]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
17 years agoFix a typo.
jhb [Mon, 12 Mar 2007 20:10:29 +0000 (20:10 +0000)]
Fix a typo.

17 years ago- Use m_gethdr(), m_get(), and m_clget() instead of the macros in
jhb [Mon, 12 Mar 2007 19:27:36 +0000 (19:27 +0000)]
- Use m_gethdr(), m_get(), and m_clget() instead of the macros in
  sosend_copyin().
- Use M_WAITOK instead of M_TRYWAIT in sosend_copyin().
- Don't check for NULL from M_WAITOK and return ENOBUFS.
  M_WAITOK/M_TRYWAIT allocations don't fail with NULL.

Reviewed by: andre
Requested by: andre (2)

17 years agoFix writes on siiprb type chips.
sos [Mon, 12 Mar 2007 15:34:08 +0000 (15:34 +0000)]
Fix writes on siiprb type chips.

17 years agoIn uipc_close(), we no longer always free the unpcb, as the last reference
rwatson [Mon, 12 Mar 2007 14:52:00 +0000 (14:52 +0000)]
In uipc_close(), we no longer always free the unpcb, as the last reference
may be dropped later.  In this case, always unlock the unpcb so as not to
leak the lock.

Found by: kris (BugMagnet)

17 years agoStop clobberring the application namespace with local
ru [Mon, 12 Mar 2007 13:54:52 +0000 (13:54 +0000)]
Stop clobberring the application namespace with local
variables such as "sect", "page", and "target"; use
underscored versions instead.

Discussed with: kan
MFC after: 3 days

17 years agoAttempt to load the kernel module only if we are going to create a
yar [Mon, 12 Mar 2007 13:08:56 +0000 (13:08 +0000)]
Attempt to load the kernel module only if we are going to create a
new interface.  In other cases loading the module is unwanted and
can lead to ill side effects.  One such effect found is as follows:
"kldunload if_foo" tells the module to kill all its interfaces,
which results in messages sent to devd; the module unloads.  Then
devd starts processing the messages, which ends up in a etc script
running ifconfig fooX, which reloads the module.

17 years agoEmit load and unload messages under bootverbose.
yar [Mon, 12 Mar 2007 12:42:14 +0000 (12:42 +0000)]
Emit load and unload messages under bootverbose.
This can help to spot bugs (which it did for me,)
and let people know which mode the vlan module is
actually using if they suspect it isn't picking its
options from the main kernel config file.

17 years agoFix some minor issues in the internal vlan lists:
yar [Mon, 12 Mar 2007 12:27:30 +0000 (12:27 +0000)]
Fix some minor issues in the internal vlan lists:

- ifv_list member of struct ifvlan is unneeded in array mode,
  it's used only in hash mode to resolve hash collisions.

- We don't need the list of trunks at all.  (The initial reason for
  having it was to be able to destroy all trunks in the MOD_UNLOAD
  handler, but a trunk is not to be destroyed forcibly -- it will
  go away when all vlan interfaces on it have been deleted.
  Note that if_clone_detach() called first of all under MOD_UNLOAD
  will delete all vlan interfaces and thus make all trunks go away
  quietly.)

- It's enough to use a single [S]LIST_FIRST() in a typical list
  destruction loop.

17 years agoAdd a pn_destroy field to pfs_node. This field points to a destructor
des [Mon, 12 Mar 2007 12:16:52 +0000 (12:16 +0000)]
Add a pn_destroy field to pfs_node.  This field points to a destructor
function which is called from pfs_destroy() before the node is reclaimed.

Modify pfs_create_{dir,file,link}() to accept a pointer to a destructor
function in addition to the usual attr / fill / vis pointers.

This breaks both the programming and binary interfaces between pseudofs
and its consumers.  It is believed that there are no pseudofs consumers
outside the source tree, so that the impact of this change is minimal.

Submitted by: Aniruddha Bohra <bohra@cs.rutgers.edu>

17 years agoAdd MAC, RX, and TX stats reporting via sysctl.
scottl [Mon, 12 Mar 2007 09:25:57 +0000 (09:25 +0000)]
Add MAC, RX, and TX stats reporting via sysctl.

17 years agoFix support for the VIA8237A SATA part.
sos [Mon, 12 Mar 2007 09:13:13 +0000 (09:13 +0000)]
Fix support for the VIA8237A SATA part.

HW sponsored by: Bob Bishop

17 years agoForced commit to note that the previous CVS comment is
mjacob [Mon, 12 Mar 2007 05:10:29 +0000 (05:10 +0000)]
Forced commit to note that the previous CVS comment is
incorrect. The MODULE_DEPENDS lines were put in (by me)
in 1.14 and removed in 1.15. The facts should be correctly
reported.

17 years agoAdd back in MODULE_DEPEND() lines that were lost in the rev 13 update.
scottl [Mon, 12 Mar 2007 05:02:42 +0000 (05:02 +0000)]
Add back in MODULE_DEPEND() lines that were lost in the rev 13 update.

17 years agoFix compilation issues found in RELENG_4 port and merge the
mjacob [Mon, 12 Mar 2007 04:54:30 +0000 (04:54 +0000)]
Fix compilation issues found in RELENG_4 port and merge the
diffs back to -current to keep versions identical.

17 years agoMinor Makefile cleanup. Do not use Makefile variables named 'version' and
kan [Mon, 12 Mar 2007 00:28:06 +0000 (00:28 +0000)]
Minor Makefile cleanup. Do not use Makefile variables named 'version' and
'target'. Latter is problematic in particular as apparently FreeBSD's
bsd.prog.mk re-defines it under some circumstances. This causes an
unexpected failures like -dumpmachine not working for cc while working
fine for c++.

Do not re-define IN_GCC in multipe places, it gets inherited from
Makefile.in anyway.

PR: gnu/110143
Submitted by: usleepless at gmail

17 years agoallow net80211 to fillin rate sets
sam [Sun, 11 Mar 2007 22:43:35 +0000 (22:43 +0000)]
allow net80211 to fillin rate sets

MFC after: 2 weeks

17 years agoBuild updates for tcsh-6.15.00 import.
mp [Sun, 11 Mar 2007 22:41:19 +0000 (22:41 +0000)]
Build updates for tcsh-6.15.00 import.

Reviewed by: ume
Reminded by: Divacky Roman
MFC after: 1 week

17 years agochange ic_modecaps to a bit vector and use setbit, et. al.
sam [Sun, 11 Mar 2007 22:37:32 +0000 (22:37 +0000)]
change ic_modecaps to a bit vector and use setbit, et. al.

17 years agoThis commit was generated by cvs2svn to compensate for changes in r167465,
mp [Sun, 11 Mar 2007 22:33:41 +0000 (22:33 +0000)]
This commit was generated by cvs2svn to compensate for changes in r167465,
which included commits to RCS files with non-trunk default branches.

17 years agoImport of tcsh-6.15.00
mp [Sun, 11 Mar 2007 22:33:41 +0000 (22:33 +0000)]
Import of tcsh-6.15.00

17 years agoMerge the following changes from NetBSD:
stefanf [Sun, 11 Mar 2007 21:47:40 +0000 (21:47 +0000)]
Merge the following changes from NetBSD:

history.c 1.32:
  # Fix memory leak found by valgrind (Julien Torres)
map.c 1.24:
  # fix debugging printf format.
read.c 1.40:
  # Fix bug with multiple pending el_pushes. Reported by Julien Torres.
tty.c 1.24:
  # Coverity CID 1216: Prevent negative index use.

MFC after: 3 weeks

17 years agoAdd dynamic acceleration to moused(8). This introduces a '-A' flag to control
philip [Sun, 11 Mar 2007 20:02:26 +0000 (20:02 +0000)]
Add dynamic acceleration to moused(8).  This introduces a '-A' flag to control
the acceleration algorithm.  It can be used together with the '-a' flag for
regular acceleration.

PR: bin/110003
Submitted by: Oliver Fromme <olli -at- lurza.secnetix.de>
MFC after: 1 week

17 years agoClarify the test comments in test-basic.sh. Have config.sh do a
kientzle [Sun, 11 Mar 2007 19:33:45 +0000 (19:33 +0000)]
Clarify the test comments in test-basic.sh.  Have config.sh do a
better job searching for the bsdtar binary to test and the gtar binary
to use for inter-operability testing.  It should now find the built
(but not installed) binary if there is one, then search for an
installed binary in a number of standard locations.

17 years agoMerge the following changes from NetBSD:
stefanf [Sun, 11 Mar 2007 18:30:22 +0000 (18:30 +0000)]
Merge the following changes from NetBSD:

chared.h 1.17, common.c 1.19, emacs.c 1.21, key.c 1.18, key.h 1.9, map.c 1.23,
term.c 1.42, term.h 1.17, vi.c 1.25:
  # Print the actual eofc, instead of ^D\b\b.
  # Change internal character decoding to prevent buffer oveflows.
key.c 1.19, key.h 1.10:
  # move declaration to header file.
term.c 1.43:
  # Coverity CID 806: Prevent NULL deref
term.c 1.44:
  # Coverity CID 1668: Plug memory leak.
term.c 1.45:
  # Fix compilation.

MFC after: 3 weeks

17 years agoDon't try to apply connection-per-ip rate limiting to unix domain
dwmalone [Sun, 11 Mar 2007 16:30:49 +0000 (16:30 +0000)]
Don't try to apply connection-per-ip rate limiting to unix domain
sockets.  Instead of rejecting all unix domain connections when the
-C flag is given, allow them instead. Aragon tested an earlier
version of the patch.

PR: 109315
MFC after: 2 weeks
Tested-by: Aragon Gouveia <aragon@phat.za.net>

17 years agoFix analog CD connectivity. HP decided to screw pinconfig settings in
ariff [Sun, 11 Mar 2007 15:26:49 +0000 (15:26 +0000)]
Fix analog CD connectivity. HP decided to screw pinconfig settings in
their latest Compaq V3000 BIOS (revision F.22). As a result, analog CD
connectivity is gone to the oblivion. Even if they decide to fix it in
future revisions, the damage has been done.

17 years agomy(4) doesn't need miibus(4).
brueffer [Sun, 11 Mar 2007 15:20:04 +0000 (15:20 +0000)]
my(4) doesn't need miibus(4).

Approved by: rwatson (mentor)
Obtained from: DragonFly
MFC after: 1 week

17 years agoAdd ncursesw, libelf, and libalias profiling libraries
rafan [Sun, 11 Mar 2007 13:37:09 +0000 (13:37 +0000)]
Add ncursesw, libelf, and libalias profiling libraries

Approved by: delphij (mentor)

17 years agoSpell .Xr without a '/'.
brooks [Sun, 11 Mar 2007 10:48:34 +0000 (10:48 +0000)]
Spell .Xr without a '/'.

Pointy hat: brooks

17 years agobsdtar 2.0.23:
kientzle [Sun, 11 Mar 2007 10:36:42 +0000 (10:36 +0000)]
bsdtar 2.0.23:
   * New test scripts exercise some basic functionality
   * Most header inclusions are now protected (portability)
   * read.c now relies on security checks in libarchive instead
     of trying to do its own (optimization)
   * -p now enabled by default for root, add --no-same-permissions
     to disable it
   * Comments, minor style fixes.

17 years agoLibarchive 2.0.23:
kientzle [Sun, 11 Mar 2007 10:29:52 +0000 (10:29 +0000)]
Libarchive 2.0.23:
   * The ACL formatter was mis-formatting entries which had a
     user/group ID but no name.  Make the parser tolerant of
     these, so that old archives can be correctly restored;
     fix the formatter to generate correct entries.
   * Fix overwrite detection by introducing a new "FAILED" return
     code that indicates the current entry cannot be continued
     but the archive as a whole is still sound.
   * Header cleanup:  Remove some unused headers, add some that
     are required with new Linux systems.

17 years agoFix a couple markup problems in the previous commit and bump the
brooks [Sun, 11 Mar 2007 10:24:37 +0000 (10:24 +0000)]
Fix a couple markup problems in the previous commit and bump the
document date.

Reported by: ru

17 years agoMerge changes to the NetBSD copyright (advertising clause removal).
stefanf [Sun, 11 Mar 2007 08:41:01 +0000 (08:41 +0000)]
Merge changes to the NetBSD copyright (advertising clause removal).

17 years agoUpdate to reflect correct usage:
sam [Sun, 11 Mar 2007 07:42:02 +0000 (07:42 +0000)]
Update to reflect correct usage:
o leave IEEE80211_RADIOTAP_HDRLEN for portability to other systems but
  correct comment about radiotap headers being padded to 64-bytes
  (hasn't been true for many years)
o remove reference to IEEE80211_RADIOTAP_FCS; it was never used, instead
  the flags are marked with IEEE80211_RADIOTAP_F_FCS to indicate whether
  or not FCS is present

Might be better to just remove IEEE80211_RADIOTAP_HDRLEN so drivers
don't bogusly pad.

MFC after: 2 weeks

17 years agochange ieee80211_fix_rate to take a rate set instead of using
sam [Sun, 11 Mar 2007 07:22:21 +0000 (07:22 +0000)]
change ieee80211_fix_rate to take a rate set instead of using
ni_rates; this lets us re-use the code to check 11n HT rates

MFC after: 2 weeks

17 years agoexpose IEEE80211_DISCARD, IEEE80211_DISCARD_IE, and IEEE80211_DISCARD_MAC
sam [Sun, 11 Mar 2007 07:15:38 +0000 (07:15 +0000)]
expose IEEE80211_DISCARD, IEEE80211_DISCARD_IE, and IEEE80211_DISCARD_MAC
so they can be used within net80211 but outside ieee80211_input.c

MFC after: 2 weeks

17 years agoimprove debug msg for ie's that are too short
sam [Sun, 11 Mar 2007 07:08:04 +0000 (07:08 +0000)]
improve debug msg for ie's that are too short

MFC after: 2 weeks

17 years agokeep tx/rx seq #'s for non-QoS traffic separate from QoS; stations
sam [Sun, 11 Mar 2007 07:06:08 +0000 (07:06 +0000)]
keep tx/rx seq #'s for non-QoS traffic separate from QoS; stations
aren't supposed mix traffic but if they did frames might be mis-handled

Obtained from: Atheros
MFC after: 2 weeks

17 years agosplit check for rate set mismatch on assoc req away from check
sam [Sun, 11 Mar 2007 06:55:59 +0000 (06:55 +0000)]
split check for rate set mismatch on assoc req away from check
for pure 11g mode so we can give meaningful diagnostic msgs

MFC after: 2 weeks

17 years agoAllow background_fsck_delay to be set to a negative value which delays
brooks [Sun, 11 Mar 2007 06:53:07 +0000 (06:53 +0000)]
Allow background_fsck_delay to be set to a negative value which delays
the background fsck indefinitely.  This allows the administrator to run
it at a convenient time.  To support running it from cron, the
forcestart argument now causes the fsck to start with no delay and all
output to be suppressed.

17 years agodisallow re-associate with a slot time mismatch
sam [Sun, 11 Mar 2007 06:52:22 +0000 (06:52 +0000)]
disallow re-associate with a slot time mismatch

Obtained from: Atheros
MFC after: 2 weeks

17 years agomove __inline before type to appease gcc 4.x
sam [Sun, 11 Mar 2007 06:44:51 +0000 (06:44 +0000)]
move __inline before type to appease gcc 4.x

Obtained from: netbsd

17 years agoAdd myself.
bushman [Sun, 11 Mar 2007 06:44:37 +0000 (06:44 +0000)]
Add myself.

Approved by: brooks (mentor)

17 years agowhite space diff reduction
sam [Sun, 11 Mar 2007 06:38:26 +0000 (06:38 +0000)]
white space diff reduction

17 years agoadd IEEE80211_KEY_UNDEFINED and use it instead of local defs
sam [Sun, 11 Mar 2007 06:36:10 +0000 (06:36 +0000)]
add IEEE80211_KEY_UNDEFINED and use it instead of local defs

Obtained from: netbsd

17 years agodiscard deauth+disassoc frames that are not destined for us; these
sam [Sun, 11 Mar 2007 06:35:27 +0000 (06:35 +0000)]
discard deauth+disassoc frames that are not destined for us; these
can be received when the interface is in promisc mode

Reviewed by: sephe
Obtained from: netbsd

17 years agochange ieee80211_mhz2ieee to use the PSB mapping when the frequency
sam [Sun, 11 Mar 2007 06:24:26 +0000 (06:24 +0000)]
change ieee80211_mhz2ieee to use the PSB mapping when the frequency
falls in the proper place, not when we're handed a 1/2 or 1/4-rate channel

MFC after: 2 weeks

17 years agoPush down the implementation of PCPU_LAZY_INC() into the machine-dependent
alc [Sun, 11 Mar 2007 05:54:29 +0000 (05:54 +0000)]
Push down the implementation of PCPU_LAZY_INC() into the machine-dependent
header file.  Reimplement PCPU_LAZY_INC() on amd64 and i386 making it
atomic with respect to interrupts.

Reviewed by: bde, jhb

17 years agofeedback from RELENG_5 port
mjacob [Sun, 11 Mar 2007 01:59:44 +0000 (01:59 +0000)]
feedback from RELENG_5 port

17 years agoCompletely eliminate "avail_start". It serves no useful purpose.
alc [Sat, 10 Mar 2007 20:26:43 +0000 (20:26 +0000)]
Completely eliminate "avail_start".  It serves no useful purpose.

17 years agoUppercase FreeBSD, sigh.
mbr [Sat, 10 Mar 2007 18:31:33 +0000 (18:31 +0000)]
Uppercase FreeBSD, sigh.

17 years agoAdd myself
mbr [Sat, 10 Mar 2007 18:30:12 +0000 (18:30 +0000)]
Add myself

17 years agoPrepend ${_chroot} to a pathname in a consistent and robust way:
yar [Sat, 10 Mar 2007 13:37:44 +0000 (13:37 +0000)]
Prepend ${_chroot} to a pathname in a consistent and robust way:
insert a slash between ${_chroot} and the pathname if and only if
${_chroot} is set to a non-empty string.  Now the pathname is very
likely to be absolute, but we shouldn't take that for granted.

17 years agoUppercase FreeBSD.
brueffer [Sat, 10 Mar 2007 12:45:46 +0000 (12:45 +0000)]
Uppercase FreeBSD.

17 years agoDocument Silicon Image SiI 3124 and SiI 3132 support.
brueffer [Sat, 10 Mar 2007 12:44:22 +0000 (12:44 +0000)]
Document Silicon Image SiI 3124 and SiI 3132 support.

17 years agoBump the original revision of c89(1).
ru [Sat, 10 Mar 2007 07:11:20 +0000 (07:11 +0000)]
Bump the original revision of c89(1).

17 years agoReduce diffs with c99(1).
ru [Sat, 10 Mar 2007 07:10:01 +0000 (07:10 +0000)]
Reduce diffs with c99(1).

17 years agoAdd myself.
mckusick [Sat, 10 Mar 2007 05:56:05 +0000 (05:56 +0000)]
Add myself.

17 years agoAdd new ST201 PCI id.
yongari [Sat, 10 Mar 2007 03:10:34 +0000 (03:10 +0000)]
Add new ST201 PCI id.

PR: kern/108150

17 years agoNew release notes: ULE 2.0, UNIX domain locking changes, driver
bmah [Sat, 10 Mar 2007 02:46:58 +0000 (02:46 +0000)]
New release notes: ULE 2.0, UNIX domain locking changes, driver
ABI/API change for interrupt filtering, zs driver gone, npe(4) (+MFC),
bpf(4) directions, IPv6 multicast forwarding loadable, setfacl(1) -h.

17 years agoFix some stupid copyright mistakes that have been there for quite some time.
mjacob [Sat, 10 Mar 2007 02:39:54 +0000 (02:39 +0000)]
Fix some stupid copyright mistakes that have been there for quite some time.

17 years agooops committed the wrong patch.
julian [Sat, 10 Mar 2007 01:02:40 +0000 (01:02 +0000)]
oops committed the wrong patch.
try this one..

17 years agoNew release notes: 802.11 support for 900 MHz cards and quarter /
bmah [Fri, 9 Mar 2007 23:49:48 +0000 (23:49 +0000)]
New release notes:  802.11 support for 900 MHz cards and quarter /
half channel support (+MFC), ng_deflate(4) (+MFC), ng_pred1(4) (+MFC),
wpa_passphrase(8) (+MFC).

Updated release notes:  ath(4) and HAL 0.9.20.3 (+MFC).

Other MFCs noted:  if_bridge(4) RSTP.

17 years agoUse sx_sleep() in the main loop of the accounting kthread.
jhb [Fri, 9 Mar 2007 23:29:31 +0000 (23:29 +0000)]
Use sx_sleep() in the main loop of the accounting kthread.

17 years agopoint-to-point address should come from ifa_dstaddr
sam [Fri, 9 Mar 2007 23:28:40 +0000 (23:28 +0000)]
point-to-point address should come from ifa_dstaddr

17 years agoAllow threads to atomically release rw and sx locks while waiting for an
jhb [Fri, 9 Mar 2007 22:41:01 +0000 (22:41 +0000)]
Allow threads to atomically release rw and sx locks while waiting for an
event.  Locking primitives that support this (mtx, rw, and sx) now each
include their own foo_sleep() routine.
- Rename msleep() to _sleep() and change it's 'struct mtx' object to a
  'struct lock_object' pointer.  _sleep() uses the recently added
  lc_unlock() and lc_lock() function pointers for the lock class of the
  specified lock to release the lock while the thread is suspended.
- Add wrappers around _sleep() for mutexes (mtx_sleep()), rw locks
  (rw_sleep()), and sx locks (sx_sleep()).  msleep() still exists and
  is now identical to mtx_sleep(), but it is deprecated.
- Rename SLEEPQ_MSLEEP to SLEEPQ_SLEEP.
- Rewrite much of sleep.9 to not be msleep(9) centric.
- Flesh out the 'RETURN VALUES' section in sleep.9 and add an 'ERRORS'
  section.
- Add __nonnull(1) to _sleep() and msleep_spin() so that the compiler will
  warn if you try to pass a NULL wait channel.  The functions already have
  a KASSERT to that effect.

17 years agoAdd support for the Silicon Image SiI 3124 and SiI 3132.
sos [Fri, 9 Mar 2007 22:23:39 +0000 (22:23 +0000)]
Add support for the Silicon Image SiI 3124 and SiI 3132.

HW donated by: Adriaan de Groot

17 years agong_apply_item should be void. It is called from the interrupt source or
julian [Fri, 9 Mar 2007 21:04:50 +0000 (21:04 +0000)]
ng_apply_item should be void. It is called from the interrupt source or
from whoever has dequeued the item from the queue. Generally they have
no interest in the result, and even if it is called by the queuer, it
should still pretend that it was queued. The queuer should be assuming
that the call was queued and giving them the false confidence that they
are getting status leads to hard to find bugs.

Make it a void and remove all the code that tried to return status through it.

17 years agoSort (once again) by month/day/year/login.
ru [Fri, 9 Mar 2007 20:31:46 +0000 (20:31 +0000)]
Sort (once again) by month/day/year/login.

17 years agos/filesystem/file system/ in six places for consistency.
bmah [Fri, 9 Mar 2007 20:16:49 +0000 (20:16 +0000)]
s/filesystem/file system/ in six places for consistency.

17 years agoAdd my record.
jkois [Fri, 9 Mar 2007 20:14:59 +0000 (20:14 +0000)]
Add my record.

Discussed with and approved by: remko@

17 years agoNew release notes: GEOM_MULTIPATH, unionfs reimplementation (+MFC),
bmah [Fri, 9 Mar 2007 20:14:07 +0000 (20:14 +0000)]
New release notes:  GEOM_MULTIPATH, unionfs reimplementation (+MFC),
addr2ascii()/ascii2addr() removal, dump(8)/restore(8) extended
attributes, objformat(1) gone, vnconfig(8) gone, ncurses wide
character support.

17 years agoWordsmithery.
bms [Fri, 9 Mar 2007 19:43:42 +0000 (19:43 +0000)]
Wordsmithery.

Pointed out by: ru

17 years agoChange the passing of callbacks to a struct in case this needs to be extended in...
thompsa [Fri, 9 Mar 2007 19:34:55 +0000 (19:34 +0000)]
Change the passing of callbacks to a struct in case this needs to be extended in the future.

17 years agoMore SATA phy/reset cleanup and simplification.
sos [Fri, 9 Mar 2007 18:45:45 +0000 (18:45 +0000)]
More SATA phy/reset cleanup and simplification.
Fix a long standing bogon in the sii_reset code.

17 years agoDocument SLEEPQ_PAUSE.
jhb [Fri, 9 Mar 2007 18:06:36 +0000 (18:06 +0000)]
Document SLEEPQ_PAUSE.

17 years agoOof, it is not my day. Unswap the descriptions of EINTR and ERESTART.
jhb [Fri, 9 Mar 2007 17:56:05 +0000 (17:56 +0000)]
Oof, it is not my day.  Unswap the descriptions of EINTR and ERESTART.

17 years agoFix a typo in the last that claimed that only masked signals were caught.
jhb [Fri, 9 Mar 2007 17:53:16 +0000 (17:53 +0000)]
Fix a typo in the last that claimed that only masked signals were caught.

17 years agoCorrect descriptions of EINTR and ERESTART errors. Masked signals don't
jhb [Fri, 9 Mar 2007 17:52:10 +0000 (17:52 +0000)]
Correct descriptions of EINTR and ERESTART errors.  Masked signals don't
interrupt sleeps.  Rather, unmasked signals interrupt restarts and can
either interrupt the system call by having it return EINTR in userland or
force the system call to be restarted.

17 years ago- Document cv_wait_unlock(9).
jhb [Fri, 9 Mar 2007 16:59:27 +0000 (16:59 +0000)]
- Document cv_wait_unlock(9).
- Don't claim that the mutex is atomically reacquired when a cv_wait
  routine returns.  There's nothing atomic or magical about the lock
  reacquire.  The only magic is that we atomically drop the lock by
  placing the thread on the sleep queue before dropping the lock.

17 years agoPlace mtx_destroy() and rw_destroy() right after mtx_init() and rw_init()
jhb [Fri, 9 Mar 2007 16:54:03 +0000 (16:54 +0000)]
Place mtx_destroy() and rw_destroy() right after mtx_init() and rw_init()
to match sx(9) and condvar(9).

17 years ago- Sort functions in the order that rwlock(9) and mutex(9) use.
jhb [Fri, 9 Mar 2007 16:52:26 +0000 (16:52 +0000)]
- Sort functions in the order that rwlock(9) and mutex(9) use.
- Markup sx_unlock() as a function rather than saying it is a macro.
  The macro part is an implementation detail, and all the other sx_*lock()
  functions are actually macros, too.
- Use the same style as rwlock(9) and mutex(9) to markup sx_assert() and
  SX_SYSINIT() with respect to headers and kernel options.
- Add a missing MLINK.

17 years ago- Sort rwlock functions similar to the order in mutex(9).
jhb [Fri, 9 Mar 2007 16:44:58 +0000 (16:44 +0000)]
- Sort rwlock functions similar to the order in mutex(9).
- Add a missing MLINK.

17 years ago! is a generic info message
xride [Fri, 9 Mar 2007 16:29:15 +0000 (16:29 +0000)]
! is a generic info message

Reviewed by: harti
Approved by: harti
MFC after: 2 weeks

17 years agoAdd two new function pointers 'lc_lock' and 'lc_unlock' to lock classes.
jhb [Fri, 9 Mar 2007 16:27:11 +0000 (16:27 +0000)]
Add two new function pointers 'lc_lock' and 'lc_unlock' to lock classes.
These functions are intended to be used to drop a lock and then reacquire
it when doing an sleep such as msleep(9).  Both functions accept a
'struct lock_object *' as their first parameter.  The 'lc_unlock' function
returns an integer that is then passed as the second paramter to the
subsequent 'lc_lock' function.  This can be used to communicate state.
For example, sx locks and rwlocks use this to indicate if the lock was
share/read locked vs exclusive/write locked.

Currently, spin mutexes and lockmgr locks do not provide working lc_lock
and lc_unlock functions.

17 years agoKERN_PROC_ALL produces a kinfo_proc for each thread in a process, which
emaste [Fri, 9 Mar 2007 16:21:40 +0000 (16:21 +0000)]
KERN_PROC_ALL produces a kinfo_proc for each thread in a process, which
caused fstat to produce duplicated output for threaded processes.  Instead
use KERN_PROC_PROC to get just one kinfo_proc per process.

MFC After: 2 weeks

17 years agoUse C99-style struct member initialization for lock classes.
jhb [Fri, 9 Mar 2007 16:19:34 +0000 (16:19 +0000)]
Use C99-style struct member initialization for lock classes.

17 years agoUse C99-style struct member initialization for lock classes.
jhb [Fri, 9 Mar 2007 16:04:44 +0000 (16:04 +0000)]
Use C99-style struct member initialization for lock classes.

17 years agoDefer calling lapic_init() until we've completed the 'MPTable: <...>'
jhb [Fri, 9 Mar 2007 15:49:57 +0000 (15:49 +0000)]
Defer calling lapic_init() until we've completed the 'MPTable: <...>'
printf.  Otherwise, printfs inside of lapic_init() (such as during a
verbose boot) can uglify the output.

17 years agoFix markup.
ru [Fri, 9 Mar 2007 14:36:18 +0000 (14:36 +0000)]
Fix markup.

17 years agoSpell "id" as "ID".
ru [Fri, 9 Mar 2007 12:45:00 +0000 (12:45 +0000)]
Spell "id" as "ID".

17 years agoBump __FreeBSD_version for ncurses wide character support
rafan [Fri, 9 Mar 2007 12:12:55 +0000 (12:12 +0000)]
Bump __FreeBSD_version for ncurses wide character support

Approved by: delphij (mentor, implicit)

17 years agoEnable ncurses wide character support
rafan [Fri, 9 Mar 2007 12:11:58 +0000 (12:11 +0000)]
Enable ncurses wide character support

Approved by: delphij (mentor)
Tested by: kris on pointyhat (early version), current@

17 years ago- style.Makefile(9) fix
rafan [Fri, 9 Mar 2007 09:54:07 +0000 (09:54 +0000)]
- style.Makefile(9) fix
  - first line is $FreeBSD$
  - Reorder special variables: DPADD, LPADD, CFLAGS
  - Use = instead of += for variables that are initially empty
  - Use space instead of tab after :
  - Use one tab after =
- Use .SUFFIXES for section 3 manual page which simplifies Makefile a lot
- Use SHAREDIR instead of /usr/share
- Remove SRCDIR in INCS since we set .PATH properly
- Use plural in variable name when it stands for more that one source file

Reviewed by: ru
Approved by: delphij (mentor)

17 years agoInitialize *pfh to NULL to quiet a gcc warning, not part of my original
trhodes [Fri, 9 Mar 2007 09:40:23 +0000 (09:40 +0000)]
Initialize *pfh to NULL to quiet a gcc warning, not part of my original
commit because it was not part of the new code.

17 years agoAdd support for dropping privileges to a specified user and/or group.
trhodes [Fri, 9 Mar 2007 09:33:19 +0000 (09:33 +0000)]
Add support for dropping privileges to a specified user and/or group.

PR: 108523
Submitted by: Dmitri Alenitchev <dmitri@dworlds.ru> (original version)
Reviewed by: mpp (first reply to PR)

17 years agoFix long standing multi playback/recording issues, caused by
ariff [Fri, 9 Mar 2007 05:35:20 +0000 (05:35 +0000)]
Fix long standing multi playback/recording issues, caused by
excessive interrupt clock timer reset, screwing interrupt generation
for already active channels. Track moving DMA pointer and call buffer
interrupt on each blocksize boundary.

PR: kern/109791
MFC after: 3 days

17 years agoMinor simplification.
pjd [Fri, 9 Mar 2007 05:22:10 +0000 (05:22 +0000)]
Minor simplification.

17 years agoBack out a chance to nfs_timer() that inadvertantly crept in the last checkin :(
mohans [Fri, 9 Mar 2007 04:07:54 +0000 (04:07 +0000)]
Back out a chance to nfs_timer() that inadvertantly crept in the last checkin :(

17 years agoOver NFS, an open() call could result in multiple over-the-wire
mohans [Fri, 9 Mar 2007 04:02:38 +0000 (04:02 +0000)]
Over NFS, an open() call could result in multiple over-the-wire
GETATTRs being generated - one from lookup()/namei() and the other
from nfs_open() (for cto consistency). This change eliminates the
GETATTR in nfs_open() if an otw GETATTR was done from the namei()
path. Instead of extending the vop interface, we timestamp each attr
load, and use this to detect whether a GETATTR was done from namei()
for this syscall. Introduces a thread-local variable that counts the
syscalls made by the thread and uses <pid, tid, thread syscalls> as
the attrload timestamp. Thanks to jhb@ and peter@ for a discussion on
thread state that could be used as the timestamp with minimal overhead.