]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoDon't be noisy in case of shared interrupts.
phk [Sat, 11 Mar 2006 15:39:22 +0000 (15:39 +0000)]
Don't be noisy in case of shared interrupts.

18 years agoadd dependency with firmware module.
damien [Sat, 11 Mar 2006 13:30:21 +0000 (13:30 +0000)]
add dependency with firmware module.
pointed out by Martin Wilke.

18 years agoGo over calcru and friends once more.
phk [Sat, 11 Mar 2006 10:48:19 +0000 (10:48 +0000)]
Go over calcru and friends once more.

Reintroduce the monotonicity for the normal case and make the two
special cases behave in what is belived to be the most sensible fasion.

18 years agoRemove last traces of disk_enumerate()
phk [Sat, 11 Mar 2006 10:24:50 +0000 (10:24 +0000)]
Remove last traces of disk_enumerate()

18 years agoFix "make distribution" when just NO_CRYPT is defined.
ru [Sat, 11 Mar 2006 08:51:16 +0000 (08:51 +0000)]
Fix "make distribution" when just NO_CRYPT is defined.

PR: misc/94309
Submitted by: Matthias Lederhofer
MFC after: 3 days

18 years agoRemove debug flag.
sos [Sat, 11 Mar 2006 08:42:08 +0000 (08:42 +0000)]
Remove debug flag.

Spotted by: ru@

18 years agoFix build.
ru [Sat, 11 Mar 2006 08:36:15 +0000 (08:36 +0000)]
Fix build.

18 years agoBlock secondary writes while expunging active unlinked files.
tegge [Sat, 11 Mar 2006 01:08:37 +0000 (01:08 +0000)]
Block secondary writes while expunging active unlinked files.

Fix detection of active unlinked files by checking VI_OWEINACT and
VI_DOINGINACT in addition to v_usecount.

Defer inactive handling for unlinked files if the file system is mostly
suspended (secondary writes being blocked).

Perform deferred inactive handling after the file system is resumed.

18 years agoFix multicast support for cs89x0 chips. Just setting the RX_MULTCAST_ACCEPT
philip [Fri, 10 Mar 2006 23:50:53 +0000 (23:50 +0000)]
Fix multicast support for cs89x0 chips.  Just setting the RX_MULTCAST_ACCEPT
flag isn't enough - the filter needs to be set up too, or no multicast frames
are accepted.

Sponsored by: Philips Industrial Applications (indirectly)
MFC after: 3 days

18 years agoMicro-optimize invalid UUID check.
jkim [Fri, 10 Mar 2006 23:33:27 +0000 (23:33 +0000)]
Micro-optimize invalid UUID check.

18 years agoo acknowledge interrupts early in the interrupt handler
damien [Fri, 10 Mar 2006 20:35:09 +0000 (20:35 +0000)]
o acknowledge interrupts early in the interrupt handler
o stop processing interrupts after a firmware fatal error or a radio kill
o clarify the possible values for the 'antenna' sysctl.
o by default, let the firmware do antenna diversity.
  the  firmware will periodically switch to another antenna to evaluate the
  signal quality.

18 years agomake use of the firmware(9) subsystem.
damien [Fri, 10 Mar 2006 19:59:09 +0000 (19:59 +0000)]
make use of the firmware(9) subsystem.
use intel's firmware version 3.0 layout.

18 years agoAdd atausb module
sos [Fri, 10 Mar 2006 19:15:20 +0000 (19:15 +0000)]
Add atausb module

18 years agoAdd "device atausb"
sos [Fri, 10 Mar 2006 19:10:05 +0000 (19:10 +0000)]
Add "device atausb"

18 years agoModule that adds support for using USB storage devices through the
sos [Fri, 10 Mar 2006 19:08:52 +0000 (19:08 +0000)]
Module that adds support for using USB storage devices through the
ATA framework. Mainly written to be able to use USB Flash keys.
This is work in progress so use with care :)
Doesn't need CAM and cannot coexist with umass.c

18 years agofix the locking.
damien [Fri, 10 Mar 2006 18:55:30 +0000 (18:55 +0000)]
fix the locking.

18 years agoAdd note about 'audit' group.
imp [Fri, 10 Mar 2006 18:40:31 +0000 (18:40 +0000)]
Add note about 'audit' group.
Add note about the '_dhcp' user.

While one is expected to run mergemaster -p before installworld, make
a note of the points where this will actively fail due to the addition of
users/groups.

18 years agoUse .Cm for all key words, not just the ones I added. More in line with
trhodes [Fri, 10 Mar 2006 12:37:19 +0000 (12:37 +0000)]
Use .Cm for all key words, not just the ones I added.  More in line with
mdoc(7) according to ru.

Requested by: ru

18 years agowrite the right unit # on verbose output.
sos [Fri, 10 Mar 2006 12:20:53 +0000 (12:20 +0000)]
write the right unit # on verbose output.

18 years agoDisable an SMP detection on pc98.
nyan [Fri, 10 Mar 2006 12:18:41 +0000 (12:18 +0000)]
Disable an SMP detection on pc98.

18 years agoHopefully fix all nearby style bugs that Bruce has mentioned.
ru [Fri, 10 Mar 2006 11:07:39 +0000 (11:07 +0000)]
Hopefully fix all nearby style bugs that Bruce has mentioned.

18 years ago- Fix variable assignment to be portable.
ru [Fri, 10 Mar 2006 11:00:11 +0000 (11:00 +0000)]
- Fix variable assignment to be portable.
- "Line up" continuations.

Submitted by: bde

18 years agoAdd entries for new pthread stubs.
davidxu [Fri, 10 Mar 2006 09:58:18 +0000 (09:58 +0000)]
Add entries for new pthread stubs.

18 years agoBlock all signals in helper threads except those should not be blocked.
davidxu [Fri, 10 Mar 2006 08:01:23 +0000 (08:01 +0000)]
Block all signals in helper threads except those should not be blocked.

18 years agoRemove including of libc sources. All the required functions
deischen [Fri, 10 Mar 2006 07:58:02 +0000 (07:58 +0000)]
Remove including of libc sources.  All the required functions
are exported by libc with prototypes in our standard headers.
I guess at one time this was necessary, but not any longer.

18 years agoWhen inserting a new component md_provsize metadata field wasn't set, which
pjd [Fri, 10 Mar 2006 07:41:31 +0000 (07:41 +0000)]
When inserting a new component md_provsize metadata field wasn't set, which
means that old problem was triggered (when two providers end at the same
offset, eg. ad0 and ad0s1 and the wrong was is picked up by gmirror/graid3).

Reported by: Michal Suszko <dry@dry.pl>
MFC after: 3 days

18 years agofix switching between agressive and non-agressive wmm modes
sam [Fri, 10 Mar 2006 06:18:03 +0000 (06:18 +0000)]
fix switching between agressive and non-agressive wmm modes

Obtained from: atheros
MFC after: 3 days

18 years agoIt is not necessary to read %gs twice.
davidxu [Fri, 10 Mar 2006 05:55:26 +0000 (05:55 +0000)]
It is not necessary to read %gs twice.

18 years agoFix stack offset to allow gcc's stack aligment code to work correctly.
davidxu [Fri, 10 Mar 2006 02:54:45 +0000 (02:54 +0000)]
Fix stack offset to allow gcc's stack aligment code to work correctly.

MFC after: 3 days

18 years agoRemove unneeded (and broken) usage of MNT_REF()/MNT_REL().
tegge [Fri, 10 Mar 2006 02:31:12 +0000 (02:31 +0000)]
Remove unneeded (and broken) usage of MNT_REF()/MNT_REL().

18 years agoMFp4: Forget the asm inlined version of in_cksum_hdr(). It doesn't work if
cognet [Thu, 9 Mar 2006 23:33:59 +0000 (23:33 +0000)]
MFp4: Forget the asm inlined version of in_cksum_hdr(). It doesn't work if
the pointer is unaligned, and it just doesn't worth it.

18 years ago- Implement serial numbers, UUID, and asset tag (turned off by default).
jkim [Thu, 9 Mar 2006 22:49:44 +0000 (22:49 +0000)]
- Implement serial numbers, UUID, and asset tag (turned off by default).
Use 'BOOT_SENSITIVE_INFO=YES' variable to turn them on.
- Use 'uint*_t' instead of 'u_int*_t', correct compilation warnings, and
update copyright while I am here.

18 years agoImplement printf 'X' conversion for both libstand and kernel.
jkim [Thu, 9 Mar 2006 22:37:34 +0000 (22:37 +0000)]
Implement printf 'X' conversion for both libstand and kernel.

18 years agomissed a case that needs WITH_SMP
sam [Thu, 9 Mar 2006 21:54:32 +0000 (21:54 +0000)]
missed a case that needs WITH_SMP

Noticed by: kensmith

18 years agoRever the previous changes. It turns out that it perfectly correct
gad [Thu, 9 Mar 2006 19:52:44 +0000 (19:52 +0000)]
Rever the previous changes.  It turns out that it perfectly correct
for a makefile to set 'NO_MAN=' when the makefile is for a program
that will not create a man page.

Based on reaction from: ru bde

18 years agoFix this to what I've actually submitted.
ru [Thu, 9 Mar 2006 17:55:08 +0000 (17:55 +0000)]
Fix this to what I've actually submitted.

18 years agoRevise the names of modules in the recent note on tdfx_linux.
yar [Thu, 9 Mar 2006 17:50:01 +0000 (17:50 +0000)]
Revise the names of modules in the recent note on tdfx_linux.

18 years agohandle KERNELS as before
sam [Thu, 9 Mar 2006 17:11:17 +0000 (17:11 +0000)]
handle KERNELS as before

Submitted by: ru, andrew

18 years agoFlip the switch and don't route interrupts to hyperthreads in a HT system.
jhb [Thu, 9 Mar 2006 16:38:52 +0000 (16:38 +0000)]
Flip the switch and don't route interrupts to hyperthreads in a HT system.
In at least one benchmark this showed around a 20% performance increase.
If other workloads do benefit from having hyperthreads service interrupts,
we can always make this a loader tunable.

MFC after: 3 days
Tested by: ps

18 years agoDon't dereference an uninitialized pointer.
imp [Thu, 9 Mar 2006 16:28:24 +0000 (16:28 +0000)]
Don't dereference an uninitialized pointer.

18 years agoRevert last change, per ru@'s objection. I misunderstood consensus
imp [Thu, 9 Mar 2006 16:24:58 +0000 (16:24 +0000)]
Revert last change, per ru@'s objection.  I misunderstood consensus

18 years agoAvoid pulling in the whole <net/pfvar.h> by opaquely declaring
yar [Thu, 9 Mar 2006 15:54:01 +0000 (15:54 +0000)]
Avoid pulling in the whole <net/pfvar.h> by opaquely declaring
the structs pflog_packet() takes pointers to.

Approved by: mlaier
MFC after: 3 days

18 years agoThe `pf' and `pflog' sources do not depend on DEV_PF or DEV_PFLOG,
yar [Thu, 9 Mar 2006 15:48:45 +0000 (15:48 +0000)]
The `pf' and `pflog' sources do not depend on DEV_PF or DEV_PFLOG,
which is normal for own files of a device driver.

DEV_FOO should be used if an unrelated kernel file needs to know of
the `foo' driver's static presence.  Obviously, module source files
should never use DEV_*.

18 years agoRevert to setting vlan and vlandev parametes synchronously, as soon
yar [Thu, 9 Mar 2006 14:58:09 +0000 (14:58 +0000)]
Revert to setting vlan and vlandev parametes synchronously, as soon
as both have been read from the command line.  Still use the callback,
but this time only to verify that both vlan and vlandev have been
found on the command line.

This should allow for control over the relative order of processing
parameters, which is needed to satisfy some caveats of the if_vlan
driver.  E.g., MTU cannot be changed on a vlan interface until it's
attached to its parent.

PR:             bin/94028
Reviewed by:    ru
MFC after:      3 days

18 years agoAdd -F option to not fail on already installed packages.
phk [Thu, 9 Mar 2006 14:49:19 +0000 (14:49 +0000)]
Add -F option to not fail on already installed packages.

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

18 years agoIn addition to -b (suppress builds) add -k (suppress kernel build) and
phk [Thu, 9 Mar 2006 14:43:07 +0000 (14:43 +0000)]
In addition to -b (suppress builds) add -k (suppress kernel build) and
-w (suppress world).

18 years agoDon't acquire a lock before calling vlan_unconfig().
ru [Thu, 9 Mar 2006 14:42:51 +0000 (14:42 +0000)]
Don't acquire a lock before calling vlan_unconfig().
This fixes a panic when doing "ifconfig ... -vlandev".

OK'ed by: glebius

18 years agoAfter revision 1.4 of `src/usr.sbin/bsnmpd/bsnmpd/Makefile' the
keramida [Thu, 9 Mar 2006 12:55:02 +0000 (12:55 +0000)]
After revision 1.4 of `src/usr.sbin/bsnmpd/bsnmpd/Makefile' the
gensnmptree utility is required at "make depend" time, because it
is used to generate `oid.h' for bsnmpd.

This means that all versions of HEAD before 2004/01/23 16:22:49 need
gensnmptree as a bootstrap util when it's not already installed as
part of the base system.  The first __FreeBSD_version that we can
assume this can work is 700014 which happened after the gensnmptree
change (in -r 1.263 of `src/sys/sys/param.h', at 2006/02/17 14:14:15).

For __FreeBSD_version values before 700014 add gensnmptree to the
bootstrap tools, to allow upgrades from versions of FreeBSD before
that date.

Approved by: ru, harti

18 years agoOnly use the channel in the request given to ata_completed if its
sos [Thu, 9 Mar 2006 10:25:35 +0000 (10:25 +0000)]
Only use the channel in the request given to ata_completed if its
still present.

18 years agotypo.
sos [Thu, 9 Mar 2006 10:24:03 +0000 (10:24 +0000)]
typo.

18 years agoOops, forgot newline.
phk [Thu, 9 Mar 2006 09:44:10 +0000 (09:44 +0000)]
Oops, forgot newline.

18 years agoAdd slop to "backwards" cpu accounting messages, 3 usec or 1% whichever
phk [Thu, 9 Mar 2006 09:33:17 +0000 (09:33 +0000)]
Add slop to "backwards" cpu accounting messages, 3 usec or 1% whichever
triggers.

This should eliminate all the trivial messages which result from minor
increases in cpu_tick frequency.

Machines which don't du cpu clock fiddling shouldn't issue "backwards"
messages now.

Laptops and other machines where the initial estimate of cputicks may be
waaaay off will still issue warnings.

18 years agosilence cpu_tick calibration and notice only (under bootverbose)
phk [Thu, 9 Mar 2006 09:30:33 +0000 (09:30 +0000)]
silence cpu_tick calibration and notice only (under bootverbose)
when the frequency increases.

18 years agoIgnore kenv strings which overflow the room we have, rather than pretend
phk [Thu, 9 Mar 2006 09:29:41 +0000 (09:29 +0000)]
Ignore kenv strings which overflow the room we have, rather than pretend
we have room for them.

18 years agoGet rid of all the "long long"/"maxint" casting around in printf's.
sos [Thu, 9 Mar 2006 08:34:44 +0000 (08:34 +0000)]
Get rid of all the "long long"/"maxint" casting around in printf's.
On all our platforms intmax == int64_t so simply using %j to print
int64_t's is safe all over, and doesn't pullute the code.

18 years agoRemove _STOPEVENT call, it is already called in issignal, simplify
davidxu [Thu, 9 Mar 2006 08:31:51 +0000 (08:31 +0000)]
Remove _STOPEVENT call, it is already called in issignal, simplify
code for SIGKILL signal.

18 years agons_name_skip was not prefixed by `__'.
ume [Thu, 9 Mar 2006 03:47:11 +0000 (03:47 +0000)]
ns_name_skip was not prefixed by `__'.

Obtained from: BIND9

18 years agoWhen a process is de-configuring a log file, also stop all of its
jkoshy [Thu, 9 Mar 2006 02:08:12 +0000 (02:08 +0000)]
When a process is de-configuring a log file, also stop all of its
PMCs that require a log file to operate.  This change should fix
PR 90269.

PR: kern/90269
MFC after: 1 week

18 years agoNow that we now spell NO_MAN=xxx MAN=, update the docs.
imp [Thu, 9 Mar 2006 01:33:38 +0000 (01:33 +0000)]
Now that we now spell NO_MAN=xxx MAN=, update the docs.

Submitted by: John Hein

18 years agoConnect librt to buildworld.
davidxu [Thu, 9 Mar 2006 01:03:58 +0000 (01:03 +0000)]
Connect librt to buildworld.

18 years agocorrect ni_txrate when using a fixed rate; fixes current rate reporting
sam [Thu, 9 Mar 2006 00:28:44 +0000 (00:28 +0000)]
correct ni_txrate when using a fixed rate; fixes current rate reporting

MFC after: 3 days

18 years agoRemove mqueue and timer, now they are in librt.
davidxu [Wed, 8 Mar 2006 23:47:04 +0000 (23:47 +0000)]
Remove mqueue and timer, now they are in librt.

18 years agoUse vn_start_secondary_write() and vn_finished_secondary_write() as a
tegge [Wed, 8 Mar 2006 23:43:39 +0000 (23:43 +0000)]
Use vn_start_secondary_write() and vn_finished_secondary_write() as a
replacement for vn_write_suspend_wait() to better account for secondary write
processing.

Close race where secondary writes could be started after ffs_sync() returned
but before the file system was marked as suspended.

Detect if secondary writes or softdep processing occurred during vnode sync
loop in ffs_sync() and retry the loop if needed.

18 years agoRemove mq.c, the POSIX mqueue is implemented in librt.
davidxu [Wed, 8 Mar 2006 23:42:41 +0000 (23:42 +0000)]
Remove mq.c, the POSIX mqueue is implemented in librt.

Suggested by: deischen

18 years agoSwitch these makefiles to use 'MAN=' to indicate they will not generate
gad [Wed, 8 Mar 2006 22:28:12 +0000 (22:28 +0000)]
Switch these makefiles to use 'MAN=' to indicate they will not generate
a man page, instead of 'NO_MAN='.  'NO_MAN=' is something users would
set, not something a makefile should be using.

Based on comments by: des

18 years agoFix overriding jail_list from command line.
fjoe [Wed, 8 Mar 2006 20:40:37 +0000 (20:40 +0000)]
Fix overriding jail_list from command line.

MFC after: 3 days

18 years agoFix exec_map resource leaks.
ups [Wed, 8 Mar 2006 20:21:54 +0000 (20:21 +0000)]
Fix exec_map resource leaks.

Tested by: kris@

18 years agoWhen parsing an RPC request in nfsrv_dorec(), KASSERT that there
simon [Wed, 8 Mar 2006 20:21:15 +0000 (20:21 +0000)]
When parsing an RPC request in nfsrv_dorec(), KASSERT that there
actually is an mbuf to process.  This catches the missing mbuf before it
would otherwise causes a NULL pointer dereference, which could be
triggered by a 0 length RPC record before the check for such records was
added in rev 1.97.

Approved by: cperciva (mentor)

18 years agokernel install script (missed in previous commit)
sam [Wed, 8 Mar 2006 18:03:09 +0000 (18:03 +0000)]
kernel install script (missed in previous commit)

18 years agoRevamp base system packaging of kernels to enable up/smp selection
sam [Wed, 8 Mar 2006 18:02:32 +0000 (18:02 +0000)]
Revamp base system packaging of kernels to enable up/smp selection
at runtime and to support distributing additional kernels:
o remove kernel from the base tarball
o add new kernel tarballs
o build + package both SMP and GENERIC kernels when an <arch>/conf/SMP
  config file is present
o add sysinstall support for multiple kernels
o update sysinstall to probe for the number of cpus on a system
  and auto-select smp/up kernel accordingly
o add a post-kernels install hook to fixup /boot/kernel
o add -ldevinfo to boot crunch for sysinstall's cpu probing logic

Notes:
1. On HEAD this code is not currently used because GENERIC kernels
   include SMP.  This work is mainly intended for RELENG_6 where the
   GENERIC kernel is UP.  If HEAD changes to match then just enable
   WITH_SMP in sysinstall/Makefile.
2. The cpu probing support is done with acpi and MPTable; this means
   some systems will require work for auto-detection to work.
3. The handling of /boot/kernel may need to be revisited; for now
   we rename one kernel at the last moment (SMP if installed, otherwise
   GENERIC).  There are other, possibly better, approaches.

Lots of help from ru, emaste, scottl, and jhb.

18 years agoFix output and exit status when daily_mailq_shorten is set to YES
matteo [Wed, 8 Mar 2006 17:26:53 +0000 (17:26 +0000)]
Fix output and exit status when daily_mailq_shorten is set to YES

PR: conf/93472
MFC after: 3

18 years agoATA_USB will need to hook into the delayed boot identify to have
sos [Wed, 8 Mar 2006 16:39:01 +0000 (16:39 +0000)]
ATA_USB will need to hook into the delayed boot identify to have
interrupts running, so externalize it.

18 years agoUpdate a DB_SET to DB_FUNC I missed yesterday.
jhb [Wed, 8 Mar 2006 15:47:48 +0000 (15:47 +0000)]
Update a DB_SET to DB_FUNC I missed yesterday.

18 years agoExplain why the first partition should start at offset 16.
pjd [Wed, 8 Mar 2006 14:17:26 +0000 (14:17 +0000)]
Explain why the first partition should start at offset 16.

Requested by: hrs

18 years agoFix spelling error
matteo [Wed, 8 Mar 2006 14:10:38 +0000 (14:10 +0000)]
Fix spelling error

MFC after: 1 day

18 years agoProperly handle the case when the packet secondary zone can't allocate
andre [Wed, 8 Mar 2006 14:05:38 +0000 (14:05 +0000)]
Properly handle the case when the packet secondary zone can't allocate
further mbuf clusters to attach to mbufs.

Reported by: kris
Tested by: kris
Sponsored by: TCP/IP Optimization Fundraise 2005
MFC after: 3 days

18 years agoAlways read the TOC in acd_read_toc even if MEDIA_CHANGED wasn't set.
sos [Wed, 8 Mar 2006 10:14:39 +0000 (10:14 +0000)]
Always read the TOC in acd_read_toc even if MEDIA_CHANGED wasn't set.
Since we cannot (reliably) detect when media has been removed before we
try to touch it, this could leave stale TOC's around.

18 years agoBail out of afd_sense if ENODEV.
sos [Wed, 8 Mar 2006 09:21:48 +0000 (09:21 +0000)]
Bail out of afd_sense if ENODEV.

18 years agoCollapse strncpy/strncat/strncat into a single snprintf, as suggested
gad [Wed, 8 Mar 2006 09:15:20 +0000 (09:15 +0000)]
Collapse strncpy/strncat/strncat into a single snprintf, as suggested
by pjd.

MFC after: 3 weeks

18 years agoFix the case where the user specifies an alternate heading for some
gad [Wed, 8 Mar 2006 08:58:44 +0000 (08:58 +0000)]
Fix the case where the user specifies an alternate heading for some
output-format keyword, and the keyword they picked is an alias to
some other keyword.                 E.g.:   ps -o stat=Zustand $$
('stat' is defined as an alias for 'state')

PR: bin/57833
MFC after: 3 weeks

18 years agoAdd some notes how to properly dump kernel onto gmirror provider.
pjd [Wed, 8 Mar 2006 08:50:52 +0000 (08:50 +0000)]
Add some notes how to properly dump kernel onto gmirror provider.

Reviewed and corrected by: brueffer
MFC after: 3 days

18 years agoAllow to dump kernel to gmirror providers.
pjd [Wed, 8 Mar 2006 08:27:33 +0000 (08:27 +0000)]
Allow to dump kernel to gmirror providers.
Some conditions have to be met to make it work properly. This will be
described in the manual page.

MFC after: 3 days

18 years agoRemove leading __ from __(inline|const|signed|volatile). They are
imp [Wed, 8 Mar 2006 06:31:46 +0000 (06:31 +0000)]
Remove leading __ from __(inline|const|signed|volatile).  They are
obsolete.  This should reduce diffs to NetBSD as well.

18 years agoRemove configure.ac.in and the 'distfile' target from
kientzle [Wed, 8 Mar 2006 02:49:12 +0000 (02:49 +0000)]
Remove configure.ac.in and the 'distfile' target from
the FreeBSD makefile.  This is a step towards cleaning
out the non-FreeBSD bits from the FreeBSD source tree.

18 years agoDon't set IN_CHANGE and IN_UPDATE on inodes for potentially suspended
tegge [Wed, 8 Mar 2006 02:14:39 +0000 (02:14 +0000)]
Don't set IN_CHANGE and IN_UPDATE on inodes for potentially suspended
file systems.  This could cause deadlocks when creating snapshots.

Reviewed by: jeff

18 years agoRemove configure.ac.in and reorganize a few other things. This is
kientzle [Wed, 8 Mar 2006 01:56:06 +0000 (01:56 +0000)]
Remove configure.ac.in and reorganize a few other things.  This is
part of a program to remove the non-FreeBSD autoconf/automake build
system for libarchive from the FreeBSD source tree.

18 years agoFix a bug in NFSv3 READDIRPLUS reply processing
cel [Wed, 8 Mar 2006 01:43:01 +0000 (01:43 +0000)]
Fix a bug in NFSv3 READDIRPLUS reply processing

The client's READDIRPLUS logic skips the attributes and
filehandle of the ".." entry.  If the server doesn't send
attributes but does send a filehandle for "..", the
client's logic doesn't account for the extra "value
follows" field that indicates whether the filehandle is
present, causing the remaining entries in the reply
to be ignored.

Sponsored by: Network Appliance, Inc.
Reviewed by: rick, mohans
Approved by: silby
MFC after: 2 weeks

18 years agoIgnore dirty pages owned by "dead" objects.
tegge [Wed, 8 Mar 2006 00:51:00 +0000 (00:51 +0000)]
Ignore dirty pages owned by "dead" objects.

18 years agoSet SNF_SYNC flag for timer, as the timer notification should be
davidxu [Wed, 8 Mar 2006 00:34:54 +0000 (00:34 +0000)]
Set SNF_SYNC flag for timer, as the timer notification should be
serialized.

18 years agoStyle nit.
jhb [Tue, 7 Mar 2006 22:17:26 +0000 (22:17 +0000)]
Style nit.

18 years agoClean up the way we handle auxiliary commands for a given ddb command
jhb [Tue, 7 Mar 2006 22:17:06 +0000 (22:17 +0000)]
Clean up the way we handle auxiliary commands for a given ddb command
table.  Previously, the ddb code knew of each linker set of auxiliary
commands and which explicit command list they were tied to.  These changes
add a simple command_table struct that contains both the static list of
commands and the pointers for any auxiliary linker set of additional
commands.  This also makes it possible for other arbitrary command tables
to be defined in other parts of the kernel w/o having to edit ddb itself.

The DB_SET macro has also been trimmed down to just creating an entry in
a linker set.  A new DB_FUNC macro does what the old DB_SET did which is
to not only add an entry to the linker set but also to include a function
prototype for the function being added.  With these changes, it's now also
possible to create aliases for ddb functions using DB_SET() directly if
desired.

18 years agoUse the Alpha PCC as a cpu ticker for process runtime accounting. This is
jhb [Tue, 7 Mar 2006 22:12:09 +0000 (22:12 +0000)]
Use the Alpha PCC as a cpu ticker for process runtime accounting.  This is
slightly more tricky than on x86 as although the PCC is 64-bits, it is not
a simple 64-bit counter like the TSC.  Instead, the upper 32-bits have
PAL-defined behavior and the lower 32-bits run as a free-running 32-bit
counter.  To handle this, we detect overflows by maintaining a small amount
of per-cpu state and use this to simulate the upper 32-bits of the counter
providing a full 64-bit counter to the consumers of cpu_ticks().

18 years ago- Simplify the i8254 timecounter for the alpha since we don't actually have
jhb [Tue, 7 Mar 2006 21:44:20 +0000 (21:44 +0000)]
- Simplify the i8254 timecounter for the alpha since we don't actually have
  an interrupt handler for the i8254.  (Our clock interrupts come from
  elsewhere.)  Instead, use the same algo that i386 uses when the lapic
  timer is in use.  This lets us remove a lot of cruft that tried to handle
  the i8254 interrupts that we weren't even using or setting up a handler
  for.
- G/C a bunch of unused cruft while I'm here.
- Fix the code to not use the rpcc timecounter (similar to TSC) on SMP
  machines to only disable that timecounter if more than one CPU is in
  use by the kernel.  Previously, a UP kernel on a machine with multiple
  CPUs would needlessly disable this timecounter.

MFC after: 1 week

18 years agoFurther refine the bridge hack in the arp code. Only do the special arp
thompsa [Tue, 7 Mar 2006 21:40:44 +0000 (21:40 +0000)]
Further refine the bridge hack in the arp code. Only do the special arp
handling for interfaces which are actually in the bridge group, ignore all
others.

MFC after: 3 days

18 years agoFor consistency sake, use >= MINCLSIZE rather than > MINCLSIZE to determine
jhb [Tue, 7 Mar 2006 21:31:20 +0000 (21:31 +0000)]
For consistency sake, use >= MINCLSIZE rather than > MINCLSIZE to determine
whether or not to allocate a full mbuf cluster rather than just a plain
mbuf when adding on additional mbufs in m_getm().  In practice, there wasn't
any resulting mem trashing since m_getm() doesn't ever allocate an mbuf with
a packet header, and MINCLSIZE is the available payload in an mbuf with a
header rather than the available payload in a plain mbuf.

Discussed with:  andre (lightly)

18 years agoDisable 802.11e support for now. Did some cleanup while I'm here.
damien [Tue, 7 Mar 2006 20:48:23 +0000 (20:48 +0000)]
Disable 802.11e support for now.  Did some cleanup while I'm here.

18 years agoRemove trailing dot from the servername in the serverlist to be able
ume [Tue, 7 Mar 2006 19:13:00 +0000 (19:13 +0000)]
Remove trailing dot from the servername in the serverlist to be able
to use portsnap from behind a http proxy.  Some HTTP proxy don't like
trailing dot in the servername.

Approved by: cperciva
MFC after: 2 days

18 years agoCorrect handling of HTTP_PROXY_AUTH. A password may have `:'.
ume [Tue, 7 Mar 2006 19:04:16 +0000 (19:04 +0000)]
Correct handling of HTTP_PROXY_AUTH.  A password may have `:'.

Approved by: cperciva
MFC after: 2 days

18 years agoFix incorrect command line examples.
hrs [Tue, 7 Mar 2006 18:34:18 +0000 (18:34 +0000)]
Fix incorrect command line examples.

Submitted by: Yoshihiko Sarumaru (mistral at imasy dot or dot jp)
PR: docs/94158
MFC after: 2 days

18 years agoFix version number when the feature was added.
hrs [Tue, 7 Mar 2006 18:25:03 +0000 (18:25 +0000)]
Fix version number when the feature was added.

Submitted by: Yoshihiko Sarumaru (mistral at imasy dot or dot jp)
PR: docs/93801 and docs/93803
MFC after: 2 days