]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
16 years agoRemove ISDN4BSD (I4B) from HEAD as it is not MPSAFE and
bz [Mon, 26 May 2008 10:40:09 +0000 (10:40 +0000)]
Remove ISDN4BSD (I4B) from HEAD as it is not MPSAFE and
parts relied on the now removed NET_NEEDS_GIANT.
Most of I4B has been disconnected from the build
since July 2007 in HEAD/RELENG_7.

This is what was removed:
- configuration in /etc/isdn
- examples
- man pages
- kernel configuration
- sys/i4b (drivers, layers, include files)
- user space tools
- i4b support from ppp
- further documentation

Discussed with: rwatson, re

16 years agoAdd a missing space between a variable and the ] for a test
dougb [Mon, 26 May 2008 10:10:11 +0000 (10:10 +0000)]
Add a missing space between a variable and the ] for a test

16 years agoTake into account possible overflow when multiplying. The casuality is
kib [Mon, 26 May 2008 10:01:13 +0000 (10:01 +0000)]
Take into account possible overflow when multiplying. The casuality is
the malloc call later, panicing kernel due to the oversized allocation.

Reported by: pho
Reviewed by: jeff

16 years agoDont call ata_start() when ata_reinit fails in the ioctl path.
sos [Mon, 26 May 2008 08:58:48 +0000 (08:58 +0000)]
Dont call ata_start() when ata_reinit fails in the ioctl path.

16 years agoWe can not call iwn_start directly in the interrupt
delphij [Mon, 26 May 2008 07:40:14 +0000 (07:40 +0000)]
We can not call iwn_start directly in the interrupt
context, where the iwn mutex is being held, and
iwn_start assumes that we do not have that mutex held.

Resolve this issue with what we do for other NICs by
splitting the iwn_start procedure into two parts,
iwn_start() do the locking, and iwn_start_locked()
assumes that the mutex is being held.  This resolves
panic when WITNESS is enabled.

16 years agoFix namespace collision after src/sys/sys/file.h:1.78.
pjd [Sun, 25 May 2008 22:34:17 +0000 (22:34 +0000)]
Fix namespace collision after src/sys/sys/file.h:1.78.

16 years agoBackout the change to remove sun4v.
jb [Sun, 25 May 2008 22:13:17 +0000 (22:13 +0000)]
Backout the change to remove sun4v.

Note that sun4v is currently broken on T2000 hardware so it's pretty
much a lame duck at this point.

16 years agoRemove netatm from HEAD as it is not MPSAFE and relies on the now removed
rwatson [Sun, 25 May 2008 22:11:40 +0000 (22:11 +0000)]
Remove netatm from HEAD as it is not MPSAFE and relies on the now removed
NET_NEEDS_GIANT.  netatm has been disconnected from the build for ten
months in HEAD/RELENG_7.  Specifics:

- netatm include files
- netatm command line management tools
- libatm
- ATM parts in rescue and sysinstall
- sample configuration files and documents
- kernel support as a module or in NOTES
- netgraph wrapper nodes for netatm
- ctags data for netatm.
- netatm-specific device drivers.

MFC after: 3 weeks
Reviewed by: bz
Discussed with: bms, bz, harti

16 years agoComment out the code that breaks with invariants. This is stuff that is
jb [Sun, 25 May 2008 20:24:07 +0000 (20:24 +0000)]
Comment out the code that breaks with invariants. This is stuff that is
still WIP along with the lockstat provider, so there is no harm leaving
it out for now.

16 years agoThe "if" semantic is not needed, just fix this.
attilio [Sun, 25 May 2008 16:11:27 +0000 (16:11 +0000)]
The "if" semantic is not needed, just fix this.

16 years agoReplace direct atomic operation for the file refcount witht the
attilio [Sun, 25 May 2008 14:57:43 +0000 (14:57 +0000)]
Replace direct atomic operation for the file refcount witht the
refcount interface.
It also introduces the correct usage of memory barriers, as sometimes
fdrop() and fhold() are used with shared locks, which don't use any
release barrier.

16 years agostyle fix for newly introduced macro.
attilio [Sun, 25 May 2008 14:50:47 +0000 (14:50 +0000)]
style fix for newly introduced macro.

16 years agoRemove sun4v from the list of arches in 'make universe'. There has been
jb [Sun, 25 May 2008 02:48:57 +0000 (02:48 +0000)]
Remove sun4v from the list of arches in 'make universe'. There has been
no active development on it for over a year now and it isn't
reliable under a simple buildworld. Developers can't be expected to
test code targeted for it.

16 years agoRemove the dependency on lockstat. That its still work in progress.
jb [Sun, 25 May 2008 02:04:18 +0000 (02:04 +0000)]
Remove the dependency on lockstat. That its still work in progress.

16 years agoAdd the vtime (virtual time) hooks for DTrace.
jb [Sun, 25 May 2008 01:44:58 +0000 (01:44 +0000)]
Add the vtime (virtual time) hooks for DTrace.

16 years agoTo date, our implementation of munmap(2) has required that the
alc [Sat, 24 May 2008 21:57:16 +0000 (21:57 +0000)]
To date, our implementation of munmap(2) has required that the
entirety of the specified range be mapped.  Specifically, it has
returned EINVAL if the entire range is not mapped.  There is not,
however, any basis for this in either SuSv2 or our own man page.
Moreover, neither Linux nor Solaris impose this requirement.  This
revision removes this requirement.

Submitted by: Tijl Coosemans
PR: 118510
MFC after: 6 weeks

16 years agoFix comments to replace SBSIZE with SBLOCKSIZE, since SBSIZE
rodrigc [Sat, 24 May 2008 20:44:14 +0000 (20:44 +0000)]
Fix comments to replace SBSIZE with SBLOCKSIZE, since SBSIZE
was renamed to SBLOCKSIZE in version 1.33

Reviewed by: mckusick

16 years agoRemove 'fasttrap' module for amd64 for now as the code is not there
bz [Sat, 24 May 2008 19:52:19 +0000 (19:52 +0000)]
Remove 'fasttrap' module for amd64 for now as the code is not there
in sys/cddl/dev/fasttrap.

16 years agoRemove redundant redeclaration of 'zone_drain'.
bz [Sat, 24 May 2008 19:30:38 +0000 (19:30 +0000)]
Remove redundant redeclaration of 'zone_drain'.

16 years agoRestore buildable state. Style ignored.
bz [Sat, 24 May 2008 19:29:02 +0000 (19:29 +0000)]
Restore buildable state. Style ignored.
Leave IDTVEC(ill) where it was unless we compile with KDTRACE_HOOKS[1].
Hide the with DTRACE case case under #ifdef KDTRACE_HOOKS.

Suggested by: attilio [1]
Reviewed by: attilio

16 years agoChange my email address to the one from the FreeBSD project.
ed [Sat, 24 May 2008 18:35:55 +0000 (18:35 +0000)]
Change my email address to the one from the FreeBSD project.

Approved by: philip (mentor)

16 years agoIn addition to the ipsec_osdep.h removal a week ago, now also eliminate
bz [Sat, 24 May 2008 15:32:46 +0000 (15:32 +0000)]
In addition to the ipsec_osdep.h removal a week ago, now also eliminate
IPSEC_SPLASSERT_SOFTNET which has been 'unused' since FreeBSD 5.0.

16 years agoFactor out the v4-only vs. the v6-only inp_flags processing in
bz [Sat, 24 May 2008 15:20:48 +0000 (15:20 +0000)]
Factor out the v4-only vs. the v6-only inp_flags processing in
ip6_savecontrol in preparation for udp_append() to no longer
need an WLOCK as we will no longer be modifying socket options.

Requested by: rwatson
Reviewed by: gnn
MFC after: 10 days

16 years ago- Add locking to all filesystem operations in fdescfs and flag it as MPSAFE.
lulf [Sat, 24 May 2008 14:51:30 +0000 (14:51 +0000)]
- Add locking to all filesystem operations in fdescfs and flag it as MPSAFE.
- Use proper synhronization primitives to protect the internal fdesc node cache
  used in fdescfs.
- Properly initialize and uninitalize hash.
- Remove unused functions.

Since fdescfs might recurse on itself, adding proper locking to it needed some
tricky workarounds in some parts to make it work. For instance, a descriptor in
fdescfs could refer to an open descriptor to itself, thus forcing the thread to
recurse on vnode locks. Because of this, other race conditions also had to be
fixed.

Tested by: pho
Reviewed by: kib (mentor)
Approved by: kib (mentor)

16 years ago Add myself to the calendar
gonzo [Sat, 24 May 2008 11:10:46 +0000 (11:10 +0000)]
  Add myself to the calendar

  Approved by: cognet (mentor)

16 years agoMFi386: revision 1.249
nyan [Sat, 24 May 2008 09:07:52 +0000 (09:07 +0000)]
MFi386: revision 1.249

  Add a cyclic hook for DTrace.

16 years agoAdd myself to calendar.freebsd
manolis [Sat, 24 May 2008 08:54:00 +0000 (08:54 +0000)]
Add myself to calendar.freebsd
This will also help me not forget my own birthday :)

Approved by: gabor (mentor)

16 years agoAdd myself to commiters-doc.dot
manolis [Sat, 24 May 2008 08:49:55 +0000 (08:49 +0000)]
Add myself to commiters-doc.dot

Approved by: gabor (mentor)

16 years agoMake the zfs module depend on the opensolaris module in preparation for it
jb [Sat, 24 May 2008 06:43:55 +0000 (06:43 +0000)]
Make the zfs module depend on the opensolaris module in preparation for it
to shared stuff with the DTrace modules.

16 years agoAdd the DTrace hooks for exception handling (Function boundary trace
jb [Sat, 24 May 2008 06:32:26 +0000 (06:32 +0000)]
Add the DTrace hooks for exception handling (Function boundary trace
-fbt- provider), cyclic clock and syscalls.

16 years agoAdd a cyclic hook for DTrace.
jb [Sat, 24 May 2008 06:27:54 +0000 (06:27 +0000)]
Add a cyclic hook for DTrace.

16 years agoAdd the DTrace hooks for exception handling (Function boundary trace
jb [Sat, 24 May 2008 06:27:02 +0000 (06:27 +0000)]
Add the DTrace hooks for exception handling (Function boundary trace
-fbt- provider), cyclic clock and syscalls.

16 years agoAdd DTrace 'proc' provider probes using the Statically Defined Trace
jb [Sat, 24 May 2008 06:22:16 +0000 (06:22 +0000)]
Add DTrace 'proc' provider probes using the Statically Defined Trace
(sdt) mechanism.

16 years agoExpand dump to allow MAX_INT dump levels.
mckusick [Sat, 24 May 2008 05:20:46 +0000 (05:20 +0000)]
Expand dump to allow MAX_INT dump levels.

PR:           bin/100732
Submitted by: Matthew Vincenz <msvincen@midway.uchicago.edu>

16 years agoLeave out a couple of DTrace modules for now.
jb [Sat, 24 May 2008 04:36:07 +0000 (04:36 +0000)]
Leave out a couple of DTrace modules for now.

16 years agoAfter converting the "snapshot" mount option to the MNT_SNAPSHOT flag,
rodrigc [Sat, 24 May 2008 00:41:32 +0000 (00:41 +0000)]
After converting the "snapshot" mount option to the MNT_SNAPSHOT flag,
delete "snapshot" from the persistent mount options list.
This should fix problems with doing a mount -o snapshot of a file system, followed by
an NFS export of the same file system.

PR: 122833
Reported by: Leon Kos <leon.kos lecad fs uni-lj si>,
Jaakko Heinonen <jh saunalahti fi>
MFC after: 1 month

16 years agoFor the following mount options, do not perform the string to flag conversions
rodrigc [Sat, 24 May 2008 00:02:12 +0000 (00:02 +0000)]
For the following mount options, do not perform the string to flag conversions
here, because we already do them further up in vfs_donmount() in vfs_mount.c

async -> MNT_ASYNC
force -> MNT_FORCE
multilabel -> MNT_MULTILABEL
noatime -> MNT_NOATIME
noclusterr -> MNT_NOCLUSTERR
noclusterw -> MNT_NOCLUSTERW

MFC after:  1 month

16 years agoDo not convert the "snapshot" string to the MNT_SNAPSHOT flag here, since
rodrigc [Fri, 23 May 2008 23:33:07 +0000 (23:33 +0000)]
Do not convert the "snapshot" string to the MNT_SNAPSHOT flag here, since
we do it further down in ffs_vfsops.c

MFC after: 1 month

16 years agoWhen using dump to generate level 0 dumps which are then rsync'ed
mckusick [Fri, 23 May 2008 23:13:14 +0000 (23:13 +0000)]
When using dump to generate level 0 dumps which are then rsync'ed
to a remote machine, the fact that the dump date is stored with
each header (inode) record makes rsync significantly less efficient
than necessary. This also applies to inode access times when they
are not important data to retain. When implementing an offsite
backup solution of this type, these dates in particular are not
important, especially if it prevents effective offsite backups.

PR:           bin/91049
Submitted by: Forrest W Christian <fwc@mt.net>

16 years agoMessing with the endian defines breaks the use of other FreeBSD headers.
jb [Fri, 23 May 2008 23:03:17 +0000 (23:03 +0000)]
Messing with the endian defines breaks the use of other FreeBSD headers.

16 years agoDon't need to include vmem.h anymore.
jb [Fri, 23 May 2008 22:44:46 +0000 (22:44 +0000)]
Don't need to include vmem.h anymore.

16 years agoDelete a couple of OpenSolaris headers which get in the way of our
jb [Fri, 23 May 2008 22:40:58 +0000 (22:40 +0000)]
Delete a couple of OpenSolaris headers which get in the way of our
implementation.

16 years agoOpenSolaris kernel module compatibility sources.
jb [Fri, 23 May 2008 22:39:28 +0000 (22:39 +0000)]
OpenSolaris kernel module compatibility sources.

16 years agoHook the DTrace kernel modules into the build.
jb [Fri, 23 May 2008 22:36:12 +0000 (22:36 +0000)]
Hook the DTrace kernel modules into the build.

16 years agoThe cyclic code is in sys/cddl/dev/cyclic to be consistent with
jb [Fri, 23 May 2008 22:23:28 +0000 (22:23 +0000)]
The cyclic code is in sys/cddl/dev/cyclic to be consistent with
the other stuff.

16 years agoThe cyclic timer device. This is a cut down version of the one in
jb [Fri, 23 May 2008 22:21:58 +0000 (22:21 +0000)]
The cyclic timer device. This is a cut down version of the one in
OpenSolaris. We don't have the lock levels that they do, so this is just
hooked into clock interrupts.

16 years agoAdd the KDTRACE_HOOKS option for DTrace support.
jb [Fri, 23 May 2008 22:17:28 +0000 (22:17 +0000)]
Add the KDTRACE_HOOKS option for DTrace support.

16 years agoAccount for IPI_PREEMPT. We don't want to call sched_preempt() with
marcel [Fri, 23 May 2008 19:53:50 +0000 (19:53 +0000)]
Account for IPI_PREEMPT. We don't want to call sched_preempt() with
interrupts disabled or with td_intr_nesting_level non-zero.

16 years agoFix nits pointed out in PR bin/39905 that have not already been
mckusick [Fri, 23 May 2008 19:17:08 +0000 (19:17 +0000)]
Fix nits pointed out in PR bin/39905 that have not already been
corrected since it was filed. With this change the PR will be closed.

PR:     bin/39905

16 years agoInvalidate the TLB in pmap_cpu_bootstrap(), so that it also happens
marcel [Fri, 23 May 2008 19:16:24 +0000 (19:16 +0000)]
Invalidate the TLB in pmap_cpu_bootstrap(), so that it also happens
on the APs.

16 years agoFix the build by including <sys/param.h> in <sys/clist.h>.
ed [Fri, 23 May 2008 18:34:33 +0000 (18:34 +0000)]
Fix the build by including <sys/param.h> in <sys/clist.h>.

Of course I was silly enough to only check LINT for build failures, but not
the userspace bits. In the mpsafetty branch I didn't notice this, because
<sys/clist.h> never got included in userspace.

Approved by: philip (mentor)
Pointy hat to: me :-(

16 years agoRev. 1.274 put the ttyrel() call before the destroy_dev() in the
kib [Fri, 23 May 2008 16:47:55 +0000 (16:47 +0000)]
Rev. 1.274 put the ttyrel() call before the destroy_dev() in the
ttyfree(), freeing the tty. Since destroy_dev() may call d_purge()
cdevsw method, that is the ttypurge() for the tty, the code ends up
accessing freed tty structure.

Put the ttyrel() after destroy_dev() in the ttyfree. To prevent the
panic the rev. 1.274 provided fix for, check the TS_GONE in sysctl
handler and refuse to provide information on such tty.

Reported, debugging help and tested by: pho
DIscussed with and reviewed by: jhb
MFC after: 1 week

16 years agoThe dev_refthread() in the tty_gettp() may fail, because Giant is taken
kib [Fri, 23 May 2008 16:46:27 +0000 (16:46 +0000)]
The dev_refthread() in the tty_gettp() may fail, because Giant is taken
in the giant_trick routines after the dev_refthread increments the
si_threadcount. Remove assert, do not perform dev_relthread() for failed
dev_refthread(), and handle failure in the tty_gettp() callers (cdevsw
tty methods).

Before kern_conf.c 1.210 and 1.211, the kernel usually paniced in the
giant_trick routines dereferencing NULL cdevsw, not taking this fault.

Reported by: Vince Hoffman <jhary unsane co uk>
Debugging help and tested by: pho
Reviewed by: jhb
MFC after: 1 week

16 years agoUse the t_state for the TS_GONE test.
kib [Fri, 23 May 2008 16:43:59 +0000 (16:43 +0000)]
Use the t_state for the TS_GONE test.

Submitted by:   jhb
MFC after: 3 days

16 years agoAssert that si_threadcount > 0 before decrementing it. This helps catching
kib [Fri, 23 May 2008 16:38:38 +0000 (16:38 +0000)]
Assert that si_threadcount > 0 before decrementing it. This helps catching
the improper use of the dev_refthread/dev_relthread.

Tested by: pho
MFC after: 1 week

16 years agoWhen vget() fails (because the vnode has been reclaimed), there is no
kib [Fri, 23 May 2008 16:36:39 +0000 (16:36 +0000)]
When vget() fails (because the vnode has been reclaimed), there is no
sense to loop trying to vget() the vnode again.

PR: 122977
Submitted by: Arthur Hartwig <arthur.hartwig nokia com>
Tested by: pho
Reviewed by: jhb
MFC after: 1 week

16 years agoMove TTY unrelated bits out of <sys/tty.h>.
ed [Fri, 23 May 2008 16:06:35 +0000 (16:06 +0000)]
Move TTY unrelated bits out of <sys/tty.h>.

For some reason, the <sys/tty.h> header file also contains routines of the
clists and console that are used inside the TTY layer. Because the clists
are not only used by the TTY layer (example: various input drivers), we'd
better move the entire clist programming interface into <sys/clist.h>. Also
remove a declaration of nonexistent variable.

The <sys/tty.h> header also contains various definitions for the console
code (tty_cons.c). Also move these to <sys/cons.h>, because they are
not implemented inside the TTY layer.

While there, create separate malloc pools for the clist and console code.

Approved by: philip (mentor)

16 years agoMove sysinstall/sade away from TIOCGSIZE.
ed [Fri, 23 May 2008 14:24:33 +0000 (14:24 +0000)]
Move sysinstall/sade away from TIOCGSIZE.

Both sysinstall and sade still seem to use the TIOCGSIZE ioctl to obtain
the terminal dimensions. We'd better use TIOCGWINSZ to do this. The
TIOCGWINSZ interface is preferred, because it also allows sizes in pixels
to be passed to the application (though this is not used here).

Approved by: philip (mentor)

16 years agoAnother problem caused by the knlist_cleardel() potentially dropping
kib [Fri, 23 May 2008 11:14:03 +0000 (11:14 +0000)]
Another problem caused by the knlist_cleardel() potentially dropping
PIPE_MTX().

Since the pipe_present is cleared before (potentially) sleeping, the
second thread may enter the pipeclose() for the reciprocal pipe end.
The test at the end of the pipeclose() for the pipe_present == 0 would
succeed, allowing the second thread to free the pipe memory. First
threads then accesses the freed memory after being woken up.

Properly track the closing state of the pipe in the pipe_present.
Introduce the intermediate state that marks the pipe as mostly
dismantled but might be sleeping waiting for the knote list to be
cleared. Free the pipe pair memory only when both ends pass that point.

Debugging help and tested by: pho
Discussed with: jmg
MFC after: 2 weeks

16 years agoDestruction of the pipe calls knlist_cleardel() to remove the knotes
kib [Fri, 23 May 2008 11:09:50 +0000 (11:09 +0000)]
Destruction of the pipe calls knlist_cleardel() to remove the knotes
monitoring the pipe. The code sets pipe_present = 0 and enters
knlist_cleardel(), where the PIPE_MTX might be dropped when knl->kl_list
cannot be cleared due to influx knotes.

If the following often encountered code fragment
                if (!(kn->kn_status & KN_DETACHED))
                        kn->kn_fop->f_detach(kn);
                knote_drop(kn, td); [1]
is executed while the knlist lock is dropped, then the knote memory is freed
by the knote_drop() without knote being removed from the knlist, since
the filt_pipedetach() contains the following:
        if (kn->kn_filter == EVFILT_WRITE) {
                if (!cpipe->pipe_peer->pipe_present) {
                        PIPE_UNLOCK(cpipe);
                        return;

Now, the memory may be reused in the zone, causing the access to the
freed memory. I got the panics caused by the marker knote appearing on
the knlist, that, I believe, manifestation of the issue. In the Peter
Holm test scenarious, we got unkillable processes too.

The pipe_peer that has the knote for write shall be present. Ignore the
pipe_present value for EVFILT_WRITE in filt_pipedetach().

Debugging help and tested by: pho
Discussed with: jmg
MFC after: 2 weeks

16 years agoRelax the strict type check because gcc as distributed doesn't create
jb [Fri, 23 May 2008 07:35:36 +0000 (07:35 +0000)]
Relax the strict type check because gcc as distributed doesn't create
debug strtabs with type SHT_STRTAB. Although we could change FreeBSD's
gcc, we really need to play nicely with gcc as distributed by the FSF.

16 years agoChange the alignment of the NOTE to match what gcc does.
jb [Fri, 23 May 2008 07:33:45 +0000 (07:33 +0000)]
Change the alignment of the NOTE to match what gcc does.

16 years agoAdd the case for SHT_AMD64_UNWIND.
jb [Fri, 23 May 2008 07:32:19 +0000 (07:32 +0000)]
Add the case for SHT_AMD64_UNWIND.

16 years agoAdd the ctf_get function and update the args to linker_file_function_listall.
jb [Fri, 23 May 2008 07:08:59 +0000 (07:08 +0000)]
Add the ctf_get function and update the args to linker_file_function_listall.

16 years agoCustom DTrace kernel module files plus FreeBSD-specific DTrace providers.
jb [Fri, 23 May 2008 05:59:42 +0000 (05:59 +0000)]
Custom DTrace kernel module files plus FreeBSD-specific DTrace providers.

16 years agoThe value le->name cannot be NULL when we're freeing an entry in the
cperciva [Fri, 23 May 2008 05:07:22 +0000 (05:07 +0000)]
The value le->name cannot be NULL when we're freeing an entry in the
hardlink table for two reasons: 1. If le->name is set to NULL, the
structure le won't be inserted into the table; 2. Even if le somehow
did manage to get into the table with le->name equal to NULL, we would
die when we dereferenced le->null before we could get to the point of
freeing the entry.

Remove the unnecessary "if (le->name != NULL)" test and just free the
pointer.

Found by: Coverity Prevent

16 years agoCheck that lseek(2) succeeds and puts us where we expect. [1]
cperciva [Fri, 23 May 2008 05:01:29 +0000 (05:01 +0000)]
Check that lseek(2) succeeds and puts us where we expect. [1]

While we're here, fix a long-standing bug in the handling of write(2)
errors: The API changed from "return # of bytes written" to "return
status code" almost 4 years ago, so instead of returning (-1) we need
to return ARCHIVE_FATAL.

Found by: Coverity Prevent [1]

16 years agoWe only use the string "?rwxrwxrwx " once, so inline it rather than
cperciva [Fri, 23 May 2008 04:57:28 +0000 (04:57 +0000)]
We only use the string "?rwxrwxrwx " once, so inline it rather than
declaring a variable which points to it.  Aside from eliminating a
line of code and one level of unnecessary indirection, this eliminates
a false positive in Coverity.

16 years agoAdd the DTrace build tools to the list of tools and pass NO_CTF to all
jb [Fri, 23 May 2008 04:22:14 +0000 (04:22 +0000)]
Add the DTrace build tools to the list of tools and pass NO_CTF to all
the stages of the build which don't require CTF conversion.

16 years agoAdd a knob to allow just the kernels to be built during a 'make universe'.
jb [Fri, 23 May 2008 04:20:59 +0000 (04:20 +0000)]
Add a knob to allow just the kernels to be built during a 'make universe'.
This is useful when wanting to retry a broken build using NOCLEAN and the
buildworlds are known to have completed.

16 years agoAdd the ctf_get method.
jb [Fri, 23 May 2008 04:06:49 +0000 (04:06 +0000)]
Add the ctf_get method.

16 years agoAllow a rendezvous with just a specified CPU too.
jb [Fri, 23 May 2008 04:05:26 +0000 (04:05 +0000)]
Allow a rendezvous with just a specified CPU too.

Make the API work in the non-smp case too so that a kernel module
can work the same regardless of whether or not it is loaded on a SMP
kernel or not.

16 years agoThe VM system no longer uses setPQL2(). Remove it and its helpers.
alc [Fri, 23 May 2008 04:03:54 +0000 (04:03 +0000)]
The VM system no longer uses setPQL2().  Remove it and its helpers.

16 years agoDTrace can enter the debugger from a probe. Add a WHY definition for this.
jb [Fri, 23 May 2008 04:00:44 +0000 (04:00 +0000)]
DTrace can enter the debugger from a probe. Add a WHY definition for this.

16 years agoAdd the CTF conversion to the generated makefile. In the case where
jb [Fri, 23 May 2008 03:55:26 +0000 (03:55 +0000)]
Add the CTF conversion to the generated makefile. In the case where
NO_CTF or !WITH_CTF, the macro is empty.

16 years agoAdd support for generating CTF data for the kernel.
jb [Fri, 23 May 2008 03:53:49 +0000 (03:53 +0000)]
Add support for generating CTF data for the kernel.

16 years agoAdd a kernel option for amd64 to compile with the frame on the stack
jb [Fri, 23 May 2008 03:52:55 +0000 (03:52 +0000)]
Add a kernel option for amd64 to compile with the frame on the stack
so that the DTrace Function Bounadry Trace (fbt) provider can get
coverage of most functions in the kernel.

16 years agoAdd the CTF source file which gets shared with link_elf.c and link_elf_obj.c.
jb [Fri, 23 May 2008 03:04:27 +0000 (03:04 +0000)]
Add the CTF source file which gets shared with link_elf.c and link_elf_obj.c.

16 years agoAdd hooks for the Compact C Type Format (CTF) data to be attached to
jb [Fri, 23 May 2008 00:49:39 +0000 (00:49 +0000)]
Add hooks for the Compact C Type Format (CTF) data to be attached to
the elf files. This is complicated by the fact that the actual CTF
parsing has to be done in CDDL'd code, so the BSD licensed code only
knows about the opaque data which it must be able to free.

16 years agoAdd support for the DTrace malloc provider which can enable probes
jb [Fri, 23 May 2008 00:43:36 +0000 (00:43 +0000)]
Add support for the DTrace malloc provider which can enable probes
on a per-malloc type basis.

16 years agoAdd the dtrace client app to the build.
jb [Thu, 22 May 2008 23:06:19 +0000 (23:06 +0000)]
Add the dtrace client app to the build.

16 years agospecial-case broadcast probe request in ieee80211_find_rxnode; we need
sam [Thu, 22 May 2008 22:22:10 +0000 (22:22 +0000)]
special-case broadcast probe request in ieee80211_find_rxnode; we need
to spam all vaps and this won't happen if the frame comes from a station
that is associated to an ap vap (and so has an entry in the table)

Noticed by: Jared Go
Reviewed by: thompsa

16 years agoThis fixes the "getfile: lost data" panic when restoring dumps
mckusick [Thu, 22 May 2008 22:19:33 +0000 (22:19 +0000)]
This fixes the "getfile: lost data" panic when restoring dumps
on a 7.0 or later system that were created on a pre-5.0 system.
We must ensure that restore zeros out the previously undefined
birthtime and external attribute size fields when reading dump
tapes made by the UFS1 dump program.

The problem is that UFS2 dump carefully zeros out the unused
birthtime and external attribute size fields in the dump header
when dumping UFS1 filesystems, but the UFS1 dump didn't know about
those fields (they were spares) so just left whatever random junk
was in them. So, when restoring one of these pre-UFS2 dumps,
the new restore would eventually trip across a header that had
a non-zero external attribute size and try to extract it. That
consumed several tape blocks which left it totally out of sync
and very unhappy (i.e., the panic). The fix is in the gethead()
function which modernizes old headers by copying old fields to
their new location (and with this fix) zeroing out previously
undefined fields.

PR: bin/120881
Review by: David Malone & Scott Lambert
MFC after: 1 week

16 years agoFollow on to fix 1.51 for "Header with wrong dumpdate" message.
mckusick [Thu, 22 May 2008 22:18:38 +0000 (22:18 +0000)]
Follow on to fix 1.51 for "Header with wrong dumpdate" message.

Must ensure that dump tapes from UFS1 filesystems properly copy
old fields of dump headers to new locations. Move check of dumpdate
to follow the code which ensures that the appropriate fields have
been copied.

PR: bin/118087
Help from: David Malone, Scott Lambert, Javier Martín Rueda
MFC after: 2 weeks

16 years agoignore sw beacon miss while scanning, just reset the beacon count and timer
sam [Thu, 22 May 2008 22:17:27 +0000 (22:17 +0000)]
ignore sw beacon miss while scanning, just reset the beacon count and timer

Reviewed by: thompsa

16 years agono need to stop the sw beacon miss timer; it's not used with adhoc or wds
sam [Thu, 22 May 2008 22:14:58 +0000 (22:14 +0000)]
no need to stop the sw beacon miss timer; it's not used with adhoc or wds

16 years agoclear out crud
sam [Thu, 22 May 2008 21:53:15 +0000 (21:53 +0000)]
clear out crud

Submitted by: "Pawel Worach" <pawel.worach@gmail.com>

16 years ago Clean out all mentions of struct proc* for it was removed from
gonzo [Thu, 22 May 2008 21:41:19 +0000 (21:41 +0000)]
  Clean out all mentions of struct proc* for it was removed from
  taskqueue(9) API more then two years ago.

  Approved by: cognet (mentor)

16 years agoImprove portability via s/struct siginfo/struct siginfo_data/. This was
cperciva [Thu, 22 May 2008 21:08:36 +0000 (21:08 +0000)]
Improve portability via s/struct siginfo/struct siginfo_data/.  This was
running into a namespace collision on an avian operating system.

16 years agoo Document two new ALT_BREAK_TO_DEBUGGER key sequences.
maxim [Thu, 22 May 2008 18:19:49 +0000 (18:19 +0000)]
o Document two new ALT_BREAK_TO_DEBUGGER key sequences.

16 years agoThe function pmap_addr_hint no longer exists. Remove its man page.
alc [Thu, 22 May 2008 16:25:25 +0000 (16:25 +0000)]
The function pmap_addr_hint no longer exists.  Remove its man page.

16 years agoIntroduce REMOVE_NEXT() macro's for SLIST and STAILQ.
ed [Thu, 22 May 2008 14:40:03 +0000 (14:40 +0000)]
Introduce REMOVE_NEXT() macro's for SLIST and STAILQ.

Even though single linked lists allow items to be removed at constant time
(when the previous element is known), the queue macro's don't allow this.
Implement new REMOVE_NEXT() macro's. Because the REMOVE() macro's also
contain the same code, make it call REMOVE_NEXT().

The OpenBSD version of SLIST_REMOVE_NEXT() needs a reference to the list
head, even though it is unused. We'd better mimic this. The STAILQ version
also needs a reference to the list. This means the prototypes of both
macro's are the same.

Approved by: philip (mentor)
PR: kern/121117

16 years agoAdd a new group definition to sys/conf.h: GID_TTY.
ed [Thu, 22 May 2008 13:46:41 +0000 (13:46 +0000)]
Add a new group definition to sys/conf.h: GID_TTY.

Our current TTY layer uses a set-uid application called ptchown to
change ownership of a PTY slave device. The new TTY layer implements
this functionality through a new ioctl().

By accident I discovered Darwin's TTY layer also uses this approach.
Because of this, they also have a GID_TTY.

Approved by: philip (mentor)

16 years agoLast but not least, add myself to the list of birthdays as well.
ed [Thu, 22 May 2008 13:21:05 +0000 (13:21 +0000)]
Last but not least, add myself to the list of birthdays as well.

Approved by: philip (mentor)

16 years agoAdd myself to this list as well.
ed [Thu, 22 May 2008 13:18:09 +0000 (13:18 +0000)]
Add myself to this list as well.

Approved by: philip (mentor)

16 years ago- Recognize the 'volume' parameter when creating a plex.
lulf [Thu, 22 May 2008 10:27:03 +0000 (10:27 +0000)]
- Recognize the 'volume' parameter when creating a plex.

PR: kern/75632
Approved by: pjd (mentor)
MFC after: 1 day

16 years agoAdd codec id for Sigmatel STAC9872AK
kevlo [Thu, 22 May 2008 09:14:12 +0000 (09:14 +0000)]
Add codec id for Sigmatel STAC9872AK

Reviewed by: ariff

16 years agoA 'special' compatibility header to plug OpenSolaris code.
jb [Thu, 22 May 2008 09:08:41 +0000 (09:08 +0000)]
A 'special' compatibility header to plug OpenSolaris code.

16 years agoAdditional compatibility headers.
jb [Thu, 22 May 2008 08:35:03 +0000 (08:35 +0000)]
Additional compatibility headers.

16 years agoCompatibility stuff for DTrace.
jb [Thu, 22 May 2008 08:33:24 +0000 (08:33 +0000)]
Compatibility stuff for DTrace.