]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agoMFlibthr:
David Xu [Mon, 19 Nov 2007 02:09:07 +0000 (02:09 +0000)]
MFlibthr:
In _pthread_key_create() ensure that libkse is initialized.

16 years agoReduce the WARNS level across the board due to a warning that occurs
John Birrell [Mon, 19 Nov 2007 00:19:01 +0000 (00:19 +0000)]
Reduce the WARNS level across the board due to a warning that occurs
on at least arm and sparc64. i386/pc98, ia64 and amd64 are all OK.

16 years agoCapitalization
Mike Makonnen [Sun, 18 Nov 2007 18:44:35 +0000 (18:44 +0000)]
Capitalization

16 years agoLet sunkbd(4) emulate an AT keyboard by default.
Marius Strobl [Sun, 18 Nov 2007 18:11:16 +0000 (18:11 +0000)]
Let sunkbd(4) emulate an AT keyboard by default.
This has the following benefits:
- allows to use the AT keyboard maps in share/syscons/keymaps with
  sunkbd(4),
- allows to use kbdmux(4) with sunkbd(4),
- allows Sun RS232 keyboards to be configured and used the same
  way as Sun USB keyboards driven by ukbd(4) (which also does AT
  keyboard emulation) with X.Org, putting an end to the problem
  of native support for the former in X.Org being broken over and
  over again.

MFC after: 3 days

16 years agoAdd generic support for chipsets that say they support AHCI. This should catch new...
Søren Schmidt [Sun, 18 Nov 2007 14:44:52 +0000 (14:44 +0000)]
Add generic support for chipsets that say they support AHCI.  This should catch new chipsets that we dont know but that we should support.
Add a few new PCI id's.
Misc cleanups.

16 years agoExpand lock class with the "virtual" function lc_assert which will offer
Attilio Rao [Sun, 18 Nov 2007 14:43:53 +0000 (14:43 +0000)]
Expand lock class with the "virtual" function lc_assert which will offer
an unified way for all the lock primitives to express lock assertions.
Currenty, lockmgrs and rmlocks don't have assertions, so just panic in
that case.
This will be a base for more callout improvements.

Ok'ed by: jhb, jeff

16 years ago- Add in missing event handler invokes for initial proc and thread.
Randall Stewart [Sun, 18 Nov 2007 13:56:51 +0000 (13:56 +0000)]
- Add in missing event handler invokes for initial proc and thread.

16 years agoUse LC_ALL=C before awk generating "optionlist"
Andrey A. Chernov [Sun, 18 Nov 2007 11:59:44 +0000 (11:59 +0000)]
Use LC_ALL=C before awk generating "optionlist"
(and before two other awk calls, just to be safe)
Without it sorting is broken for locales with ASCII collating equivalence
like (v,w) in sv_SE

16 years agoo English lesson from bde@: "iff" is not a typo, it means "if and only if".
Maxim Konovalov [Sun, 18 Nov 2007 09:21:30 +0000 (09:21 +0000)]
o English lesson from bde@: "iff" is not a typo, it means "if and only if".
Backout previous.

16 years agoo Fix Dd format.
Maxim Konovalov [Sun, 18 Nov 2007 09:18:20 +0000 (09:18 +0000)]
o Fix Dd format.

Pointed out by: ru

16 years agoUse an explicit pointer cast since NULL isn't guaranteed to be defined
John Birrell [Sun, 18 Nov 2007 09:13:08 +0000 (09:13 +0000)]
Use an explicit pointer cast since NULL isn't guaranteed to be defined
as a pointer type.

Submitted by: Christoph Mallon - christoph mallon at gmx de

16 years agoMFp4: Several fixes to tmpfs which makes it to survive from pho@'s
Xin LI [Sun, 18 Nov 2007 04:52:40 +0000 (04:52 +0000)]
MFp4: Several fixes to tmpfs which makes it to survive from pho@'s
strees2 suite, to quote his letter, this change:

1. It removes the tn_lookup_dirent stuff. I think this cannot be fixed,
   because nothing protects vnode/tmpfs node between lookup is done, and
   actual operation is performed, in the case the vnode lock is dropped.
   At least, this is the case with the from vnode for rename.

   For now, we do the linear lookup in the parent node. This has its own
   drawbacks. Not mentioning speed (that could be fixed by using hash), the
   real problem is the situation where several hardlinks exist in the dvp.
   But, I think this is fixable.

2. The patch restores the VV_ROOT flag on the root vnode after it became
   reclaimed and allocated again. This fixes MPASS assertion at the start
   of the tmpfs_lookup() reported by many.

Submitted by: kib

16 years agoMFp4: Fix several style(9) bugs.
Xin LI [Sun, 18 Nov 2007 04:40:42 +0000 (04:40 +0000)]
MFp4: Fix several style(9) bugs.

Submitted by: des

16 years agoUse a NULL instead of a zero to avoid a compiler warning about a missing sentinel.
John Birrell [Sun, 18 Nov 2007 03:45:55 +0000 (03:45 +0000)]
Use a NULL instead of a zero to avoid a compiler warning about a missing sentinel.

16 years agoThis makefile builds contrib code, so I won't try to fix all the
John Birrell [Sun, 18 Nov 2007 03:29:10 +0000 (03:29 +0000)]
This makefile builds contrib code, so I won't try to fix all the
casts from pointer to int here.

16 years agoFix signed/unsigned comparison compiler warning.
John Birrell [Sun, 18 Nov 2007 03:17:56 +0000 (03:17 +0000)]
Fix signed/unsigned comparison compiler warning.

16 years agoMFi386: revision 1.18
Yoshihiro Takahashi [Sun, 18 Nov 2007 03:14:06 +0000 (03:14 +0000)]
MFi386: revision 1.18

  Add a trailing \0 to the read error string so that read errors don't print
  out two error messages.

16 years agoConstify the first argument to expand_number() so that it can
John Birrell [Sun, 18 Nov 2007 02:20:02 +0000 (02:20 +0000)]
Constify the first argument to expand_number() so that it can
be called with a const without the compiler grisling.

16 years agoReduce the WARNS level to avoid a compiler warning about a variable
John Birrell [Sun, 18 Nov 2007 01:53:07 +0000 (01:53 +0000)]
Reduce the WARNS level to avoid a compiler warning about a variable
possibly being clobbered by a longjmp or a fork with gcc4.

16 years agoUse a cast that doesn't create a compiler warning.
John Birrell [Sun, 18 Nov 2007 00:33:23 +0000 (00:33 +0000)]
Use a cast that doesn't create a compiler warning.

16 years agoReduce the WARNS levels in a couple of places so that we can enable
John Birrell [Sun, 18 Nov 2007 00:29:36 +0000 (00:29 +0000)]
Reduce the WARNS levels in a couple of places so that we can enable
-Werror at higher WARNS levels like we did before gcc4 was imported.

16 years agoFix a compiler warning.
John Birrell [Sun, 18 Nov 2007 00:25:18 +0000 (00:25 +0000)]
Fix a compiler warning.

16 years agoAdd a function to list symbols in a file and their values at the
John Birrell [Sun, 18 Nov 2007 00:23:31 +0000 (00:23 +0000)]
Add a function to list symbols in a file and their values at the
same time rather than having to list the symbols and then go back
and look each one up by name.

16 years agoUse variable types which match the function prototypes.
John Birrell [Sat, 17 Nov 2007 23:21:38 +0000 (23:21 +0000)]
Use variable types which match the function prototypes.

16 years agoFix a compiler warning by using a printf format matching the variable
John Birrell [Sat, 17 Nov 2007 23:17:05 +0000 (23:17 +0000)]
Fix a compiler warning by using a printf format matching the variable
type.

16 years agoUse uintptr_t to cast a pointer to an integer to avoid compiler warnings
John Birrell [Sat, 17 Nov 2007 23:14:06 +0000 (23:14 +0000)]
Use uintptr_t to cast a pointer to an integer to avoid compiler warnings
on processors where sizeof(void *) > sizeof(int).

16 years agoUse printf formats which match the variable types without casts so we
John Birrell [Sat, 17 Nov 2007 23:09:39 +0000 (23:09 +0000)]
Use printf formats which match the variable types without casts so we
can go back to using -Werror in higher WARNS levels.

16 years agoPrevent the leakage of wired pages in the following circumstances:
Alan Cox [Sat, 17 Nov 2007 22:52:29 +0000 (22:52 +0000)]
Prevent the leakage of wired pages in the following circumstances:
First, a file is mmap(2)ed and then mlock(2)ed.  Later, it is truncated.
Under "normal" circumstances, i.e., when the file is not mlock(2)ed, the
pages beyond the EOF are unmapped and freed.  However, when the file is
mlock(2)ed, the pages beyond the EOF are unmapped but not freed because
they have a non-zero wire count.  This can be a mistake.  Specifically,
it is a mistake if the sole reason why the pages are wired is because of
wired, managed mappings.  Previously, unmapping the pages destroys these
wired, managed mappings, but does not reduce the pages' wire count.
Consequently, when the file is unmapped, the pages are not unwired
because the wired mapping has been destroyed.  Moreover, when the vm
object is finally destroyed, the pages are leaked because they are still
wired.  The fix is to reduce the pages' wired count by the number of
wired, managed mappings destroyed.  To do this, I introduce a new pmap
function pmap_page_wired_mappings() that returns the number of managed
mappings to the given physical page that are wired, and I use this
function in vm_object_page_remove().

Reviewed by: tegge
MFC after: 6 weeks

16 years ago- New sysctl variable: net.inet.ip.dummynet.io_fast
Oleg Bulyzhin [Sat, 17 Nov 2007 21:54:57 +0000 (21:54 +0000)]
- New sysctl variable: net.inet.ip.dummynet.io_fast
  If it is set to zero value (default) dummynet module will try to emulate
  real link as close as possible (bandwidth & latency): packet will not leave
  pipe faster than it should be on real link with given bandwidth.
  (This is original behaviour of dummynet which was altered in previous commit)
  If it is set to non-zero value only bandwidth is enforced: packet's latency
  can be lower comparing to real link with given bandwidth.

- Document recently introduced dummynet(4) sysctl variables.

Requested by: luigi, julian
MFC after: 3 month

16 years agoAdd thread support for arm.
Olivier Houchard [Sat, 17 Nov 2007 21:30:03 +0000 (21:30 +0000)]
Add thread support for arm.

MFC After: 1 week

16 years agoBuild libthread_db for arm as well.
Olivier Houchard [Sat, 17 Nov 2007 21:29:02 +0000 (21:29 +0000)]
Build libthread_db for arm as well.

MFC after: 1 week

16 years agoAdd arm support in libthread_db.
Olivier Houchard [Sat, 17 Nov 2007 21:27:53 +0000 (21:27 +0000)]
Add arm support in libthread_db.

16 years agoExpress in the usage() and SYNOPSIS that -F depends on -s, and
Ruslan Ermilov [Sat, 17 Nov 2007 21:01:22 +0000 (21:01 +0000)]
Express in the usage() and SYNOPSIS that -F depends on -s, and
that -f and -i are exclusive.

16 years agoSwap "source"/"target" where appropriate, to match documentation.
Ruslan Ermilov [Sat, 17 Nov 2007 20:24:58 +0000 (20:24 +0000)]
Swap "source"/"target" where appropriate, to match documentation.

16 years agoThere's no reason why we couldn't attach to a live process on arm.
Olivier Houchard [Sat, 17 Nov 2007 18:41:04 +0000 (18:41 +0000)]
There's no reason why we couldn't attach to a live process on arm.

MFC After: 1 week

16 years agoAdd a trailing \0 to the read error string so that read errors don't print
John Baldwin [Sat, 17 Nov 2007 17:32:40 +0000 (17:32 +0000)]
Add a trailing \0 to the read error string so that read errors don't print
out two error messages.

MFC after: 3 days

16 years agoAdd a kernel config file for the Hot-e HL200 (AT91RM92 based).
Olivier Houchard [Sat, 17 Nov 2007 17:25:22 +0000 (17:25 +0000)]
Add a kernel config file for the Hot-e HL200 (AT91RM92 based).
Many thanks to John Nicholls from Thinklinx for sending sample hardware.

16 years agoo Mask maximum file permissions we get from mount_ntfs -m
Maxim Konovalov [Sat, 17 Nov 2007 17:05:01 +0000 (17:05 +0000)]
o Mask maximum file permissions we get from mount_ntfs -m
with ACCESSPERMS.  Document in mount_ntfs(8) only the nine
low-order bits of mask are used (taken from mount_msdosfs(8)).

PR: kern/114856
Submitted by: Ighighi
MFC after: 1 month

16 years agoo Fix a typo in the comment.
Maxim Konovalov [Sat, 17 Nov 2007 16:19:48 +0000 (16:19 +0000)]
o Fix a typo in the comment.

16 years agoMove the priv check before the malloc call for so_pcb.
Bjoern A. Zeeb [Fri, 16 Nov 2007 22:35:33 +0000 (22:35 +0000)]
Move the priv check before the malloc call for so_pcb.
In case attach fails because of the priv check we leaked the
memory and left so_pcb as fodder for invariants.

Reported  by: Pawel Worach
Reviewed by: rwatson

16 years agoMention less v415 update
Xin LI [Fri, 16 Nov 2007 22:28:00 +0000 (22:28 +0000)]
Mention less v415 update

16 years agoThis commit was generated by cvs2svn to compensate for changes in r173686,
Xin LI [Fri, 16 Nov 2007 22:26:10 +0000 (22:26 +0000)]
This commit was generated by cvs2svn to compensate for changes in r173686,
which included commits to RCS files with non-trunk default branches.

16 years agoResolve conflicts; Update defines.h to reflect latest change.
Xin LI [Fri, 16 Nov 2007 22:24:31 +0000 (22:24 +0000)]
Resolve conflicts; Update defines.h to reflect latest change.

16 years agoThis commit was generated by cvs2svn to compensate for changes in r173682,
Xin LI [Fri, 16 Nov 2007 22:22:17 +0000 (22:22 +0000)]
This commit was generated by cvs2svn to compensate for changes in r173682,
which included commits to RCS files with non-trunk default branches.

16 years agoVirgin import of less v415.
Xin LI [Fri, 16 Nov 2007 22:22:17 +0000 (22:22 +0000)]
Virgin import of less v415.

16 years agoTeach kgdb how to handle double fault frames on i386:
John Baldwin [Fri, 16 Nov 2007 22:17:37 +0000 (22:17 +0000)]
Teach kgdb how to handle double fault frames on i386:
- Save td_oncpu in 'struct kthr' so the i386 target code can see which CPU
  a thread is running on.
- Add a new frame unwinder for double fault frames.  This unwinder is used
  when "dblfault_handler" is encountered in the stack.  It uses the CPU of
  the current thread to lookup the base address of the TSS used for the
  double fault from the GDT.  It then fetches the various registers out
  of the TSS similar to how the current trapframe unwinder fetches
  registers out of the trapframe.

MFC after: 3 days

16 years agoRemove empty setup and cleanup functions for the pfkey test.
Bjoern A. Zeeb [Fri, 16 Nov 2007 21:24:45 +0000 (21:24 +0000)]
Remove empty setup and cleanup functions for the pfkey test.

Add regression tests for privileged and supposedly unprivileged
IP_IPSEC_POLICY,IPV6_IPSEC_POLICY setsockopt cases.

We may need to review the current 'good' results to make
sure they reflect what we really want.

Discussed with: rwatson
Reviewed by: rwatson

16 years agoMake VPD register access more robust:
Jung-uk Kim [Fri, 16 Nov 2007 20:49:34 +0000 (20:49 +0000)]
Make VPD register access more robust:

- Implement timing out of VPD register access.[1]
- Fix an off-by-one error of freeing malloc'd space when checksum is invalid.
- Fix style(9) bugs, i.e., sizeof cannot be followed by space.
- Retire now obsolete 'hw.pci.enable_vpd' tunable.

Submitted by: cokane (initial revision)[1]
Reviewed by: marius (intermediate revision)
Silence from: jhb, jmg, rwatson
Tested by: cokane, jkim
MFC after: 3 days

16 years agoo s/resiserfs_sb/reiserfs_sb/.
Maxim Konovalov [Fri, 16 Nov 2007 19:43:26 +0000 (19:43 +0000)]
o s/resiserfs_sb/reiserfs_sb/.

Submitted by: Ighighi

16 years agoFix exit code when using -l on a non-existent md(4) device.
Florent Thoumie [Fri, 16 Nov 2007 17:40:24 +0000 (17:40 +0000)]
Fix exit code when using -l on a non-existent md(4) device.

PR: conf/116177
Submitted by: Remi Guyomarch <rguyom@pobox.com>
MFC after: 3 days

16 years agoDo not report MAC, TX, and RX stats via sysctl(8) with BCM5705+.
Jung-uk Kim [Fri, 16 Nov 2007 16:39:27 +0000 (16:39 +0000)]
Do not report MAC, TX, and RX stats via sysctl(8) with BCM5705+.
The register layout is little different from memory-mapped stats
in the previous generation chips.  In fact, it is bad because
registers in this range are cleared after reading them.

Reviewed by: scottl
MFC after: 3 days

16 years agouse the private task q thread instead of the shared system thread
Sam Leffler [Fri, 16 Nov 2007 15:51:47 +0000 (15:51 +0000)]
use the private task q thread instead of the shared system thread

Reviewed by: yongari
MFC after: 1 week

16 years agoFix bug in handling SDP continuation state.
Maksim Yevmenkin [Fri, 16 Nov 2007 15:13:12 +0000 (15:13 +0000)]
Fix bug in handling SDP continuation state.

Reported by: Eric Millbrandt emillbrandt at coldhaus dot com
MFC after: 3 days

16 years agoAdd change missing from previous commit: Remove temporary file.
Colin Percival [Fri, 16 Nov 2007 14:00:22 +0000 (14:00 +0000)]
Add change missing from previous commit: Remove temporary file.

16 years agoFix "freebsd-update rollback" applied to minor-version upgrades: Old
Colin Percival [Fri, 16 Nov 2007 13:57:41 +0000 (13:57 +0000)]
Fix "freebsd-update rollback" applied to minor-version upgrades:  Old
shared object files which have the same name as currently-installed
shared object files should be reinstalled after binaries are rolled
back.  The order for rolling back updates is therefore
1. Install any old shared object files which can be installed without
overwriting a new shared object file.
2. Rollback everything which isn't a shared object or kernel file.
3. Rollback any shared object files which we didn't deal with in (1).
4. Rollback to the old kernel.

Bug reported by: Jan Henrik Sylvester
MFC after: 3 days

16 years agoThe i386 disc1 size wandered above what fits on 650Mb media which has been
Ken Smith [Fri, 16 Nov 2007 13:14:27 +0000 (13:14 +0000)]
The i386 disc1 size wandered above what fits on 650Mb media which has been
our target previously.  So i386 joins the others that have switched over
to a separate livefs iso image.

16 years agoMFP4(128855, 129015):
Tai-hwa Liang [Fri, 16 Nov 2007 11:22:18 +0000 (11:22 +0000)]
MFP4(128855, 129015):
- Trying to eliminate another racing by replacing the timeout(9) with
  callout APIs.  In addition to that, the callout_drain() in an_detach()
  help us to avoid a possible panic-on-free due to the callout API tries
  to lock a destroyed mutex.
- In an_stats_update(), check the return value of an_read_record(). This
  should reduce the chance of device removal(PCCARD) panic [2].
- Adding a comment to state the fact that an_stats_update() is now called
  via callout(9) with a lock held [2].

Submitted by: jhb [1], ambrisko [2]
Reviewed by: jhb, ambrisko
Reported by: dhw
Tested by: dhw
MFC after: 3 days

16 years agoReset autonegotation timer if media option is not IFM_AUTO.
Pyun YongHyeon [Fri, 16 Nov 2007 10:39:18 +0000 (10:39 +0000)]
Reset autonegotation timer if media option is not IFM_AUTO.
Make mii_ticks advance, autonegiation is retried every
mii_anegticks seconds.

16 years agoRead MII_ANAR register and get common denominator ability.
Pyun YongHyeon [Fri, 16 Nov 2007 10:32:10 +0000 (10:32 +0000)]
Read MII_ANAR register and get common denominator ability.

PR: 92599

16 years agoIEEE 802.3 Annex 28B.3 explicitly specifies the following relative
Pyun YongHyeon [Fri, 16 Nov 2007 10:25:36 +0000 (10:25 +0000)]
IEEE 802.3 Annex 28B.3 explicitly specifies the following relative
priorities of the technologies supported by 802.3 Selector Field
value.

1000BASE-T full duplex
1000BASE-T
100BASE-T2 full duplex
100BASE-TX full duplex
100BASE-T2
100BASE-T4
100BASE-TX
10BASE-T full duplex
10BAST-T

However PHY drivers didn't honor the order such that 100BASE-T4 had
higher priority than 100BASE-TX full duplex. Fix that long standing
bugs such that have PHY drivers choose the highest common denominator
ability.
Fix a bug in dcphy which inadvertently aceepts 100BASE-T4.

PR: 92599

16 years agoCatch up to changes in rescue on ia64.
Marcel Moolenaar [Fri, 16 Nov 2007 04:03:08 +0000 (04:03 +0000)]
Catch up to changes in rescue on ia64.

16 years agoAdd PCI device support for Intel S7000FC4UR in usb and ichwd, systems
Jack F Vogel [Thu, 15 Nov 2007 23:59:36 +0000 (23:59 +0000)]
Add PCI device support for Intel S7000FC4UR in usb and ichwd, systems
will not install without the usb changes in the install kernel, so I
would like to MFC this in time for 7.0 RC

MFC: 3 days

16 years agoFix bad rule and bad dependency for nsparser.h that can
Ruslan Ermilov [Thu, 15 Nov 2007 22:39:15 +0000 (22:39 +0000)]
Fix bad rule and bad dependency for nsparser.h that can
cause the build to fail because y.tab.c can have a more
recent modification time than y.tab.h, and the bad rule
relied on the opposite.

(The last write to y.tab.c by yacc(1) happens after the
last write to y.tab.h, according to truss(1).)

Reported by: kensmith

16 years agoAdd support for cross double fault frames in stack traces:
John Baldwin [Thu, 15 Nov 2007 22:00:57 +0000 (22:00 +0000)]
Add support for cross double fault frames in stack traces:
- Populate the register values for the trapframe put on the stack by the
  double fault handler.
- Teach DDB's trace routine to treat a double fault like other trap frames.

MFC after: 3 days

16 years agoAcquire the process mutex and spin locks before calling thread_exit() in
John Baldwin [Thu, 15 Nov 2007 21:45:17 +0000 (21:45 +0000)]
Acquire the process mutex and spin locks before calling thread_exit() in
kthread_exit() to fix panics when using INVARIANTS.

16 years agoFix a change in the previous commit that was actually a type-o.
Scott Long [Thu, 15 Nov 2007 16:23:38 +0000 (16:23 +0000)]
Fix a change in the previous commit that was actually a type-o.

16 years agoAdd ttys lines for pts/0-pts/255.
Robert Watson [Thu, 15 Nov 2007 16:22:59 +0000 (16:22 +0000)]
Add ttys lines for pts/0-pts/255.

MFC after: 3 days

16 years ago- Adds event handlers for process_ctor,process_dtor, process_init,
Randall Stewart [Thu, 15 Nov 2007 14:20:07 +0000 (14:20 +0000)]
- Adds event handlers for process_ctor,process_dtor, process_init,
  process_fini, thread_ctor, thread_dtor, thread_init, thread_fini. This
  will allow us to extend dynamically areas in proc/thread for dtrace ;-)
Reviewed by:    rwatson

16 years agoFix build.
Gleb Smirnoff [Thu, 15 Nov 2007 14:16:20 +0000 (14:16 +0000)]
Fix build.

16 years agoAdds an event handler for:
Randall Stewart [Thu, 15 Nov 2007 13:28:54 +0000 (13:28 +0000)]
Adds an event handler for:
  - process_ctor,dtor, init and fini
  - thread_ctor,dtor, init and fini
This allows the ability to add on additional things
during construction/destruction of threads and processes.

Reviewed by: rwatson

16 years agoFix (uncritical) potential buffer overflow: the length of "buf" was not
Stefan Eßer [Thu, 15 Nov 2007 07:31:11 +0000 (07:31 +0000)]
Fix (uncritical) potential buffer overflow: the length of "buf" was not
extended when "jaiL" was added to fmt_flags().
MFC after: 1 week

16 years agoThis time REALLY copy the name from the proc to the thread as a default.
Julian Elischer [Thu, 15 Nov 2007 06:35:26 +0000 (06:35 +0000)]
This time REALLY copy the name from the proc to the thread as a default.

16 years agoBriefly document what the -c option of the "label" command does.
Ruslan Ermilov [Thu, 15 Nov 2007 06:31:18 +0000 (06:31 +0000)]
Briefly document what the -c option of the "label" command does.

MFC after: 3 days

16 years agoRevise the markup and apply some wordsmithing.
Ruslan Ermilov [Thu, 15 Nov 2007 06:23:33 +0000 (06:23 +0000)]
Revise the markup and apply some wordsmithing.

Reviewed by: pjd
MFC after: 3 days

16 years agoWhen forking, the new thread deserves a name too. Don't just use the
Julian Elischer [Thu, 15 Nov 2007 02:13:44 +0000 (02:13 +0000)]
When forking, the new thread deserves a name too. Don't just use the
td_startcopy section as it is not the right thing to do
in other cases (e.g. if starting a new thread from one that is already named).

16 years agoThis commit was generated by cvs2svn to compensate for changes in r173619,
David E. O'Brien [Wed, 14 Nov 2007 22:58:36 +0000 (22:58 +0000)]
This commit was generated by cvs2svn to compensate for changes in r173619,
which included commits to RCS files with non-trunk default branches.

16 years agoBring in FSF revision 1.9 (Change type of ps_strings from int to u_long).
David E. O'Brien [Wed, 14 Nov 2007 22:58:36 +0000 (22:58 +0000)]
Bring in FSF revision 1.9 (Change type of ps_strings from int to u_long).
This fixes a situation where GDB gives incorrect backtraces when a signal
handler is in the call chain.

Submitted by: Stu Grossman <grossman@juniper.net>

16 years agocorrect MIMO power save action frame format;
Sam Leffler [Wed, 14 Nov 2007 21:28:48 +0000 (21:28 +0000)]
correct MIMO power save action frame format;
this one really is from the D2.04 spec

MFC after: 3 days

16 years agoRemove a bogus KASSERT which will prevent rwlock to be acquired
Attilio Rao [Wed, 14 Nov 2007 21:21:48 +0000 (21:21 +0000)]
Remove a bogus KASSERT which will prevent rwlock to be acquired
recursively in exclusive mode with debugging kernels.

Submitted by: kmacy
Approved by: jeff

16 years agoAdd reload functionality.
Henrik Brix Andersen [Wed, 14 Nov 2007 21:19:15 +0000 (21:19 +0000)]
Add reload functionality.

PR: conf/116659
Approved by: sam, erwin (mentor)

16 years agoo Rename cpu_thread_setup() to cpu_thread_alloc() to better
Marcel Moolenaar [Wed, 14 Nov 2007 20:21:54 +0000 (20:21 +0000)]
o  Rename cpu_thread_setup() to cpu_thread_alloc() to better
   communicate that it relates to (is called by) thread_alloc()
o  Add cpu_thread_free() which is called from thread_free()
   to counter-act cpu_thread_alloc().

i386: Have cpu_thread_free() call cpu_thread_clean() to
preserve behaviour.
ia64: Have cpu_thread_free() call mtx_destroy() for the
mutex initialized in cpu_thread_alloc().

PR: ia64/118024

16 years agoAdd wlan_amrr to the required devices/modules
Benjamin Close [Wed, 14 Nov 2007 12:51:02 +0000 (12:51 +0000)]
Add wlan_amrr to the required devices/modules

Submitted By: glebius

16 years agoMFi386: revision 1.661
Yoshihiro Takahashi [Wed, 14 Nov 2007 12:29:09 +0000 (12:29 +0000)]
MFi386: revision 1.661

  Drastically simplify the i386 pcpu backend by merging parts of the
  amd64 mechanism over.

16 years ago- Merge all the ng_send_fn2* functions into one - ng_send_fn2(),
Gleb Smirnoff [Wed, 14 Nov 2007 11:25:58 +0000 (11:25 +0000)]
- Merge all the ng_send_fn2* functions into one - ng_send_fn2(),
  removing some copy&pasted code.
- Reduce copy and paste in ng_apply_item().
- Resurrect ng_send_fn() as a valid symbol, not a define.

Reviewed by: mav, julian

16 years agoFix build of the "package" target, ignoring non existing
Luigi Rizzo [Wed, 14 Nov 2007 07:38:22 +0000 (07:38 +0000)]
Fix build of the "package" target, ignoring non existing
floppy types (in fact, we have only one left!)
Also cleanup some code, using || and && in some places, and
using "export VAR=value" instead of two separate lines.

16 years agoA bunch more files that should probably print out a thread name
Julian Elischer [Wed, 14 Nov 2007 06:51:33 +0000 (06:51 +0000)]
A bunch more files that should probably print out a thread name
instead of a process name.

16 years agogenerally we are interested in what thread did something as
Julian Elischer [Wed, 14 Nov 2007 06:21:24 +0000 (06:21 +0000)]
generally we are interested in what thread did something as
opposed to what process. Since threads by default have teh name of the
process unless over-written with more useful information, just print the
thread name instead.

16 years agoMake sure there is a good default thread name for all threads.
Julian Elischer [Wed, 14 Nov 2007 06:04:57 +0000 (06:04 +0000)]
Make sure there is a good default thread name for all threads.

16 years agoApply the same sort of locking done in
Julian Elischer [Wed, 14 Nov 2007 05:43:55 +0000 (05:43 +0000)]
Apply the same sort of locking done in
   sys/dev/acpica/acpi.c rev 1.196 a while ago:

Grab Giant around calls to DEVICE_SUSPEND/RESUME in
acpi_SetSleepState().
If we are resuming non-MPSAFE drivers, they need Giant held for them.
This may fix some obscure suspend/resume problems.  It has fixed keyrate
setting problems that were triggered by cardbus (MPSAFE) changing the
ordering for syscons resume (non-MPSAFE).  Also, add some asserts that
Giant is held in our suspend/resume and shutdown methods.

Submitted by: Marko Zec

16 years agoFix picobsd builds. Changes include:
Luigi Rizzo [Wed, 14 Nov 2007 05:42:54 +0000 (05:42 +0000)]
Fix picobsd builds. Changes include:
- use proper make configuration for the build, using
  ${BINMAKE} as evaluated by ${VERSION}/src/Makefile
- remove -lmytinfo from crunch.conf
- remove support for login_access in tinyware/login
- remove "machine i386" from the kernel config file

It might actually be interesting to extend the script to do
a cross build for !i386 ...

MFC after: 3 days

16 years agoReserve a bit for use when capturing callchains.
Joseph Koshy [Wed, 14 Nov 2007 04:48:58 +0000 (04:48 +0000)]
Reserve a bit for use when capturing callchains.

16 years agoDrastically simplify the i386 pcpu backend by merging parts of the
Peter Wemm [Tue, 13 Nov 2007 23:00:24 +0000 (23:00 +0000)]
Drastically simplify the i386 pcpu backend by merging parts of the
amd64 mechanism over.  Instead of page table hackery that isn't
actually needed, just use 'struct pcpu __pcpu[MAXCPU]' for backing like
all the other platforms do.  Get rid of 'struct privatespace' and a
while mess of #ifdef SMP garbage that set it up.  As a bonus, this
returns the 4MB of KVA that we stole to implement it the old way.
This also allows you to read the pcpu data for each cpu when reading a
minidump.

Background information:  Originally, pcpu stuff was implemented as having
per-cpu page tables and magic to make different data structures appear
at the same actual address.  In order to share page tables, we switched
to using the GDT and %fs/%gs to access it.  But we still did the evil
magic to set it up for the old way.  The "idle stacks" are not used
for the idle process anymore and are just used for a few functions during
bootup, then ignored.  (excercise for reader: free these afterwards).

16 years agoSort.
Ruslan Ermilov [Tue, 13 Nov 2007 20:26:27 +0000 (20:26 +0000)]
Sort.

16 years agoo Do not leak inodes hash table at module unload.
Maxim Konovalov [Tue, 13 Nov 2007 19:34:06 +0000 (19:34 +0000)]
o Do not leak inodes hash table at module unload.

PR: kern/118017
Submitted by: Ighighi
MFC after: 1 week

16 years agoAdd instructions about how to upgrade less in base system
Xin LI [Tue, 13 Nov 2007 18:54:35 +0000 (18:54 +0000)]
Add instructions about how to upgrade less in base system

16 years agoXref wpi(4).
Christian Brueffer [Tue, 13 Nov 2007 16:42:22 +0000 (16:42 +0000)]
Xref wpi(4).

16 years agoCleanup, mostly typo fixes.
Christian Brueffer [Tue, 13 Nov 2007 16:40:23 +0000 (16:40 +0000)]
Cleanup, mostly typo fixes.

16 years agoDocument hardware wpi(4) supports.
Christian Brueffer [Tue, 13 Nov 2007 16:27:29 +0000 (16:27 +0000)]
Document hardware wpi(4) supports.

16 years agoRemove the IEEE80211_C_WEP capability flag - setting it makes net80211 expect
Rink Springer [Tue, 13 Nov 2007 16:12:59 +0000 (16:12 +0000)]
Remove the IEEE80211_C_WEP capability flag - setting it makes net80211 expect
that the driver will handle WEP encryption. However, this does not seem to be
implemented by this driver (or maybe the chipset doesn't support it?)

Removing the flag makes my wpi card work using wpa_supplicant(8) on a
network with 802.1x security (without this change it authenticated fine, but
tcpdump only saw garbage packets)

Reviewed by: benjsc, imp (mentor)
Approved by: imp (mentor), sam

16 years agoSplit decr_init() into two, with the section that reads the timebase
Peter Grehan [Tue, 13 Nov 2007 15:47:55 +0000 (15:47 +0000)]
Split decr_init() into two, with the section that reads the timebase
frequency from OpenFirmware moved out and into a routine that is called
from cpu_startup().

This allows correct reporting of the CPU clockspeed when printing out
CPU information at boot time.

Reported by: numerous
Reviewed by: marcel
MFC after: 1 day