pmap_kremove as they were only used by pmap_qenter
  and pmap_qremove
- in pmap_copy don't copy wired bit to destination pmap
- mpte was unused in pmap_enter_object - remove
- pmap_enter_quick_locked is not called on the kernel_pmap, remove check
- move pmap_remove_write specific logic out of tte_clear_phys_bit
- in pmap_protect check for removal of execute bit
- panic in the presence of a wired page in pmap_remove_all
- pmap_zero_range can call hwblkclr if offset is zero and size is PAGE_SIZE
- tte_clear_virt_bit is only used by pmap_change_wiring - thus it can be
  greatly simplified
- pmap_invalidate_page need only be called in tte_clear_phys_bit if there
  is a match with flags
- lock the pmap in tte_clear_phys_bit so that clearing the page bits is
  atomic with invalidating the page

- these changes result in 100s reduction in buildworld from a malloc backed
  disk to a malloc backed disk - ~2.5%

17 years agoDon't count on the first phys_avail range being greater than zero
Kip Macy [Sun, 24 Dec 2006 07:47:10 +0000 (07:47 +0000)]
Don't count on the first phys_avail range being greater than zero

17 years agoTCP Window scaling was being recognised but the recorded settings were being
Darren Reed [Sun, 24 Dec 2006 02:18:36 +0000 (02:18 +0000)]
TCP Window scaling was being recognised but the recorded settings were being
clobbered and thus effectively disabled.

MFC after: 7 days

17 years ago- resizing the tte_hash in pmap_copy is not likely to occur
Kip Macy [Sun, 24 Dec 2006 01:56:35 +0000 (01:56 +0000)]
- resizing the tte_hash in pmap_copy is not likely to occur
- the implementation also made the mistake of assuming the
  dst_pmap is the current pmap

17 years agoFix typo, p_slptime should be td_slptime.
David Xu [Sun, 24 Dec 2006 01:52:27 +0000 (01:52 +0000)]
Fix typo, p_slptime should be td_slptime.

17 years agoAdd missing articles.
Ruslan Ermilov [Sat, 23 Dec 2006 23:26:05 +0000 (23:26 +0000)]
Add missing articles.

17 years agoAdd missing things: a prototype and a const qualifier.
Yaroslav Tykhiy [Sat, 23 Dec 2006 22:39:38 +0000 (22:39 +0000)]
Add missing things: a prototype and a const qualifier.

Found by: WARNS=4

17 years agoRe-wrap comments following de-indentation.
Robert Watson [Sat, 23 Dec 2006 22:21:13 +0000 (22:21 +0000)]
Re-wrap comments following de-indentation.

17 years agoDrop all received data mbufs from a socket's queue if the MT_SONAME
Bruce M Simpson [Sat, 23 Dec 2006 21:07:07 +0000 (21:07 +0000)]
Drop all received data mbufs from a socket's queue if the MT_SONAME
mbuf is dropped, to preserve the invariant in the PR_ADDR case.

Add a regression test to detect this condition, but do not hook it
up to the build for now.

PR:             kern/38495
Submitted by:   James Juran
Reviewed by:    sam, rwatson
Obtained from:  NetBSD
MFC after:      2 weeks

17 years agoNits.
Ruslan Ermilov [Sat, 23 Dec 2006 19:15:39 +0000 (19:15 +0000)]
Nits.

17 years agoDynamically resize the Disk column. It was too narrow for modern
Yaroslav Tykhiy [Sat, 23 Dec 2006 18:54:49 +0000 (18:54 +0000)]
Dynamically resize the Disk column.  It was too narrow for modern
disk device names such as da0s1b.  So we also get rid of the nasty
constant 5 scattered over the code.

Implementing this change is a good chance to improve other bits
around it: init saved lengths early, always check return value from
kvm_getswapinfo().

17 years agoCall vnode_create_vobject() in VOP_OPEN. Makes mmap work on UDF filesystem.
Pav Lucistnik [Sat, 23 Dec 2006 18:53:22 +0000 (18:53 +0000)]
Call vnode_create_vobject() in VOP_OPEN.  Makes mmap work on UDF filesystem.

PR: kern/92040
Approved by: scottl
MFC after: 1 week

17 years agoClear to EOL after the end of meter so that its reading can decrease.
Yaroslav Tykhiy [Sat, 23 Dec 2006 18:03:10 +0000 (18:03 +0000)]
Clear to EOL after the end of meter so that its reading can decrease.

17 years agoMake it possible for meter to reach 100% mark when swap is totally full.
Yaroslav Tykhiy [Sat, 23 Dec 2006 17:46:32 +0000 (17:46 +0000)]
Make it possible for meter to reach 100% mark when swap is totally full.

17 years agoFix a deadlock in detach/shutdown.
Alfred Perlstein [Sat, 23 Dec 2006 17:18:18 +0000 (17:18 +0000)]
Fix a deadlock in detach/shutdown.

The problem was that I was acquiring the driver sx lock and then waiting
for a taskqueue to drain, however the taskqueue itself would try to
acquire the lock as well leading to a deadlock.

To fix the problem roll my own exclusive lock that allows for lock
cancellation.  This is a normal exclusive lock, however if someone
marks it as "dead" then all waiters who request an error return will
get back an error instead of continuing to wait for the lock.

In this particular case, the shutdown and detach functions kill the
lock while the async task thread tries to acquire the lock but will
abort if the lock returns an error.

The other option was to drop the driver lock mid-detach and mid-shutdown,
mid-detach was a ok, however mid-shutdown was not.

While I'm here, fix a bug in what appears to be the mii link status
word in the softc going out to lunch.  Explicitly set the status
word to 1 after initializing the mii.  This would result in an interface
that would never respond to "if_start" requests as the mii interface
would always look down.

17 years agoImprove markup.
Ruslan Ermilov [Sat, 23 Dec 2006 17:14:01 +0000 (17:14 +0000)]
Improve markup.

17 years agoImprove style:
Yaroslav Tykhiy [Sat, 23 Dec 2006 17:02:09 +0000 (17:02 +0000)]
Improve style:
- Don't define vars inside loops.
- Avoid useless casts.
- Use C idioms.
- Do alike things in a consistent way.

17 years agoReposition the "(swap not configured)" sign WRT the new layout.
Yaroslav Tykhiy [Sat, 23 Dec 2006 16:48:29 +0000 (16:48 +0000)]
Reposition the "(swap not configured)" sign WRT the new layout.

17 years agoEliminate a couple of screen coordinate variables
Yaroslav Tykhiy [Sat, 23 Dec 2006 16:39:09 +0000 (16:39 +0000)]
Eliminate a couple of screen coordinate variables
that were useless and just obfuscated the code.

17 years agoAdd some vertical whitespace for easier reading.
Yaroslav Tykhiy [Sat, 23 Dec 2006 16:31:06 +0000 (16:31 +0000)]
Add some vertical whitespace for easier reading.

17 years agoFix the swap display further:
Yaroslav Tykhiy [Sat, 23 Dec 2006 15:40:41 +0000 (15:40 +0000)]
Fix the swap display further:

1) Resize the Used column to avoid screen overflow if BLOCKSIZE is long.
2) Track the current swap configuration so that its changes don't break
   the display.

Suggested by: bde (1)

17 years agoFix the Total line shown if there are >1 swap devices.
Yaroslav Tykhiy [Sat, 23 Dec 2006 14:30:50 +0000 (14:30 +0000)]
Fix the Total line shown if there are >1 swap devices.

17 years agoStart fixing the "swap" display by saving one horizontal position.
Yaroslav Tykhiy [Sat, 23 Dec 2006 14:25:17 +0000 (14:25 +0000)]
Start fixing the "swap" display by saving one horizontal position.
Now the display won't overflow the 80-char row if BLOCKSIZE=1024.
The new spacing is also consistent with the "pigs" display.

17 years agoFix a typo in ata-queue.c
Remko Lodder [Sat, 23 Dec 2006 12:40:54 +0000 (12:40 +0000)]
Fix a typo in ata-queue.c

PR: kern/107100
Submitted by: Stephen J. Roznowski <sjr at comcast dot net>
Approved by: sos

17 years agoOne more nit.
Ruslan Ermilov [Sat, 23 Dec 2006 09:25:23 +0000 (09:25 +0000)]
One more nit.

17 years agoFix markup.
Ruslan Ermilov [Sat, 23 Dec 2006 09:22:06 +0000 (09:22 +0000)]
Fix markup.

17 years agoFic typo in previous commit.
Colin Percival [Sat, 23 Dec 2006 08:16:54 +0000 (08:16 +0000)]
Fic typo in previous commit.

MFC after: 3 days

17 years agoAdd two checks for inappropriate options: -U only makes sense in "x"
Colin Percival [Sat, 23 Dec 2006 07:06:37 +0000 (07:06 +0000)]
Add two checks for inappropriate options: -U only makes sense in "x"
mode, and --strip-components only makes sense in "x" and "t" modes.

MFC after: 3 days

17 years agoCorrect the names of some options.
Colin Percival [Sat, 23 Dec 2006 07:01:59 +0000 (07:01 +0000)]
Correct the names of some options.

MFC after: 3 days

17 years agoFix a panic when rebooting a SMP machine, when option STOP_NMI is used,
David Xu [Sat, 23 Dec 2006 03:30:50 +0000 (03:30 +0000)]
Fix a panic when rebooting a SMP machine, when option STOP_NMI is used,
nmi handler is used to stop other processors, nmi hander calls trap(),
however, trap() now accepts a pointer rather than a reference, this was
changed by kmacy@.

17 years agoUpdate comments to reflect changes in the extattrctl() code.
Robert Watson [Sat, 23 Dec 2006 00:30:03 +0000 (00:30 +0000)]
Update comments to reflect changes in the extattrctl() code.

Clean up comment formatting.

Obtained from: TrustedBSD Project

17 years agoImplement chunk allocation/deallocation hysteresis by caching one spare
Jason Evans [Sat, 23 Dec 2006 00:18:51 +0000 (00:18 +0000)]
Implement chunk allocation/deallocation hysteresis by caching one spare
chunk per arena, rather than immediately deallocating all unused chunks.
This fixes a potential performance issue when allocating/deallocating
an object of size (4kB..1MB] in a loop.

Reported by: davidxu

17 years agoConnect vfs_extattr.c to build by default.
Robert Watson [Sat, 23 Dec 2006 00:11:02 +0000 (00:11 +0000)]
Connect vfs_extattr.c to build by default.

17 years agoFollowing a repo-copy of vfs_syscalls.c to vfs_extattr.c, remove
Robert Watson [Sat, 23 Dec 2006 00:10:36 +0000 (00:10 +0000)]
Following a repo-copy of vfs_syscalls.c to vfs_extattr.c, remove
non-extattr functions from vfs_extattr.c, and extattr functions from
vfs_syscalls.c.

Change copyright/license on vfs_extattr.c to my copyright/license on
the extended attribute implementation (from extattr.h).

Clean up includes a bit.

Obtained from: TrustedBSD Project

17 years agoForced commit to recognize the repo-copy of vfs_syscalls.c to
Robert Watson [Fri, 22 Dec 2006 23:36:13 +0000 (23:36 +0000)]
Forced commit to recognize the repo-copy of vfs_syscalls.c to
vfs_extattr.c.  This follows a naming convention similar to the break-out
of the extended attribute system calls (etc) in NetBSD.

Thanks to: simon

17 years agoMove src/sys/sys/mac_policy.h, the kernel interface between the MAC
Robert Watson [Fri, 22 Dec 2006 23:34:47 +0000 (23:34 +0000)]
Move src/sys/sys/mac_policy.h, the kernel interface between the MAC
Framework and security modules, to src/sys/security/mac/mac_policy.h,
completing the removal of kernel-only MAC Framework include files from
src/sys/sys.  Update the MAC Framework and MAC policy modules.  Delete
the old mac_policy.h.

Third party policy modules will need similar updating.

Obtained from: TrustedBSD Project

17 years agoForced commit to recognize repo-copy of src/sys/sys/mac_policy.h to
Robert Watson [Fri, 22 Dec 2006 23:32:23 +0000 (23:32 +0000)]
Forced commit to recognize repo-copy of src/sys/sys/mac_policy.h to
src/sys/security/mac/mac_policy.h.

Thanks to: simon

17 years agoTweak some wording and markup.
Ruslan Ermilov [Fri, 22 Dec 2006 23:23:59 +0000 (23:23 +0000)]
Tweak some wording and markup.

17 years agoFix distorted sound on Intel Desktop Board D101GGC (ATI SB450/ALC861).
Ariff Abdullah [Fri, 22 Dec 2006 11:55:59 +0000 (11:55 +0000)]
Fix distorted sound on Intel Desktop Board D101GGC (ATI SB450/ALC861).

Turn on ALC861 quirk as default since it seems affecting all hardwares
with the same codec.

Reported/Tested by: Darryl Yeoh <drl@MyBSD.org.my>

17 years ago- Mention umask(2) when first referring to it.
Ruslan Ermilov [Fri, 22 Dec 2006 07:28:38 +0000 (07:28 +0000)]
- Mention umask(2) when first referring to it.
- Add missing markup.

Submitted by: Eugene Grosbein <eugen@grosbein.pp.ru>

17 years agoAvoid 'bogus' uninitialized warning by initializing rather than
Warner Losh [Fri, 22 Dec 2006 05:54:19 +0000 (05:54 +0000)]
Avoid 'bogus' uninitialized warning by initializing rather than
playing cute games.  It is much simpler, clearer and easier to follow
and understand.  Besides, gcc4 likes it better.

17 years agoThe second argument (type) to socket(2) is an enum, not a bitmask, so parse
Craig Rodrigues [Fri, 22 Dec 2006 05:07:21 +0000 (05:07 +0000)]
The second argument (type) to socket(2) is an enum, not a bitmask, so parse
it as an enum.

If an SCTP SOCK_SEQPACKET socket was opened, kdump would display this
wrong output:
 socket(PF_INET,SOCK_STREAM|SOCK_RDM|SOCK_SEQPACKET,0x84)

instead of this correct output:
 socket(PF_INET,SOCK_SEQPACKET,0x84)

MFC after: 2 weeks

17 years agoNote the second coming of MSI support in the bge driver.
John Polstra [Fri, 22 Dec 2006 03:03:31 +0000 (03:03 +0000)]
Note the second coming of MSI support in the bge driver.

17 years agoRe-enable MSI support for those chips on which it is believed to work
John Polstra [Fri, 22 Dec 2006 02:59:58 +0000 (02:59 +0000)]
Re-enable MSI support for those chips on which it is believed to work
properly.

17 years agoCorrect the BGE_CHIPID_BCM5750_B0 constant.
John Polstra [Fri, 22 Dec 2006 01:10:36 +0000 (01:10 +0000)]
Correct the BGE_CHIPID_BCM5750_B0 constant.

MFC after: 1 week

17 years agoACPIIO_BATT_GET_UNITS would always return ENXIO. However, it should never
Warner Losh [Fri, 22 Dec 2006 00:57:26 +0000 (00:57 +0000)]
ACPIIO_BATT_GET_UNITS would always return ENXIO.  However, it should never
return an error since it returns a count of battery devices in the system.
Set it to 0 explicitly, since it is the only switch branch that doesn't set
it.

# I guess no one uses it.

17 years agofix botch
Matt Jacob [Thu, 21 Dec 2006 23:20:51 +0000 (23:20 +0000)]
fix botch

17 years agoPrevent a line from being broken on a line boundary.
Ruslan Ermilov [Thu, 21 Dec 2006 22:59:07 +0000 (22:59 +0000)]
Prevent a line from being broken on a line boundary.

17 years agoImprove markup.
Ruslan Ermilov [Thu, 21 Dec 2006 22:44:41 +0000 (22:44 +0000)]
Improve markup.

17 years agoDocument MTX_NOPROFILE flag.
Pawel Jakub Dawidek [Thu, 21 Dec 2006 22:42:18 +0000 (22:42 +0000)]
Document MTX_NOPROFILE flag.

17 years agoAllow this module to get its options from the kernel build directory
Yaroslav Tykhiy [Thu, 21 Dec 2006 21:35:49 +0000 (21:35 +0000)]
Allow this module to get its options from the kernel build directory
instead of always hard coding them in CFLAGS.  POLA is kept here:
The module file built with GENERIC stays the same.

Tested with: md5(1)