]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
18 years agoBecause JOB control has higher priority than single threading in
davidxu [Tue, 21 Mar 2006 08:41:15 +0000 (08:41 +0000)]
Because JOB control has higher priority than single threading in
thread_suspend_check(), call thread_stopped() to report SIGCHLD
if there is JOB control in progress.

18 years agoOpen the lock file write-only.
des [Tue, 21 Mar 2006 08:08:58 +0000 (08:08 +0000)]
Open the lock file write-only.

18 years agoPrepare to autogenerate the src.conf(5) manpage.
ru [Tue, 21 Mar 2006 07:50:50 +0000 (07:50 +0000)]
Prepare to autogenerate the src.conf(5) manpage.

18 years agoRemove an outdated README.
ru [Tue, 21 Mar 2006 07:26:51 +0000 (07:26 +0000)]
Remove an outdated README.

18 years agoRemove stale KSE code.
davidxu [Tue, 21 Mar 2006 06:46:27 +0000 (06:46 +0000)]
Remove stale KSE code.

Reviewed by: alc

18 years agoApply more thorough fixes while dealing with device opening and closing:
ariff [Tue, 21 Mar 2006 06:35:48 +0000 (06:35 +0000)]
Apply more thorough fixes while dealing with device opening and closing:

- Determine open direction using 'flags', not 'mode'. This bug exist since
  past 4 years.
- Don't allow opening the same device twice, be it in a same or different
  direction.
- O_RDWR is allowed, provided that it is done by a single open (for example
  by mixer(8)) and the underlying hardware support true full-duplex operation.
- Do various paranoid checking in case other process/thread trying to hijack
  the same device twice (or more).

MFC after: 5 days

18 years agoRecover (?) support for AD1815 based ISA soundcards.
ariff [Tue, 21 Mar 2006 03:47:25 +0000 (03:47 +0000)]
Recover (?) support for AD1815 based ISA soundcards.

PR: kern/94388
Submitted by: Krzysztof Kotlenga <piernik at gmail dot com>
MFC after: 3 days

18 years agoRemove unused leaked debug function prototype.
tegge [Tue, 21 Mar 2006 01:04:24 +0000 (01:04 +0000)]
Remove unused leaked debug function prototype.

18 years agoAdd descriptions for the sysctls:
keramida [Mon, 20 Mar 2006 21:44:12 +0000 (21:44 +0000)]
Add descriptions for the sysctls:

    net.inet.icmp.drop_redirect
    net.inet.icmp.log_redirect
    net.inet.icmp.icmplim
    net.inet.icmp.icmplim_output

Approved & text by: andre

18 years agoMinor style(9) and KNF elimination as I prepare to fix a bug.
imp [Mon, 20 Mar 2006 21:28:46 +0000 (21:28 +0000)]
Minor style(9) and KNF elimination as I prepare to fix a bug.

18 years agoregen
netchild [Mon, 20 Mar 2006 19:48:02 +0000 (19:48 +0000)]
regen

18 years agoFix tinderbox on alpha.
netchild [Mon, 20 Mar 2006 19:46:56 +0000 (19:46 +0000)]
Fix tinderbox on alpha.

Tested by: cross-compile

18 years agoDrop some unneeded casts since we program the kernel in C rather than C++.
jhb [Mon, 20 Mar 2006 19:39:08 +0000 (19:39 +0000)]
Drop some unneeded casts since we program the kernel in C rather than C++.

18 years agoregen: fix of linuxolator with testing in a cross-build
netchild [Mon, 20 Mar 2006 18:54:29 +0000 (18:54 +0000)]
regen: fix of linuxolator with testing in a cross-build

18 years agoFix the linuxolator on amd64 (cross-build).
netchild [Mon, 20 Mar 2006 18:53:26 +0000 (18:53 +0000)]
Fix the linuxolator on amd64 (cross-build).

18 years ago- Replace awk parts with shell.
flz [Mon, 20 Mar 2006 18:00:14 +0000 (18:00 +0000)]
- Replace awk parts with shell.

Approved by: brooks
MFC after: 2 weeks

18 years agoKeep references sorted alphabetically.
wkoszek [Mon, 20 Mar 2006 17:10:09 +0000 (17:10 +0000)]
Keep references sorted alphabetically.

Spotted by: brueffer
Approved by: cognet (mentor)

18 years agoThere is no need to pass NULL to the pam_error() as the last argument.
wkoszek [Mon, 20 Mar 2006 16:56:08 +0000 (16:56 +0000)]
There is no need to pass NULL to the pam_error() as the last argument.
Remove it.

Reviewed by: des
Approved by: cognet (mentor)

18 years agoAdd reference to conscontrol(8).
wkoszek [Mon, 20 Mar 2006 16:53:00 +0000 (16:53 +0000)]
Add reference to conscontrol(8).

Reviewed by: ru
Approved by: cognet (mentor)

18 years agoUpdate the pmc(3) manual page's date string and freshen the year in the (c) line for
jkoshy [Mon, 20 Mar 2006 15:28:59 +0000 (15:28 +0000)]
Update the pmc(3) manual page's date string and freshen the year in the (c) line for
pmc.3 and libpmc.c.

18 years agoExtend coverage of the MK_IPX build option to the following:
ru [Mon, 20 Mar 2006 14:24:58 +0000 (14:24 +0000)]
Extend coverage of the MK_IPX build option to the following:

- <netipx> headers [1]
- IPX library (libipx)
- IPX support in ifconfig(8)
- IPXrouted(8)
- new MK_NCP option

New MK_NCP build option controls:

- <netncp> and <fs/nwfs> headers
- NCP library (libncp)
- ncplist(1) and ncplogin(1)
- mount_nwfs(8)
- ncp and nwfs kernel modules

User knobs: WITHOUT_IPX, WITHOUT_IPX_SUPPORT, WITHOUT_NCP.

[1] <netsmb/netbios.h> unconditionally uses <netipx> headers
so they are still installed.  This needs to be dealt with.

18 years agoDocument the requirement to run rpc.lockd(8) on both client
ru [Mon, 20 Mar 2006 12:27:08 +0000 (12:27 +0000)]
Document the requirement to run rpc.lockd(8) on both client
and server when installing onto an NFS file system.

18 years agoSeparate completely full runs from runs that are merely almost full, so
jasone [Mon, 20 Mar 2006 04:05:05 +0000 (04:05 +0000)]
Separate completely full runs from runs that are merely almost full, so
that no linear searching is necessary if we resort to allocating from a
run that is known to be mostly full.  There are pathological edge cases
that could have caused severely degraded performance, and this change
fixes that.

18 years agoSet default contention scope to system.
davidxu [Mon, 20 Mar 2006 03:14:14 +0000 (03:14 +0000)]
Set default contention scope to system.

18 years agoRestore fd optimization with a few minor tweaks, to quote tegge:
csjp [Mon, 20 Mar 2006 00:13:47 +0000 (00:13 +0000)]
Restore fd optimization with a few minor tweaks, to quote tegge:

"fdinit() fails to initialize newfdp->fd_fd.fd_lastfile to -1.  This breaks
fdcopy() which will incorrectly set newfdp->fd_freefile to 1 if no files are
open and the last file descriptor marked as unused for fdp was 0.  This later
causes descriptor 0 to be unavailable in newfdp when the optimization is
enabled.

When the last file descriptor previously marked as used is nonzero and marked
as unused, fdunused() incorrectly sets fdp->fd_lastfile to fd - 1 due to
fd_last_used() returning (size - 1).  This hides the problem that breaks the
optimization."

This allows us to keep the optimization, while un-breaking it.

This is a RELENG_6 candidate.

PR: kern/87208
MFC after: 1 week
Submitted by: tegge

18 years agoAllow compilation when not using softupdates.
tegge [Sun, 19 Mar 2006 22:16:44 +0000 (22:16 +0000)]
Allow compilation when not using softupdates.

18 years agoLet snapshots make a copy of old contents for all buffers taking part in a
tegge [Sun, 19 Mar 2006 21:43:36 +0000 (21:43 +0000)]
Let snapshots make a copy of old contents for all buffers taking part in a
cluster instead of just the first buffer.

Delay buf_start() calls until snapshots have a copy of old content.

PR: kern/93942

18 years agoAdd kludge to avoid deadlock when unlinking snapshot.
tegge [Sun, 19 Mar 2006 21:29:20 +0000 (21:29 +0000)]
Add kludge to avoid deadlock when unlinking snapshot.

18 years agoReduce probability of unmount failing after having unmounted snapshots.
tegge [Sun, 19 Mar 2006 21:09:19 +0000 (21:09 +0000)]
Reduce probability of unmount failing after having unmounted snapshots.

18 years agoEnsure that vnode for directory isn't reclaimed before ffs_snapshot() has
tegge [Sun, 19 Mar 2006 21:05:10 +0000 (21:05 +0000)]
Ensure that vnode for directory isn't reclaimed before ffs_snapshot() has
completed expunging unlinked files.  It could come back at another memory
location causing a lock order reversal.

18 years agoCall vn_start_write() before locking vnode.
tegge [Sun, 19 Mar 2006 20:45:06 +0000 (20:45 +0000)]
Call vn_start_write() before locking vnode.

18 years agoDon't call vn_finished_write() if vn_start_write() failed.
tegge [Sun, 19 Mar 2006 20:43:07 +0000 (20:43 +0000)]
Don't call vn_finished_write() if vn_start_write() failed.

18 years ago - Correct an assert in vop_rename_pre. fdvp may be locked if it is either
jeff [Sun, 19 Mar 2006 20:14:46 +0000 (20:14 +0000)]
 - Correct an assert in vop_rename_pre.  fdvp may be locked if it is either
   the target directory or file.  This case should fail in the filesystem
   anyway and perhaps kern_rename() should catch it.

Sponsored by: Isilon Systems, Inc.

18 years agoFix spelling.
ru [Sun, 19 Mar 2006 19:10:39 +0000 (19:10 +0000)]
Fix spelling.

18 years agoOptimize realloc() to reallocate in place if the old and new sizes are
jasone [Sun, 19 Mar 2006 18:28:06 +0000 (18:28 +0000)]
Optimize realloc() to reallocate in place if the old and new sizes are
close enough to each other that reallocation would allocate a new region
of the same size.  This improves the performance of repeated incremental
reallocations by up to three orders of magnitude. [1]

Fix arena_new() to properly constrain run size if a small chunk size was
specified during runtime configuration.

Suggested by: se [1]

18 years agoMerge Perforce change 93581 from TrustedBSD audit3 branch:
rwatson [Sun, 19 Mar 2006 17:34:00 +0000 (17:34 +0000)]
Merge Perforce change 93581 from TrustedBSD audit3 branch:

  Mega-style patch.

Obtained from: TrustedBSD Project

18 years agoMerge Perforce changes 93512, 93514, 93515 from TrustedBSD audit3
rwatson [Sun, 19 Mar 2006 16:03:43 +0000 (16:03 +0000)]
Merge Perforce changes 93512, 93514, 93515 from TrustedBSD audit3
branch:

  Integrate audit.c to audit_worker.c, so as to migrate the worker
  thread implementation to its own .c file.

  Populate audit_worker.c using parts now removed from audit.c:

  - Move audit rotation global variables.
  - Move audit_record_write(), audit_worker_rotate(),
    audit_worker_drain(), audit_worker(), audit_rotate_vnode().
  - Create audit_worker_init() from relevant parts of audit_init(),
    which now calls this routine.
  - Recreate audit_free(), which wraps uma_zfree() so that
    audit_record_zone can be static to audit.c.
  - Unstaticize various types and variables relating to the audit
    record queue so that audit_worker can get to them.  We may want
    to wrap these in accessor methods at some point.
  - Move AUDIT_PRINTF() to audit_private.h.

  Addition of audit_worker.c to kernel configuration, missed in
  earlier submit.

Obtained from: TrustedBSD Project

18 years agoAlso refer to csup in the example SUP setting.
mux [Sun, 19 Mar 2006 15:54:45 +0000 (15:54 +0000)]
Also refer to csup in the example SUP setting.

18 years agoMerge Perforce change 93572 from TrustedBSD audit3 branch:
rwatson [Sun, 19 Mar 2006 15:47:53 +0000 (15:47 +0000)]
Merge Perforce change 93572 from TrustedBSD audit3 branch:

  Simple audit pipe ioctl regression test.

Obtained from: TrustedBSD Project

18 years agoMerge Perforce change 93569 from TrustedBSD audit3 branch:
rwatson [Sun, 19 Mar 2006 15:44:53 +0000 (15:44 +0000)]
Merge Perforce change 93569 from TrustedBSD audit3 branch:

  Do install sys/security/audit include files.  It would be nice just
  to install audit_ioctl.h, but we seem only to support installing
  directories, so we get them all.  The two not intended for extra-
  kernel use have !_KERNEL #error's, which should help.

Obtained from: TrustedBSD Project

18 years agoMerge Perforce change 93570 from TrustedBSD audit3 branch:
rwatson [Sun, 19 Mar 2006 15:39:03 +0000 (15:39 +0000)]
Merge Perforce change 93570 from TrustedBSD audit3 branch:

  Add audit pipe ioctls to query minimum and maximum audit queue
  lengths.

Obtained from: TrustedBSD Project

18 years agoMerge Perforce change 93567 from TrustedBSD audit3 branch:
rwatson [Sun, 19 Mar 2006 15:38:03 +0000 (15:38 +0000)]
Merge Perforce change 93567 from TrustedBSD audit3 branch:

  Bump default queue limit for audit pipes from 32 to 128, since 32 is
  pretty small.

Obtained from: TrustedBSD Project

18 years agoMerge Perforce change 93568 from TrustedBSD audit3 branch:
rwatson [Sun, 19 Mar 2006 15:37:04 +0000 (15:37 +0000)]
Merge Perforce change 93568 from TrustedBSD audit3 branch:

  Normalize nested include guards.

Obtained from: TrustedBSD Project

18 years agoChange the default SUP command to use csup instead of CVSup for
mux [Sun, 19 Mar 2006 15:36:56 +0000 (15:36 +0000)]
Change the default SUP command to use csup instead of CVSup for
"make update" when SUP_UPDATE is defined.

18 years agoMerge Perforce change 93506 from TrustedBSD audit3 branch:
rwatson [Sun, 19 Mar 2006 15:36:10 +0000 (15:36 +0000)]
Merge Perforce change 93506 from TrustedBSD audit3 branch:

  Add ioctls to audit pipes in order to allow querying of the current
  record queue state, setting of the queue limit, and querying of pipe
  statistics.

Obtained from: TrustedBSD Project

18 years agoActually I wanted 'nolockd' here instead of 'lockd'.
pjd [Sun, 19 Mar 2006 13:27:37 +0000 (13:27 +0000)]
Actually I wanted 'nolockd' here instead of 'lockd'.

MFC after: 2 days

18 years agoUpdate copyright for 2006.
pjd [Sun, 19 Mar 2006 12:55:51 +0000 (12:55 +0000)]
Update copyright for 2006.

18 years agoMake net.inet.ip.portrange.reservedhigh and
dwmalone [Sun, 19 Mar 2006 11:48:48 +0000 (11:48 +0000)]
Make net.inet.ip.portrange.reservedhigh and
net.inet.ip.portrange.reservedlow apply to IPv6 aswell as IPv4.

We could have made new sysctls for IPv6, but that potentially makes
things complicated for mapped addresses. This seems like the least
confusing option and least likely to cause obscure problems in the
future.

This change makes the mac_portacl module useful with IPv6 apps.

Reviewed by: ume
MFC after: 1 month

18 years agokern.geom.raid3.sync_requests=2 seems to be a better default - it still
pjd [Sun, 19 Mar 2006 11:18:33 +0000 (11:18 +0000)]
kern.geom.raid3.sync_requests=2 seems to be a better default - it still
keeps disks very busy, but makes system much more responsive.

While here, kill extra space.

18 years agoRegen.
ru [Sun, 19 Mar 2006 11:12:41 +0000 (11:12 +0000)]
Regen.

18 years agoUnbreak COMPAT_LINUX32 option support on amd64.
ru [Sun, 19 Mar 2006 11:10:33 +0000 (11:10 +0000)]
Unbreak COMPAT_LINUX32 option support on amd64.

Broken by: netchild

18 years agokern.geom.mirror.sync_requests=2 seems to be a better default - it still
pjd [Sun, 19 Mar 2006 10:49:05 +0000 (10:49 +0000)]
kern.geom.mirror.sync_requests=2 seems to be a better default - it still
keeps disks very busy, but makes system much more responsive.

While here, kill extra space.

18 years agoFix build until I find a way to handle this case properly.
ru [Sun, 19 Mar 2006 08:52:49 +0000 (08:52 +0000)]
Fix build until I find a way to handle this case properly.

18 years agoimplements section 2.2 of RFC4191, regarding the reserved preference value (10)
suz [Sun, 19 Mar 2006 06:38:39 +0000 (06:38 +0000)]
implements section 2.2 of RFC4191, regarding the reserved preference value (10)

Obtained from: KAME
MFC after: 1 day

18 years agoRevert last delta.
ru [Sun, 19 Mar 2006 06:14:30 +0000 (06:14 +0000)]
Revert last delta.

18 years agoCommit a forgotten part of NO_PROFILE/NO_LIB32 conversion.
ru [Sun, 19 Mar 2006 06:14:07 +0000 (06:14 +0000)]
Commit a forgotten part of NO_PROFILE/NO_LIB32 conversion.

18 years agos/ifconfig tunnel greN/ifconfig greN tunnel/
brd [Sun, 19 Mar 2006 05:52:35 +0000 (05:52 +0000)]
s/ifconfig tunnel greN/ifconfig greN tunnel/

Submitted by: ankon@#bsddocs
Approved by: ceri@

18 years agoComment out MK_PROFILE until ru@ can fix this properly
phk [Sun, 19 Mar 2006 04:49:11 +0000 (04:49 +0000)]
Comment out MK_PROFILE until ru@ can fix this properly

18 years agoAdd default for MK_PROFILE
phk [Sun, 19 Mar 2006 04:39:42 +0000 (04:39 +0000)]
Add default for MK_PROFILE

18 years agoupdates net.inet6.ip6.kame_version as the proof of the latest KAME merge
suz [Sun, 19 Mar 2006 02:11:42 +0000 (02:11 +0000)]
updates net.inet6.ip6.kame_version as the proof of the latest KAME merge

Reviewed by: KAME
MFC after: 2 days

18 years agoBack out fd optimization introduced in revision 1.280 as it appears to be
csjp [Sat, 18 Mar 2006 23:27:21 +0000 (23:27 +0000)]
Back out fd optimization introduced in revision 1.280 as it appears to be
really breaking things. Simple "close(0); dup(fd)" does not return descriptor
"0" in some cases. Further, this change also breaks some MAC interactions with
mac_execve_will_transition().  Under certain circumstances, fdcheckstd() can
be called in execve(2) causing an assertion that checks to make sure that
stdin, stdout and stderr reside at indexes 0, 1 and 2 in the process fd table
to fail, resulting in a kernel panic when INVARIANTS is on.

This should also kill the "dup(2) regression on 6.x" show stopper item on the
6.1-RELEASE TODO list.

This is a RELENG_6 candidate.

PR: kern/87208
Silence from: des
MFC after: 1 week

18 years agoConvert NO_PROFILE and NO_LIB32 to new style.
ru [Sat, 18 Mar 2006 21:37:05 +0000 (21:37 +0000)]
Convert NO_PROFILE and NO_LIB32 to new style.

18 years agoAdd reference to PR to TOSHIBA TransMemory quirk entry.
joerg [Sat, 18 Mar 2006 21:13:14 +0000 (21:13 +0000)]
Add reference to PR to TOSHIBA TransMemory quirk entry.

18 years agoregen
netchild [Sat, 18 Mar 2006 20:49:01 +0000 (20:49 +0000)]
regen

18 years agoFixup some problems in my previous commit (COMPAT_43).
netchild [Sat, 18 Mar 2006 20:47:36 +0000 (20:47 +0000)]
Fixup some problems in my previous commit (COMPAT_43).

Pointyhat to: netchild

18 years agoEnable global pages TLB extension on Application Processors.
ups [Sat, 18 Mar 2006 19:32:46 +0000 (19:32 +0000)]
Enable global pages TLB extension on Application Processors.

MFC after: 3 days

18 years agoMerge perforce 93507:
rwatson [Sat, 18 Mar 2006 18:32:12 +0000 (18:32 +0000)]
Merge perforce 93507:

  Correct comment: this print is now from audit_record_write(), not
  audit_worker().

Obtained from: TrustedBSD Project

18 years agoMerge perforce change 93199:
rwatson [Sat, 18 Mar 2006 18:31:24 +0000 (18:31 +0000)]
Merge perforce change 93199:

  Change send_trigger() prototype to return an int, so that user
  space callers can tell if the message was successfully placed
  in the trigger queue.  This isn't quite the same as it being
  successfully received, but is close enough that we can generate
  a more useful warning message in audit(8).

Obtained from: TrustedBSD Project

18 years ago- Add a PR number for future reference.
flz [Sat, 18 Mar 2006 18:28:13 +0000 (18:28 +0000)]
- Add a PR number for future reference.

Approved by: njl

18 years agoregen after COMPAT_43 removal
netchild [Sat, 18 Mar 2006 18:24:38 +0000 (18:24 +0000)]
regen after COMPAT_43 removal

18 years agoGet rid of the need of COMPAT_43 in the linuxolator.
netchild [Sat, 18 Mar 2006 18:20:17 +0000 (18:20 +0000)]
Get rid of the need of COMPAT_43 in the linuxolator.

Submitted by: Divacky Roman <xdivac02@stud.fit.vutbr.cz>
Obtained from: DragonFly (some parts)

18 years agoOops - fix the build in the !USB_DEBUG case.
iedowse [Sat, 18 Mar 2006 17:57:34 +0000 (17:57 +0000)]
Oops - fix the build in the !USB_DEBUG case.

18 years agoAdd the new "showconfig" target that displays build configuration.
ru [Sat, 18 Mar 2006 14:30:23 +0000 (14:30 +0000)]
Add the new "showconfig" target that displays build configuration.

Suggested by: phk

18 years agoLet the EHCI hardware track the toggle state for bulk and interrupt
iedowse [Sat, 18 Mar 2006 13:55:16 +0000 (13:55 +0000)]
Let the EHCI hardware track the toggle state for bulk and interrupt
transfers. This fixes some cases where the software toggle tracking
was not doing the right thing. For example, a short transfer that
transferred 0 bytes of the requested qTD transfer size does cause
a toggle change, but the existing code was assuming it didn't.

Reported and tested by: pav
MFC after: 2 weeks

18 years agoProperly use the sense key info
sos [Sat, 18 Mar 2006 13:14:38 +0000 (13:14 +0000)]
Properly use the sense key info

18 years agoProvide alternate default for SHLIBDIR before bsd.own.mk does this.
ru [Sat, 18 Mar 2006 11:01:06 +0000 (11:01 +0000)]
Provide alternate default for SHLIBDIR before bsd.own.mk does this.

Reported by: phk

18 years agoReplace .sinclude by .if exists()/.include/.endif so that older make(1)'s can
ru [Sat, 18 Mar 2006 10:54:09 +0000 (10:54 +0000)]
Replace .sinclude by .if exists()/.include/.endif so that older make(1)'s can
handle it as well.

18 years agoadd an entry for ozfod..
jmg [Sat, 18 Mar 2006 10:40:28 +0000 (10:40 +0000)]
add an entry for ozfod..

MFC after: 3 days

18 years agoWhen deconfiguring a log, only stop PMCs that are in the RUNNING
jkoshy [Sat, 18 Mar 2006 03:54:07 +0000 (03:54 +0000)]
When deconfiguring a log, only stop PMCs that are in the RUNNING
state.

18 years agoMFp4:
imp [Sat, 18 Mar 2006 01:45:29 +0000 (01:45 +0000)]
MFp4:

Add bus attachment for the ohci device on this chip.  The bus and hub
are detected correctly, but the children devices aren't detected
correctly for reasons unknown.

18 years agoAdd ohci controller mapping.
imp [Sat, 18 Mar 2006 01:43:54 +0000 (01:43 +0000)]
Add ohci controller mapping.

18 years agoMFp4:
imp [Sat, 18 Mar 2006 01:43:25 +0000 (01:43 +0000)]
MFp4:

o update TODO list
o Better use of busdma
o mark RX dtors as COHERENT.  This helps performance a lot by not requiring
  so many EXPENSIVE cache flushes.  The cost of accessing it non-cached
  is much smaller.
o Copy data from Rx buffers to make IP header 4 byte aligned.
o CRC length included in reported length, so cope
o Don't free TX buffer twice
o Manage TX buffers better.
o Enable just the interrupts we want.
o Manage OACTIVE better

# Some of these done by cognet
# These changes let us get to # via NFS root.

18 years agoMFP4:
imp [Sat, 18 Mar 2006 01:39:23 +0000 (01:39 +0000)]
MFP4:
Gratuitously sort alphabetically.

18 years agoMFP4:
imp [Sat, 18 Mar 2006 01:38:25 +0000 (01:38 +0000)]
MFP4:

GC and fix definitions.

# some of this may have been done by cognet

18 years agoMFp4:
imp [Sat, 18 Mar 2006 01:35:49 +0000 (01:35 +0000)]
MFp4:

o Add memory barrier to bus space
o Allow for up to 3 IRQs per device
o Move to table driven population of children devices.
o Add support for usb ohci memory mapped controller resource allocation.
o Clean up a bunch of extra writes to disable interrupts that are now
  done elsewhere.
o Force all system interrupt handlers be fast.  We get deadlock if they
  aren't.

18 years agoMFp4:
imp [Sat, 18 Mar 2006 01:30:31 +0000 (01:30 +0000)]
MFp4:

o Disable all interrupts that the ST can generate until we have an ISR
  to service them.
o Correct clock calculation to make DELAY the right length...

Submitted by: cognet (#2)

18 years agofix spelling of ozfod... I've been wondering why we'd need to fill a page
jmg [Sat, 18 Mar 2006 01:06:13 +0000 (01:06 +0000)]
fix spelling of ozfod...  I've been wondering why we'd need to fill a page
w/ non-zero data, and it turns out we don't...  This is really optimized
zero filled on demand, or pages that were already zero'd for us...

MFC after: 3 days

18 years agoIf an NFS server returns more than a few EJUKEBOX errors for a given RPC
cel [Fri, 17 Mar 2006 22:14:23 +0000 (22:14 +0000)]
If an NFS server returns more than a few EJUKEBOX errors for a given RPC
request, the FreeBSD NFS client will quickly back off to a excessively
long wait (days, then weeks) before retrying the request.

Change the behavior of the FreeBSD NFS client to match the behavior of
the reference NFS client implementation (Solaris).  This provides a fixed
delay of 10 seconds between each retry by default.  A sysctl, called
nfs3_jukebox_delay, is now available to tune the delay.  Unlike Solaris,
the sysctl value on FreeBSD is in seconds, rather than in HZ.

Sponsored by: Network Appliance, Incorporated
Reviewed by: rick
Approved by: silby
MFC after: 3 days

18 years agoA few minor corrections to the mkuzip.8 man page.
sobomax [Fri, 17 Mar 2006 20:48:10 +0000 (20:48 +0000)]
A few minor corrections to the mkuzip.8 man page.

PR: 92576
Submitted by: Stefan Bethke

18 years agoModify netatalk to ensure, and assert, that pcb's remain attached to
rwatson [Fri, 17 Mar 2006 20:40:17 +0000 (20:40 +0000)]
Modify netatalk to ensure, and assert, that pcb's remain attached to
sockets as long as the sockets have not been aborted or detached.  Do
not try to free the socket in pru_detach(), since sofree() will do so,
if needed, once pru_detach() returns.

Annotate a bug in ddp_abort(), which fails to free the socket; this
is probably OK as ddp_abort() should never be called, so should
instead be deleted.

18 years agoChange s_net to s_node when printing the destination address.
rwatson [Fri, 17 Mar 2006 20:14:25 +0000 (20:14 +0000)]
Change s_net to s_node when printing the destination address.

18 years agoReimplementation of world/kernel build options. For details, see:
ru [Fri, 17 Mar 2006 18:54:44 +0000 (18:54 +0000)]
Reimplementation of world/kernel build options.  For details, see:

http://lists.freebsd.org/pipermail/freebsd-current/2006-March/061725.html

The src.conf(5) manpage is to follow in a few days.

Brought to you by: imp, jhb, kris, phk, ru (all bugs are mine)

18 years agoChange so_pcb invariants in netnatm, such that netnatm sockets always have
rwatson [Fri, 17 Mar 2006 18:25:57 +0000 (18:25 +0000)]
Change so_pcb invariants in netnatm, such that netnatm sockets always have
pcb's allocated:

- Universally ensure (and assert) that so_pcb is not NULL, removing lots
  of checks and error cases.  Don't free the pcb without clearing the
  so_pcb pointer.

- Don't try to free the socket in pru_detach(), since the caller will
  immediately free the socket.

- Do retain the sotryfree() in pru_abort() for now, although eventually
  the caller will do it unconditionally.

18 years agoSony DSC cameras with revision number 6.0 require the same quirk
iedowse [Fri, 17 Mar 2006 18:16:22 +0000 (18:16 +0000)]
Sony DSC cameras with revision number 6.0 require the same quirk
as those reporting revision 5.0.

Reported by: Brian Reichert
MFC after: 1 week

18 years agoThis script symlinks /boot/kernel to the directory that contains the kernel
des [Fri, 17 Mar 2006 16:28:12 +0000 (16:28 +0000)]
This script symlinks /boot/kernel to the directory that contains the kernel
you booted from, unless /boot/kernel already exists and is not a symlink.
This should only affect people like me who juggle multiple kernels and
have KODIR = /boot/${KERN_IDENT} in /etc/make.conf to keep them apart.

18 years agoModify UNIX domain sockets to guarantee, and assume, that so_pcb is always
rwatson [Fri, 17 Mar 2006 13:52:57 +0000 (13:52 +0000)]
Modify UNIX domain sockets to guarantee, and assume, that so_pcb is always
defined for an in-use socket.  This allows us to eliminate countless tests
of whether so_pcb is non-NULL, eliminating dozens of error cases.  For
now, retain the call to sotryfree() in the uipc_abort() path, but this
will eventually move to soabort().

These new assumptions should be largely correct, and will become more so
as the socket/pcb reference model is fixed.  Removing the notion that
so_pcb can be non-NULL is a critical step towards further fine-graining
of the UNIX domain socket locking, as the so_pcb reference no longer
needs to be protected using locks, instead it is a property of the socket
life cycle.

18 years agoCorrect the requirements for g_new_consumer().
des [Fri, 17 Mar 2006 13:13:18 +0000 (13:13 +0000)]
Correct the requirements for g_new_consumer().

Approved by: phk

18 years agoDon't build the ipw and iwi on pc98.
nyan [Fri, 17 Mar 2006 13:06:19 +0000 (13:06 +0000)]
Don't build the ipw and iwi on pc98.

18 years agoThis commit was generated by cvs2svn to compensate for changes in r156803,
harti [Fri, 17 Mar 2006 10:04:10 +0000 (10:04 +0000)]
This commit was generated by cvs2svn to compensate for changes in r156803,
which included commits to RCS files with non-trunk default branches.

18 years agoThis is actually a section 3 man page, not section 1. Also fix a couple
harti [Fri, 17 Mar 2006 10:04:10 +0000 (10:04 +0000)]
This is actually a section 3 man page, not section 1. Also fix a couple
of grammatical errors.

Submitted by: jkoshy@