]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoChange the second (and last) argument of cpu_set_upcall(). Previously
Marcel Moolenaar [Wed, 4 Jun 2003 21:13:21 +0000 (21:13 +0000)]
Change the second (and last) argument of cpu_set_upcall(). Previously
we were passing in a void* representing the PCB of the parent thread.
Now we pass a pointer to the parent thread itself.
The prime reason for this change is to allow cpu_set_upcall() to copy
(parts of) the trapframe instead of having it done in MI code in each
caller of cpu_set_upcall(). Copying the trapframe cannot always be
done with a simply bcopy() or may not always be optimal that way. On
ia64 specifically the trapframe contains information that is specific
to an entry into the kernel and can only be used by the corresponding
exit from the kernel. A trapframe copied verbatim from another frame
is in most cases useless without some additional normalization.

Note that this change removes the assignment to td->td_frame in some
implementations of cpu_set_upcall(). The assignment is redundant.
A previous call to cpu_thread_setup() already did the exact same
assignment. An added benefit of removing the redundant assignment is
that we can now change td_pcb without nasty side-effects.

This change officially marks the ability on ia64 for 1:1 threading.

Not tested on: amd64, powerpc
Compile & boot tested on: alpha, sparc64
Functionally tested on: i386, ia64

21 years agoBatton down the hatches!!!
John Baldwin [Wed, 4 Jun 2003 21:10:15 +0000 (21:10 +0000)]
Batton down the hatches!!!

Always route PCI interrupts on i386 UP machines.  I was planning to enable
this for i386 anyways once SMP support is done.  Having this enabled fixes
problems on many people's laptops.

Requested by: imp

21 years ago - Add further vm object locking to vm_object_deallocate(), specifically,
Alan Cox [Wed, 4 Jun 2003 21:07:42 +0000 (21:07 +0000)]
 - Add further vm object locking to vm_object_deallocate(), specifically,
   for accessing a vm object's shadows.

21 years agoMake this compile without warnings on 64-bit architectures:
Marcel Moolenaar [Wed, 4 Jun 2003 20:38:54 +0000 (20:38 +0000)]
Make this compile without warnings on 64-bit architectures:
Don't cast thread_self() to int. Cast to uintptr_t. Pull in the
predefined printf format from <inttypes.h>

21 years agoMake this compile without warnings on 64-bit architectures:
Marcel Moolenaar [Wed, 4 Jun 2003 20:36:08 +0000 (20:36 +0000)]
Make this compile without warnings on 64-bit architectures:
In recurse(), cast the pointer difference to int to match the format.
The difference is expected to fit in an int.

21 years ago - Add VM_OBJECT_TRYLOCK().
Alan Cox [Wed, 4 Jun 2003 19:59:23 +0000 (19:59 +0000)]
 - Add VM_OBJECT_TRYLOCK().

21 years agoGRRR. Do not force Dangerously Dedicated mode on amd64 when you select
Peter Wemm [Wed, 4 Jun 2003 19:28:39 +0000 (19:28 +0000)]
GRRR.  Do not force Dangerously Dedicated mode on amd64 when you select
'use entire disk'.  Neither for ia64 while I'm here - it needs a MBR if
its going to use fdisk+disklabel.  The ia64 case is mostly academic though
because you'd be creating two partitions (dos + freebsd) rather than
a single freebsd-only partition.

21 years agoCorrect a spelling error.
Ceri Davies [Wed, 4 Jun 2003 19:24:24 +0000 (19:24 +0000)]
Correct a spelling error.

21 years agoIntroduce g_provider_by_name() function, and use it.
Poul-Henning Kamp [Wed, 4 Jun 2003 18:17:52 +0000 (18:17 +0000)]
Introduce g_provider_by_name() function, and use it.

21 years agoMake this a true GEOM class:
Poul-Henning Kamp [Wed, 4 Jun 2003 18:16:09 +0000 (18:16 +0000)]
Make this a true GEOM class:
        Attach to the component devices using GEOM semantics.
        Create a GEOM provider instead of using disk_create()
        Use the GEOM OAM api for configuration.

I saw approx ~1% speedup in througput and ~7% in latency in a
simple minded test of a two-disk striped device.

This file was repo-copied from src/sys/dev/ccd/ccd.c.

This is not yet linked into the build.

21 years agodetach RNG test facility on device detach
Sam Leffler [Wed, 4 Jun 2003 17:58:20 +0000 (17:58 +0000)]
detach RNG test facility on device detach

21 years agoremove bogon (undefined struct decl)
Sam Leffler [Wed, 4 Jun 2003 17:57:52 +0000 (17:57 +0000)]
remove bogon (undefined struct decl)

21 years agoAdd a KASSERT to prevent the same GEOM class from being processed loaded
Poul-Henning Kamp [Wed, 4 Jun 2003 17:51:10 +0000 (17:51 +0000)]
Add a KASSERT to prevent the same GEOM class from being processed loaded
twice.

Enforce that classes should have different names while we are here.

21 years agoDrop this MAINTAINER bit. I'll reclaim an "Advisory Maintainership"
Mark Murray [Wed, 4 Jun 2003 16:10:20 +0000 (16:10 +0000)]
Drop this MAINTAINER bit. I'll reclaim an "Advisory Maintainership"
for this area later.

21 years agoRemove this maintainer bit. Another version will be added a
Mark Murray [Wed, 4 Jun 2003 16:05:38 +0000 (16:05 +0000)]
Remove this maintainer bit. Another version will be added a
bit later.

21 years agoDrop stale maintainer bits.
Mark Murray [Wed, 4 Jun 2003 16:02:17 +0000 (16:02 +0000)]
Drop stale maintainer bits.

21 years agoDrop stale MAINTAINER bit.
Mark Murray [Wed, 4 Jun 2003 16:01:37 +0000 (16:01 +0000)]
Drop stale MAINTAINER bit.

21 years agoDrop MAINTAINER Bit. Not needed any more.
Mark Murray [Wed, 4 Jun 2003 15:59:13 +0000 (15:59 +0000)]
Drop MAINTAINER Bit. Not needed any more.

21 years agoDrop MAINTAINER bits from here.
Mark Murray [Wed, 4 Jun 2003 15:58:03 +0000 (15:58 +0000)]
Drop MAINTAINER bits from here.

21 years agoMake the internal statistics sysctl an array of integers rather than a string.
Hartmut Brandt [Wed, 4 Jun 2003 15:38:32 +0000 (15:38 +0000)]
Make the internal statistics sysctl an array of integers rather than a string.
While a string is readable without a tool, an array is easier to process for
a monitoring application. This also prevents the extra hoops we need with
sbufs and locking.

Move the mtx_init() in en_attach() higher before the first failure point so
that we can unconditionally destroy it in en_destroy().

21 years agoUpdate some library names. Libraries come, libraries go.
Mark Murray [Wed, 4 Jun 2003 15:36:57 +0000 (15:36 +0000)]
Update some library names. Libraries come, libraries go.

21 years agoFix substitution when the regex matches the zero-length string.
Tony Finch [Wed, 4 Jun 2003 15:31:55 +0000 (15:31 +0000)]
Fix substitution when the regex matches the zero-length string.
There are two bugs: in the s///g case, the substitution didn't occur
at the end of the line; in the s///N case, the code didn't count
forwards along the line properly. See the sg, s3, s4, and s5 tests
in src/tools/regression/usr.bin/sed/.

Reviewed by: tjr

21 years agoI'm now happy that this is no longer needed. Libcrypto has
Mark Murray [Wed, 4 Jun 2003 15:26:34 +0000 (15:26 +0000)]
I'm now happy that this is no longer needed. Libcrypto has
all its functionality, and all its consumers have been converted.

21 years agoAdd tests for the behaviour of substitution when the regex can match
Tony Finch [Wed, 4 Jun 2003 15:26:23 +0000 (15:26 +0000)]
Add tests for the behaviour of substitution when the regex can match
a zero-length string. This shows bugs in the s///g and s///2+ cases.

21 years agoOnly init _thread_sigact once, needn't init it again after a fork().
David Xu [Wed, 4 Jun 2003 12:40:21 +0000 (12:40 +0000)]
Only init _thread_sigact once, needn't init it again after a fork().

Obtained from: deischen

21 years agoDespite whether threaded mode is turned on, always save thread's
David Xu [Wed, 4 Jun 2003 12:38:21 +0000 (12:38 +0000)]
Despite whether threaded mode is turned on, always save thread's
signal mask.

21 years agoRemoved garbage:
Bruce Evans [Wed, 4 Jun 2003 11:21:18 +0000 (11:21 +0000)]
Removed garbage:
- -elf in CFLAGS had no effect except to reduce portability.
- -elf in LDFLAGS had even less effect, since LDFLAGS is not used.
- -Wall in CFLAGS had no effect except to reduce portability and break
  overriding of WARNS, since the setting of WARNS implies -Wall.

21 years agoMove unifdef to the original section of the file.
Tony Finch [Wed, 4 Jun 2003 10:04:12 +0000 (10:04 +0000)]
Move unifdef to the original section of the file.

21 years agoCompensate for decreasing the minimum retransmit timeout.
Jeffrey Hsu [Wed, 4 Jun 2003 10:03:55 +0000 (10:03 +0000)]
Compensate for decreasing the minimum retransmit timeout.

Reviewed by: jlemon

21 years agoMAINTAINER line moved to MAINTAINERS file.
Tony Finch [Wed, 4 Jun 2003 10:01:10 +0000 (10:01 +0000)]
MAINTAINER line moved to MAINTAINERS file.

21 years agoCopy the (normalized) MAINTAINER= entries from Makefile's and other sources,
Doug Barton [Wed, 4 Jun 2003 09:55:26 +0000 (09:55 +0000)]
Copy the (normalized) MAINTAINER= entries from Makefile's and other sources,
and suggest consolidating them into this file instead.

21 years agoRemove MAINTAINER line in favor of the entry in src/MAINTAINERS
Doug Barton [Wed, 4 Jun 2003 09:38:04 +0000 (09:38 +0000)]
Remove MAINTAINER line in favor of the entry in src/MAINTAINERS

21 years agoMake C applications statically compiled with libthr work. Previously,
Mike Makonnen [Wed, 4 Jun 2003 08:23:05 +0000 (08:23 +0000)]
Make C applications statically compiled with libthr work. Previously,
an application compiled -static with libthr would dump core in
malloc(3) because the stub thread initialization routine in libc would
be used instead of the libthr supplied one.

21 years agoIf the library is not able to create a thread because resources
Mike Makonnen [Wed, 4 Jun 2003 08:16:32 +0000 (08:16 +0000)]
If the library is not able to create a thread because resources
don't allow it at the moment, the correct thing to do is try again.
Otherwise, libthr would fail this test because it doesn't allow
an unlimited number of concurrent threads per application.

21 years ago - Add vm object locking to vm_object_deallocate(). (Still more
Alan Cox [Wed, 4 Jun 2003 06:00:55 +0000 (06:00 +0000)]
 - Add vm object locking to vm_object_deallocate().  (Still more
   changes are required.)
 - Remove special-case macros for kmem object locking.  They are
   no longer used.

21 years agoSet CSTD to gnu99. We can only use on of the gnu?9 C languages.
David E. O'Brien [Wed, 4 Jun 2003 05:42:04 +0000 (05:42 +0000)]
Set CSTD to gnu99. We can only use on of the gnu?9 C languages.
We can't use c89 due to use of 'inline', and c99 produces bad code.

21 years agoAdd instrumentation which tells us how much work softclock() does
Poul-Henning Kamp [Wed, 4 Jun 2003 05:25:58 +0000 (05:25 +0000)]
Add instrumentation which tells us how much work softclock() does
per invocation.

21 years agoGive ia64 the exact same semantics as i386 with respect to non-512
Poul-Henning Kamp [Wed, 4 Jun 2003 05:25:04 +0000 (05:25 +0000)]
Give ia64 the exact same semantics as i386 with respect to non-512
byte sector devices.

21 years agoHave the port readmes follow the same logic as buildworld. We skip
Marcel Moolenaar [Wed, 4 Jun 2003 05:17:18 +0000 (05:17 +0000)]
Have the port readmes follow the same logic as buildworld. We skip
doing make readmes if /tmp/.skip_ports exists. This file is created
after a complete run, which avoids that we do the lengthy operation
again for a rerelease.
When NOPORTS or NOPORTREADMES are defined, we create the file prior
to starting the release.

21 years agoAdd id for TI's 1394b link chip.
Hidetoshi Shimokawa [Wed, 4 Jun 2003 04:26:14 +0000 (04:26 +0000)]
Add id for TI's 1394b link chip.

21 years agoUserspace prototypes for the extattr_list_*() system calls.
Robert Watson [Wed, 4 Jun 2003 04:04:24 +0000 (04:04 +0000)]
Userspace prototypes for the extattr_list_*() system calls.

21 years agoDocument the BUG (feature) that you can pass in an empty name to
Robert Watson [Wed, 4 Jun 2003 04:03:58 +0000 (04:03 +0000)]
Document the BUG (feature) that you can pass in an empty name to
VOP_GETEXTATTR() to retrieve the attribute name list on some file
systems, and note that this will be replaced with its own VOP
in due course.

Pointed out by: Dominic Giampaolo <dbg@apple.com>

21 years agoThe vnode operations for extended attributes no longer suffer from
Robert Watson [Wed, 4 Jun 2003 04:01:44 +0000 (04:01 +0000)]
The vnode operations for extended attributes no longer suffer from
the features (bugs) in the BUGS section related to querying the
required buffer size, or telling if an overflow occured.

21 years agoDocument the new explicit listing API for extended attributes; note
Robert Watson [Wed, 4 Jun 2003 04:00:26 +0000 (04:00 +0000)]
Document the new explicit listing API for extended attributes; note
that the old API (passing "" as the attribute name to the _get_
interface) is now deprecated (and was probably a bad idea).

Pointed out by: Dominic Giampaolo <dbg@apple.com>
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoImplementations of extattr_list_fd(), extattr_list_file(), and
Robert Watson [Wed, 4 Jun 2003 03:57:28 +0000 (03:57 +0000)]
Implementations of extattr_list_fd(), extattr_list_file(), and
extattr_list_link() system calls, which return a least of extended
attributes defined for a vnode referenced by a file descriptor
or path name.  Currently, we just invoke VOP_GETEXTATTR() since
it will convert a request for an empty name into a query for a
name list, which was the old (more hackish) API.  At some point
in the near future, we'll push the distinction between get and
list down to the vnode operation layer, but this provides access
to the new API for applications in the short term.

Pointed out by: Dominic Giampaolo <dbg@apple.com>
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoRegen from syscalls.master:1.149, addition of extended attribute
Robert Watson [Wed, 4 Jun 2003 03:50:20 +0000 (03:50 +0000)]
Regen from syscalls.master:1.149, addition of extended attribute
list system calls for fd, file, link.

21 years agoAdd system calls to explicitly list extended attributes on a
Robert Watson [Wed, 4 Jun 2003 03:49:31 +0000 (03:49 +0000)]
Add system calls to explicitly list extended attributes on a
file/directory/link, rather than using a less explicit hack on
the extattr retrieval API:

  extattr_list_fd()
  extattr_list_file()
  extattr_list_link()

The existing API was counter-intuitive, and poorly documented.
The prototypes for these system calls are identical to
extattr_get_*(), but without a specific attribute name to
leave NULL.

Pointed out by: Dominic Giampaolo <dbg@apple.com>
Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

21 years agoKMF_DONE is now in /sys/sys/kse.h, no longer need to define it here.
David Xu [Wed, 4 Jun 2003 03:22:59 +0000 (03:22 +0000)]
KMF_DONE is now in /sys/sys/kse.h, no longer need to define it here.

21 years agoFix ALIGNED_POINTER(). sizeof((u_int32_t)) is not legal C.
Peter Wemm [Wed, 4 Jun 2003 02:15:13 +0000 (02:15 +0000)]
Fix ALIGNED_POINTER().  sizeof((u_int32_t)) is not legal C.

21 years agoUnbreak ia64. 'nuff said.
Marcel Moolenaar [Wed, 4 Jun 2003 02:01:05 +0000 (02:01 +0000)]
Unbreak ia64. 'nuff said.

21 years agoChange handling to support strong alignment architectures such as alpha and
Bernd Walter [Wed, 4 Jun 2003 01:17:37 +0000 (01:17 +0000)]
Change handling to support strong alignment architectures such as alpha and
sparc64.

PR: alpha/50658
Submitted by: rizzo
Tested on: alpha

21 years agoAssert the vnode lock when returning successfully from vn_open_cred().
Robert Watson [Wed, 4 Jun 2003 00:54:27 +0000 (00:54 +0000)]
Assert the vnode lock when returning successfully from vn_open_cred().

21 years agoRemove un-needed code.
Julian Elischer [Wed, 4 Jun 2003 00:12:57 +0000 (00:12 +0000)]
Remove un-needed code.
Don't copyin() data we are about to overwrite.
Add a flag to tell userland that KSE is officially "DONE" with the
mailbox and has gone away.

Obtained from: davidxu@

21 years agoDon't call sbp_do_attach() recursively after agent reset
Hidetoshi Shimokawa [Tue, 3 Jun 2003 23:01:43 +0000 (23:01 +0000)]
Don't call sbp_do_attach() recursively after agent reset
while device probing.

21 years agoAllocate zeroed space for fwdev.
Hidetoshi Shimokawa [Tue, 3 Jun 2003 22:58:22 +0000 (22:58 +0000)]
Allocate zeroed space for fwdev.

21 years agoOpen fwmem device exclusively.
Hidetoshi Shimokawa [Tue, 3 Jun 2003 22:52:55 +0000 (22:52 +0000)]
Open fwmem device exclusively.

21 years agoAdd a geom_vol_ffs module and hook up to the build.
Gordon Tetlow [Tue, 3 Jun 2003 20:32:13 +0000 (20:32 +0000)]
Add a geom_vol_ffs module and hook up to the build.

21 years agoAdd vm object locking to vm_object_coalesce().
Alan Cox [Tue, 3 Jun 2003 19:37:01 +0000 (19:37 +0000)]
Add vm object locking to vm_object_coalesce().

21 years agoFix a potential bucket leak where when freeing to an empty bucket
Bosko Milekic [Tue, 3 Jun 2003 19:19:13 +0000 (19:19 +0000)]
Fix a potential bucket leak where when freeing to an empty bucket
we failed to put the bucket back into the general cache/container.

Also, fix a bad assumption.  There was a KASSERT() that aimed to
guarantee that whenever the pcpu container's mc_starved was > 0,
that whatever the bucket we were freeing to was an empty bucket,
assuming it belonged to the pcpu container cache. However, there
is at least one case where this is not true anymore; consider:
1) All containers empty, next thread to try to alloc will touch
   a pcpu container, notice it's empty, and increment the pcpu
   container's mc_starved.
2) Some other thread frees an mbuf belonging to a bucket in
   the general cache/container.  Then it frees another mbuf
   belonging to the same bucket (still in gen container).
3) Some third thread tries to allocate an mbuf from the pcpu
   container and, since empty, grabs one mbuf now available
   in the general cache and moves the non-empty bucket from
   which it took 1 mbuf and to which the thread in (2) freed
   to, and moves it to the pcpu container.
4) A final thread tries to free an mbuf belonging to the
   NON-EMPTY bucket mentionned in (2) and (3) and, since
   the pcpu container's mc_starved is > 0, but the bucket
   is obviously non-empty, it trips on the KASSERT.
This meant that one could potentially get a panic in some
cases when out of mbufs and clusters.  The problem could
be mitigated by commenting out some cv_signal() calls,
but I'm assuming that was pure coincidence and this is
the correct fix.

21 years agoBah, revert the previous commit for the time being due to inadequate testing
John Baldwin [Tue, 3 Jun 2003 18:44:31 +0000 (18:44 +0000)]
Bah, revert the previous commit for the time being due to inadequate testing
on my part.  The output asm looks correct with the previous commit in place
and it works on amd64, but on my laptop I got a spew of AE_BAD_PARAMETER
errors trying to unlock the acpi global lock.

21 years agoUpdate MPILIB from code received from LSI. Make changes in the rest of
Matt Jacob [Tue, 3 Jun 2003 17:47:48 +0000 (17:47 +0000)]
Update MPILIB from code received from LSI. Make changes in the rest of
the driver based upon some somewhat gratuitous name changes.

21 years agoMake ed compile in the NOCRYPT case.
John Hay [Tue, 3 Jun 2003 17:03:48 +0000 (17:03 +0000)]
Make ed compile in the NOCRYPT case.

Reviewed by: markm

21 years agoChange fatal to warnx() for unknown encoding
Andrey A. Chernov [Tue, 3 Jun 2003 16:58:31 +0000 (16:58 +0000)]
Change fatal to warnx() for unknown encoding

21 years agoFix the asm constraints so that we use the correct constants when acquiring
John Baldwin [Tue, 3 Jun 2003 14:46:16 +0000 (14:46 +0000)]
Fix the asm constraints so that we use the correct constants when acquiring
and releasing ACPI global locks instead of (ab)using the pointers to those
locks as the constants.  Also, rather than require that the address of
the lock be stored in a register, use a memory constraint allowing the
memory address to be used directly.

Noticed by: peter

21 years agoMdoc: fix hard sentence breaks.
Hiten Pandya [Tue, 3 Jun 2003 14:19:24 +0000 (14:19 +0000)]
Mdoc: fix hard sentence breaks.

Approved by: des (mentor)

21 years agoDocument that kldload(2) can also return EEXIST.
Hiten Pandya [Tue, 3 Jun 2003 12:29:34 +0000 (12:29 +0000)]
Document that kldload(2) can also return EEXIST.

Approved by: des (mentor)

21 years agoCatch up with revs 1.49-1.50 of df.c: don't mention -t in the BUGS section
Bruce Evans [Tue, 3 Jun 2003 12:00:35 +0000 (12:00 +0000)]
Catch up with revs 1.49-1.50 of df.c: don't mention -t in the BUGS section
since it has been fixed.

21 years agoFixed exit code in previous commit. "var++" to set a flag to nonzero
Bruce Evans [Tue, 3 Jun 2003 11:54:42 +0000 (11:54 +0000)]
Fixed exit code in previous commit.  "var++" to set a flag to nonzero
is a style bug at best.  When the variable isn't a flag, it potentially
overflows after a large number of settings.  Here the number of settings
is limited by ARG_MAX, but the variable is the exit code so it became
bogus after the second setting and effectively overflowed to 0 after
approx. 128 settings.

Fixed some style bugs involving comments in and near previous commit.

Clarification of previous commit message: df -t didn't give undefined
behaviour, and the behaviour used to conform perfectly with the man
page, since the buggy behaviour is documented in the BUGS section.  -t
just worked when no files or file systems were specified, and was just
ignored if a file or file system was specified.

21 years agoTeach recent changes in the umtx structure in the kernel to the libthr
Mike Makonnen [Tue, 3 Jun 2003 09:31:33 +0000 (09:31 +0000)]
Teach recent changes in the umtx structure in the kernel to the libthr
initialiazer.

Found by: tinderbox

21 years agoRemove SI_SUB_VINUM. SI_SUB_RAID makes more sense.
Greg Lehey [Tue, 3 Jun 2003 08:41:04 +0000 (08:41 +0000)]
Remove SI_SUB_VINUM.  SI_SUB_RAID makes more sense.

Submitted by: hmp

21 years agoThe midway driver has been busdmaified and is not i386-specific anymore.
Hartmut Brandt [Tue, 3 Jun 2003 07:58:39 +0000 (07:58 +0000)]
The midway driver has been busdmaified and is not i386-specific anymore.
Repo-copy it from the i386 specific man4 section to the common one.
Remove the i386 keyword from the man page header.
Document the sysctl interface to the driver.
Fix a spelling error (ALL -> AAL).

21 years ago - Remove the blocked pointer from the umtx structure.
Jeff Roberson [Tue, 3 Jun 2003 05:24:46 +0000 (05:24 +0000)]
 - Remove the blocked pointer from the umtx structure.
 - Use a hash of umtx queues to queue blocked threads.  We hash on pid and the
   virtual address of the umtx structure.  This eliminates cases where we
   previously held a lock across a casuptr call.

Reviwed by: jhb (quickly)

21 years agoDon't deregister the kld on resetconfig, only when we're really ready
Greg Lehey [Tue, 3 Jun 2003 04:51:31 +0000 (04:51 +0000)]
Don't deregister the kld on resetconfig, only when we're really ready
to unload.  This would cause a panic on the second resetconfig.

Start Vinum at boot time at SI_SUB_RAID, not SI_SUB_VINUM.
SI_SUB_VINUM was there first, but there's no real distinction, and
SI_SUB_RAID is a more neutral name.

Submitted by:  hmp

21 years agoFree memory of internal low level lock when mutex and condition variable
David Xu [Tue, 3 Jun 2003 02:21:01 +0000 (02:21 +0000)]
Free memory of internal low level lock when mutex and condition variable
are destroyed.

Submitted by: tegge

21 years agoWhen scanning for changed containers, don't assume that the controller
Scott Long [Tue, 3 Jun 2003 02:10:55 +0000 (02:10 +0000)]
When scanning for changed containers, don't assume that the controller
will respond in a sane manner.

Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering
me to fix it.

21 years agoTwo enhancements for kdump.
Peter Wemm [Tue, 3 Jun 2003 01:44:43 +0000 (01:44 +0000)]
Two enhancements for kdump.
1) add a "-p pid", which is rather useful for selecting a single pid in
a combined trace file (eg: with ktrace -i).
2) display binary genio data in a more precise format.

21 years agoTurn on btxld for amd64.
Peter Wemm [Tue, 3 Jun 2003 01:38:16 +0000 (01:38 +0000)]
Turn on btxld for amd64.

21 years agoAdd AMD64 support.
Peter Wemm [Tue, 3 Jun 2003 01:37:32 +0000 (01:37 +0000)]
Add AMD64 support.

21 years agoAdd support for generic PMCICA ATA CARD
Warner Losh [Tue, 3 Jun 2003 01:30:55 +0000 (01:30 +0000)]
Add support for generic PMCICA ATA CARD

21 years agoSync to 1.51
Warner Losh [Tue, 3 Jun 2003 01:29:58 +0000 (01:29 +0000)]
Sync to 1.51

21 years agoAdd another PCMCIA IDE CARD
Warner Losh [Tue, 3 Jun 2003 01:29:01 +0000 (01:29 +0000)]
Add another PCMCIA IDE CARD

21 years agoAccount for packets processed at layer-2 (i.e. net.link.ether.ipfw=1).
Kelly Yancey [Mon, 2 Jun 2003 23:54:09 +0000 (23:54 +0000)]
Account for packets processed at layer-2 (i.e. net.link.ether.ipfw=1).

MFC after: 2 weeks

21 years agoMerge from private version:
Sam Leffler [Mon, 2 Jun 2003 23:45:32 +0000 (23:45 +0000)]
Merge from private version:

o add hash operations
o make AES operations useful: split into 3 ops based on key length (specify
  aes for AES w/ 128 bit keys, aes192 for 192-bit keys, and aes256 for
  256-bit keys
o add -c option to force encrypt->decrypt result to be compared against the
  original plaintext (need to compare it against a known good implementation
  like openssl)
o change multi-threaded test output to be more meaningingful
o fix default block size selection to consider algorithm's blocking needs

21 years agoRedo locking for better SMP suport:
Sam Leffler [Mon, 2 Jun 2003 23:34:59 +0000 (23:34 +0000)]
Redo locking for better SMP suport:

o adding locking to op submission
o mark interrupt handler MPSAFE
o don't use locking on detach; disabling interrupts should be sufficient
o change mutex string names so witness printouts are more meaningful

Note: locking is still pretty brute-force but it's probably not worth
improving it given the relatively low performance of hifn-based cards.

21 years agoRedo locking for proper SMP operation:
Sam Leffler [Mon, 2 Jun 2003 23:32:03 +0000 (23:32 +0000)]
Redo locking for proper SMP operation:

o replace driver-global lock with three locks: one for the handling of mcr1
  operations, one for handling of mcr2 operations, and one for the mcr1
  free list
o mark the interrupt handler MPSAFE
o don't use locking on detach; disabling interrupts is sufficient (I think)

21 years agoFlush my local cache of cryto subsystem fixes:
Sam Leffler [Mon, 2 Jun 2003 23:28:06 +0000 (23:28 +0000)]
Flush my local cache of cryto subsystem fixes:

o add a ``done'' flag for crypto operations; this is set when the operation
  completes and is intended for callers to check operations that may complete
  ``prematurely'' because of direct callbacks
o close a race for operations where the crypto driver returns ERESTART: we
  need to hold the q lock to insure the blocked state for the driver and any
  driver-private state is consistent; otherwise drivers may take an interrupt
  and notify the crypto subsystem that it can unblock the driver but operations
  will be left queued and never be processed
o close a race in /dev/crypto where operations can complete before the caller
  can sleep waiting for the callback: use a per-session mutex and the new done
  flag to handle this
o correct crypto_dispatch's handling of operations where the driver returns
  ERESTART: the return value must be zero and not ERESTART, otherwise the
  caller may free the crypto request despite it being queued for later handling
  (this typically results in a later panic)
o change crypto mutex ``names'' so witness printouts and the like are more
  meaningful

21 years agoCosmetic; record offsets used within jmpbuf
Peter Wemm [Mon, 2 Jun 2003 22:37:53 +0000 (22:37 +0000)]
Cosmetic; record offsets used within jmpbuf

21 years agoFrom the df man page:
Jordan K. Hubbard [Mon, 2 Jun 2003 22:33:12 +0000 (22:33 +0000)]
From the df man page:
-t Only print out statistics for filesystems of the specified types.

Make the behavior of df(1) conform to its man page (behavior is otherwise
undefined).

Submitted by: Rob Braun <bbraun@apple.com>
Obtained from: Apple

21 years agoUpdate for amd64 after repocopy from i386/
Peter Wemm [Mon, 2 Jun 2003 22:30:59 +0000 (22:30 +0000)]
Update for amd64 after repocopy from i386/

21 years agoPort libc_r to amd64, and turn it back on for amd64. It passes all of
Peter Wemm [Mon, 2 Jun 2003 22:22:52 +0000 (22:22 +0000)]
Port libc_r to amd64, and turn it back on for amd64.  It passes all of
the same src/lib/libc_r/test/* tests that the other platforms pass.

21 years agoFix sigsetjmp(). It helps if we do not try to save the old signal mask
Peter Wemm [Mon, 2 Jun 2003 21:59:13 +0000 (21:59 +0000)]
Fix sigsetjmp().  It helps if we do not try to save the old signal mask
to a random memory location.  Perl works a lot better with this.

21 years agoFix restarted syscalls. When we rewind %rip, we also need to restore
Peter Wemm [Mon, 2 Jun 2003 21:56:08 +0000 (21:56 +0000)]
Fix restarted syscalls.  When we rewind %rip, we also need to restore
all the argument registers etc since we have almost certainly have trashed
them by now.  Take particular car of %r10 since it held the original value
of %rcx (which we saved in tf_rcx on entry and doreti doesn't know this).

21 years agoMake this more compatable with libc_r. Make the internal types for storing
Peter Wemm [Mon, 2 Jun 2003 21:49:35 +0000 (21:49 +0000)]
Make this more compatable with libc_r.  Make the internal types for storing
registers an array of longs rather than int.

21 years agoTidy the code up a fraction. Re-release with a 2-clause BSD license
Mark Murray [Mon, 2 Jun 2003 21:43:14 +0000 (21:43 +0000)]
Tidy the code up a fraction. Re-release with a 2-clause BSD license
with the kind permission of the author/copyright holder.

Thanks to: phk

21 years agoFurther devilification of CCD:
Poul-Henning Kamp [Mon, 2 Jun 2003 21:29:04 +0000 (21:29 +0000)]
Further devilification of CCD:

Change the list interface to simplify things.
Remove old list ioctls which bogusly exported the softc to userland.
Move the softc and associated structures from the public header to
the source file.

21 years agoUse GEOM OAM api to retrive list of configured ccd devices.
Poul-Henning Kamp [Mon, 2 Jun 2003 20:50:59 +0000 (20:50 +0000)]
Use GEOM OAM api to retrive list of configured ccd devices.

Link against libgeom.

21 years agoBegin deevilification of CCD:
Poul-Henning Kamp [Mon, 2 Jun 2003 20:50:03 +0000 (20:50 +0000)]
Begin deevilification of CCD:

Make CCD a GEOM class.

For now only use this for implementing a OAM config method which
can return a list of configured CCD devices in the format which
"ccdconfig -g[v]" would normally output.

21 years agoInitialize td->td_pcb->pcb_ext in cpu_thread_setup() since a garbage
Tor Egge [Mon, 2 Jun 2003 20:43:28 +0000 (20:43 +0000)]
Initialize td->td_pcb->pcb_ext in cpu_thread_setup() since a garbage
value (e.g. 0xd0d0d0d0) can cause a kernel panic.

21 years agoReturn an indicative error message.
Poul-Henning Kamp [Mon, 2 Jun 2003 20:35:37 +0000 (20:35 +0000)]
Return an indicative error message.

21 years agoAdd some missing items to the make clean target.
John Baldwin [Mon, 2 Jun 2003 20:15:12 +0000 (20:15 +0000)]
Add some missing items to the make clean target.