]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoAdd pccbb_isa.c and pccbb_pci.c
imp [Mon, 16 Aug 2004 15:23:10 +0000 (15:23 +0000)]
Add pccbb_isa.c and pccbb_pci.c

20 years agoUpgrading a lock does not play well together with acquiring an exclusive lock
kan [Mon, 16 Aug 2004 15:01:22 +0000 (15:01 +0000)]
Upgrading a lock does not play well together with acquiring an exclusive lock
and can lead to two threads being granted exclusive access. Check that no one
has the same lock in exclusive  mode before proceeding to acquire it.

The LK_WANT_EXCL and LK_WANT_UPGRADE bits act as mini-locks and can block
other threads.  Normally this is not a problem since the mini locks are
upgraded to full locks and the release of the locks will unblock the other
threads.  However if a thread reset the bits without obtaining a full lock
other threads are not awoken. Add missing wakeups for these cases.

PR: kern/69964
Submitted by: Stephan Uphoff <ups at tree dot com>
Very good catch by: Stephan Uphoff <ups at tree dot com>

20 years agoStore a pointer to "null" in struct ndblock's defn member instead of a
tjr [Mon, 16 Aug 2004 14:18:22 +0000 (14:18 +0000)]
Store a pointer to "null" in struct ndblock's defn member instead of a
duplicate allocated on the heap; the address defn points to is significant,
and is checked against the address of "null" in certain conditionals.

PR: 59883
MFC after: 1 week

20 years agoFix compile, s/tp_dtv/tp_tdv/g.
davidxu [Mon, 16 Aug 2004 14:07:38 +0000 (14:07 +0000)]
Fix compile, s/tp_dtv/tp_tdv/g.

20 years agoAdd MP_WATCHDOG option information to pc98, as it uses the i386
rwatson [Mon, 16 Aug 2004 13:59:01 +0000 (13:59 +0000)]
Add MP_WATCHDOG option information to pc98, as it uses the i386
mp_machdep.c, which relies on the option's include file and
defines.

Constancy of: tinderbox

20 years agoAdd /dev/mem and /dev/kmem to powerpc.
ssouhlal [Mon, 16 Aug 2004 13:07:40 +0000 (13:07 +0000)]
Add /dev/mem and /dev/kmem to powerpc.

Approved by: grehan (mentor)

20 years agoComplete 'IA32' -> 'COMPAT_IA32' change for the Linuxulator32.
obrien [Mon, 16 Aug 2004 12:51:33 +0000 (12:51 +0000)]
Complete 'IA32' -> 'COMPAT_IA32' change for the Linuxulator32.

20 years agoUn-comment LINPROCFS.
tjr [Mon, 16 Aug 2004 12:39:27 +0000 (12:39 +0000)]
Un-comment LINPROCFS.

20 years agoAMD64 on-CPU GART support.
obrien [Mon, 16 Aug 2004 12:25:48 +0000 (12:25 +0000)]
AMD64 on-CPU GART support.
This also applies to AMD64 HW running 'i386' OS.

Submitted by: Jung-uk Kim <jkim@niksun.com>
Integration by: obrien

20 years agostyle.9.
obrien [Mon, 16 Aug 2004 12:23:53 +0000 (12:23 +0000)]
style.9.

20 years agoFix the 'DEBUG' argument code to unbreak the amd64 LINT build.
obrien [Mon, 16 Aug 2004 12:15:07 +0000 (12:15 +0000)]
Fix the 'DEBUG' argument code to unbreak the amd64 LINT build.

20 years agoPut the 'antispoof' opcode in the proper place in the opcode list such
obrien [Mon, 16 Aug 2004 12:05:19 +0000 (12:05 +0000)]
Put the 'antispoof' opcode in the proper place in the opcode list such
that it doesn't break the ipfw2 ABI.

20 years agoUpdate low level code for Cronyx Tau-PCI:
rik [Mon, 16 Aug 2004 12:00:07 +0000 (12:00 +0000)]
Update low level code for Cronyx Tau-PCI:
1. Add support for Cronyx Tau-PCI/E3 rev.B.
2. Improve model identification and firmware load procedure.

20 years agoAdd 3653/tcp and 3653/udp, since they were specifically requested.
cperciva [Mon, 16 Aug 2004 11:52:22 +0000 (11:52 +0000)]
Add 3653/tcp and 3653/udp, since they were specifically requested.

PR: conf/63907
Submitted by: Marc Blanchet
MFC after: 3 days

20 years agoI missed an 'IA32' in the documentation.
obrien [Mon, 16 Aug 2004 11:15:46 +0000 (11:15 +0000)]
I missed an 'IA32' in the documentation.

20 years agoFix the 'DEBUG' argument code to unbreak the amd64 LINT build.
obrien [Mon, 16 Aug 2004 11:12:57 +0000 (11:12 +0000)]
Fix the 'DEBUG' argument code to unbreak the amd64 LINT build.

20 years agoI'm not sure what tjr envisioned for turning on FreeBSD/i386 rt support,
obrien [Mon, 16 Aug 2004 11:09:59 +0000 (11:09 +0000)]
I'm not sure what tjr envisioned for turning on FreeBSD/i386 rt support,
but make it COMPAT_IA32 for now.
Fix the 'DEBUG' argument code to unbreak the amd64 LINT build.

20 years agoFix the 'DEBUG' argument code to unbreak the amd64 LINT build.
obrien [Mon, 16 Aug 2004 10:54:25 +0000 (10:54 +0000)]
Fix the 'DEBUG' argument code to unbreak the amd64 LINT build.

20 years agoRemove Wiretek UBRJ4 from the list of supported devices. While it is
simon [Mon, 16 Aug 2004 10:49:45 +0000 (10:49 +0000)]
Remove Wiretek UBRJ4 from the list of supported devices.  While it is
detected by the driver, it doesn't really work as soon as it really used
for network traffic.  Perhaps it can be re-added later when the issues
are resolved.

Reported by: Soeren Vrist <seet@seet.dk>

20 years agoAdd myself as geom_raid3 maintainer.
pjd [Mon, 16 Aug 2004 10:43:41 +0000 (10:43 +0000)]
Add myself as geom_raid3 maintainer.

20 years agoFix the 'DEBUG' argument code to unbreak the LINT build.
obrien [Mon, 16 Aug 2004 10:36:12 +0000 (10:36 +0000)]
Fix the 'DEBUG' argument code to unbreak the LINT build.

20 years agoMinor style.9 cleanup.
obrien [Mon, 16 Aug 2004 10:33:35 +0000 (10:33 +0000)]
Minor style.9 cleanup.

20 years agoBring back 'NO_FLOPPIES', but as a user external knob.
obrien [Mon, 16 Aug 2004 10:25:43 +0000 (10:25 +0000)]
Bring back 'NO_FLOPPIES', but as a user external knob.
The internal logic knob remains MAKE_FLOPPIES.

Suggested by: ru

20 years agoWhen looking for some extra data to include in the hash, use the
dwmalone [Mon, 16 Aug 2004 10:00:44 +0000 (10:00 +0000)]
When looking for some extra data to include in the hash, use the
address of the dirhash, rather than the first sizeof(struct dirhash
*) bytes of the structure (which, thankfully, seem to be constant).

Submitted by: Ted Unangst <tedu@zeitbombe.org>
MFC after: 2 weeks

20 years agoImprove MIME handling. This patch is based on Eugene's patch, but
dwmalone [Mon, 16 Aug 2004 09:38:34 +0000 (09:38 +0000)]
Improve MIME handling. This patch is based on Eugene's patch, but
with the following changes:

1) Don't make a mime_types.h 'cos we should avoid creating variables
   in header files,
2) Use strrchr to find the extension, rather than strchr,
3) Slightly simplify the mime-type matching loop.

any goof are likely to be mine. Note that there are links to more
improvements by Eugene in the PR.

PR: 29725
Submitted by: Eugene Grosbein <eugen@kuzbass.ru>

20 years agoRemove some debug cruft from the Makefiles and make them ${.OBJDIR}
dfr [Mon, 16 Aug 2004 09:35:49 +0000 (09:35 +0000)]
Remove some debug cruft from the Makefiles and make them ${.OBJDIR}
friendly.

20 years agoImprove (hopefully) on the workaround code for devices that doesn't
sos [Mon, 16 Aug 2004 09:32:35 +0000 (09:32 +0000)]
Improve (hopefully) on the workaround code for devices that doesn't
interrupt when command is done, ie some ATAPI CD drives with no
media loaded.

20 years agoAdd Hungarian calendar entries.
dwmalone [Mon, 16 Aug 2004 09:31:09 +0000 (09:31 +0000)]
Add Hungarian calendar entries.

PR: 42725
Submitted by: Janos Mohacsi <janos.mohacsi@bsd.hu>

20 years agoFix Promise sx4/sx4000 support that broke on the latest race fixes.
sos [Mon, 16 Aug 2004 09:29:46 +0000 (09:29 +0000)]
Fix Promise sx4/sx4000 support that broke on the latest race fixes.

20 years agoAdd regression tests for GEOM_RAID3.
pjd [Mon, 16 Aug 2004 09:09:23 +0000 (09:09 +0000)]
Add regression tests for GEOM_RAID3.

20 years agoAccept the fact that floppies either aren't applicable or relevant on most
obrien [Mon, 16 Aug 2004 08:58:04 +0000 (08:58 +0000)]
Accept the fact that floppies either aren't applicable or relevant on most
platforms.  So follow 'MAKE_ISOS''s style and s/NO_FLOPPIES/MAKE_FLOPPIES/g.

20 years agoFollow 'MAKE_ISOS' in naming and s/FLOPPIES/MAKE_FLOPPIES/g
obrien [Mon, 16 Aug 2004 08:56:50 +0000 (08:56 +0000)]
Follow 'MAKE_ISOS' in naming and s/FLOPPIES/MAKE_FLOPPIES/g

Submitted by: ru

20 years agoIncrease the scaling of VM_KMEM_SIZE_MAX.
obrien [Mon, 16 Aug 2004 08:35:22 +0000 (08:35 +0000)]
Increase the scaling of VM_KMEM_SIZE_MAX.

Submitted by: alc

20 years agos/MAX_SAFE_MAXVNODES/MAXVNODES_MAX/g
obrien [Mon, 16 Aug 2004 08:33:37 +0000 (08:33 +0000)]
s/MAX_SAFE_MAXVNODES/MAXVNODES_MAX/g

20 years agoDecrease debug level to 0.
pjd [Mon, 16 Aug 2004 08:33:04 +0000 (08:33 +0000)]
Decrease debug level to 0.

20 years agoAccept the fact that floppies either aren't applicable or relevant on most
obrien [Mon, 16 Aug 2004 08:28:58 +0000 (08:28 +0000)]
Accept the fact that floppies either aren't applicable or relevant on most
platforms.  So make the floppy logic easier to read.

20 years agoFix warning.
pjd [Mon, 16 Aug 2004 08:21:31 +0000 (08:21 +0000)]
Fix warning.

20 years agoAdd support for 32-bit Linux binary emulation on amd64:
tjr [Mon, 16 Aug 2004 08:19:18 +0000 (08:19 +0000)]
Add support for 32-bit Linux binary emulation on amd64:
- include <machine/../linux32/linux.h> instead of <machine/../linux/linux.h>
  if building with the COMPAT_LINUX32 option.
- make minimal changes to the i386 linprocfs_docpuinfo() function to support
  amd64. We return a fake CPU family of 6 for now.

20 years agoFix alpha build and add __FBSDID.
dfr [Mon, 16 Aug 2004 08:12:29 +0000 (08:12 +0000)]
Fix alpha build and add __FBSDID.

PR: 70518

20 years agoRegen.
tjr [Mon, 16 Aug 2004 08:07:06 +0000 (08:07 +0000)]
Regen.

20 years agoAdd preliminary support for running 32-bit Linux binaries on amd64, enabled
tjr [Mon, 16 Aug 2004 07:55:06 +0000 (07:55 +0000)]
Add preliminary support for running 32-bit Linux binaries on amd64, enabled
with the COMPAT_LINUX32 option. This is largely based on the i386 MD Linux
emulations bits, but also builds on the 32-bit FreeBSD and generic IA-32
binary emulation work.

Some of this is still a little rough around the edges, and will need to be
revisited before 32-bit and 64-bit Linux emulation support can coexist in
the same kernel.

20 years agoUpdate the date.
pjd [Mon, 16 Aug 2004 07:53:43 +0000 (07:53 +0000)]
Update the date.

Reminded by: ru

20 years agoThis patch merges the sort fields for both pages, so you can (for
alfred [Mon, 16 Aug 2004 07:51:22 +0000 (07:51 +0000)]
This patch merges the sort fields for both pages, so you can (for
example) view io stats while sorting by process size.  Also adds
voluntary and involuntary context-switch stats to the io page because
there was lots of room.

Submitted by: Dan Nelson dnelson at allantgroup.com

20 years agoChanges to MI Linux emulation code necessary to run 32-bit Linux binaries
tjr [Mon, 16 Aug 2004 07:28:16 +0000 (07:28 +0000)]
Changes to MI Linux emulation code necessary to run 32-bit Linux binaries
on AMD64, and the general case where the emulated platform has different
size pointers than we use natively:
- declare certain structure members as l_uintptr_t and use the new PTRIN
  and PTROUT macros to convert to and from native pointers.
- declare some structures __packed on amd64 when the layout would differ
  from that used on i386.
- include <machine/../linux32/linux.h> instead of <machine/../linux/linux.h>
  if compiling with COMPAT_LINUX32. This will need to be revisited before
  32-bit and 64-bit Linux emulation support can coexist in the same kernel.
- other small scattered changes.

This should be a no-op on i386 and Alpha.

20 years agoAdd a new type, l_uintptr_t, which is an unsigned integer type with the
tjr [Mon, 16 Aug 2004 07:05:44 +0000 (07:05 +0000)]
Add a new type, l_uintptr_t, which is an unsigned integer type with the
same width as a pointer under Linux. Add two new macros, PTRIN and PTROUT,
which convert between l_uintptr_t and native pointers.

20 years agoUse a local "compress" symbol corresponding to a variable in BSS,
ru [Mon, 16 Aug 2004 07:02:14 +0000 (07:02 +0000)]
Use a local "compress" symbol corresponding to a variable in BSS,
rather than the one from libz, corresponding to a function, when
linking statically.

PR: bin/70392

20 years agoInform hier(7) about /usr/include/geom/raid3/ directory.
pjd [Mon, 16 Aug 2004 06:39:10 +0000 (06:39 +0000)]
Inform hier(7) about /usr/include/geom/raid3/ directory.

20 years agoConnect RAID3 GEOM class to the build.
pjd [Mon, 16 Aug 2004 06:36:21 +0000 (06:36 +0000)]
Connect RAID3 GEOM class to the build.

20 years agoDon't need to declare cbb module. don't know why I never saw
imp [Mon, 16 Aug 2004 06:33:58 +0000 (06:33 +0000)]
Don't need to declare cbb module.  don't know why I never saw
duplicate messages..

20 years agoMention a potential net.inet.ip.random_id drawback.
maxim [Mon, 16 Aug 2004 06:30:58 +0000 (06:30 +0000)]
Mention a potential net.inet.ip.random_id drawback.

Submitted by: ru

20 years agoAdvertise that color is supported so that syscons doesn't come up
grehan [Mon, 16 Aug 2004 06:26:15 +0000 (06:26 +0000)]
Advertise that color is supported so that syscons doesn't come up
in monochrome mode when run as init.

20 years agoIntroduce GEOM RAID3 class, i.e. kernel module, which implements RAID3
pjd [Mon, 16 Aug 2004 06:23:14 +0000 (06:23 +0000)]
Introduce GEOM RAID3 class, i.e. kernel module, which implements RAID3
transformation and graid3(8) userland utility, which can be used for
configuration. No manual page yet, sorry.

Hardware provided by: Daniel Seuffert

20 years ago - Introduce and use a new tunable "debug.mpsafevm". At present, setting
alc [Mon, 16 Aug 2004 06:16:12 +0000 (06:16 +0000)]
 - Introduce and use a new tunable "debug.mpsafevm".  At present, setting
   "debug.mpsafevm" results in (almost) Giant-free execution of zero-fill
   page faults.  (Giant is held only briefly, just long enough to determine
   if there is a vnode backing the faulting address.)

   Also, condition the acquisition and release of Giant around calls to
   pmap_remove() on "debug.mpsafevm".

   The effect on performance is significant.  On my dual Opteron, I see a
   3.6% reduction in "buildworld" time.

 - Use atomic operations to update several counters in vm_fault().

20 years agoBring PPC up to date with latest TLS changes.
grehan [Mon, 16 Aug 2004 05:41:39 +0000 (05:41 +0000)]
Bring PPC up to date with latest TLS changes.

20 years agoPlug a memory leak in error case.
davidxu [Mon, 16 Aug 2004 05:20:12 +0000 (05:20 +0000)]
Plug a memory leak in error case.

20 years agoDocument the fact that matching against a process command-name will
gad [Mon, 16 Aug 2004 04:49:43 +0000 (04:49 +0000)]
Document the fact that matching against a process command-name will
only work on the first MAXCOMLEN (19) characters of that name.

Noticed by: Peter Holm

20 years agoAlways acquire the UNIX domain socket subsystem lock (UNP lock)
rwatson [Mon, 16 Aug 2004 04:41:03 +0000 (04:41 +0000)]
Always acquire the UNIX domain socket subsystem lock (UNP lock)
before dereferencing sotounpcb() and checking its value, as so_pcb
is protected by protocol locking, not subsystem locking.  This
prevents races during close() by one thread and use of ths socket
in another.

unp_bind() now assert the UNP lock, and uipc_bind() now acquires
the lock around calls to unp_bind().

20 years ago1. Use libpthread's exported symbols to calcuate offset in data structure
davidxu [Mon, 16 Aug 2004 03:30:16 +0000 (03:30 +0000)]
1. Use libpthread's exported symbols to calcuate offset in data structure
2. Enable TLS debugger support.

20 years ago1. Add macro DTV_OFFSET to calculate dtv offset in tcb.
davidxu [Mon, 16 Aug 2004 03:27:29 +0000 (03:27 +0000)]
1. Add macro DTV_OFFSET to calculate dtv offset in tcb.
2. Export symbols needed by debugger.

20 years agoAdd a file to collection all symbols will be needed by debugger.
davidxu [Mon, 16 Aug 2004 03:25:07 +0000 (03:25 +0000)]
Add a file to collection all symbols will be needed by debugger.

20 years agoAs with the non-rescue version don't build fore_dnld when NOATM is defined.
marius [Mon, 16 Aug 2004 03:16:48 +0000 (03:16 +0000)]
As with the non-rescue version don't build fore_dnld when NOATM is defined.

20 years agoFix installing from SCSI tape drives by catching up device names with
bmah [Mon, 16 Aug 2004 03:13:12 +0000 (03:13 +0000)]
Fix installing from SCSI tape drives by catching up device names with
reality.

PR: 70279
Submitted by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk>

20 years agoRemove the BUGS entry for vslock(9) not actually holding user memory in place.
green [Mon, 16 Aug 2004 03:12:10 +0000 (03:12 +0000)]
Remove the BUGS entry for vslock(9) not actually holding user memory in place.

20 years agoRather than bringing back all of the changes to make VM map deletion
green [Mon, 16 Aug 2004 03:11:09 +0000 (03:11 +0000)]
Rather than bringing back all of the changes to make VM map deletion
wait for system wires to disappear, do so (much more trivially) by
instead only checking for system wires of user maps and not kernel maps.

Alternative by: tor
Reviewed by: alc

20 years agoAdd the missing knote_fdclose().
green [Mon, 16 Aug 2004 03:09:01 +0000 (03:09 +0000)]
Add the missing knote_fdclose().

20 years agoAllocate the marker, when scanning a kqueue, from the "heap" instead of the
green [Mon, 16 Aug 2004 03:08:38 +0000 (03:08 +0000)]
Allocate the marker, when scanning a kqueue, from the "heap" instead of the
stack.  When swapped out, a process's kernel stack would be unavailable,
and we could get a page fault when scanning the same kqueue.

PR: kern/61849

20 years agoDidn't intend to commit debugging code enabled
imp [Mon, 16 Aug 2004 01:57:06 +0000 (01:57 +0000)]
Didn't intend to commit debugging code enabled

20 years agoAnnotate the current UNIX domain socket locking strategies, order,
rwatson [Mon, 16 Aug 2004 01:52:04 +0000 (01:52 +0000)]
Annotate the current UNIX domain socket locking strategies, order,
strengths, and weaknesses in a comment.  Assert a copyright over the
changes made as part of the locking work.

20 years agoVarious updates for installation documentation. Most notably:
bmah [Mon, 16 Aug 2004 01:42:43 +0000 (01:42 +0000)]
Various updates for installation documentation.  Most notably:

o Update to match 5-CURRENT reality.
o Bump up minimum system requirements.
o Make examples work.

PR: docs/70485
Submitted by: Gavin Atkinson <gavin.atkinson@ury.york.ac.uk>

20 years agoMajor enhancements to pipe memory usage:
silby [Mon, 16 Aug 2004 01:27:24 +0000 (01:27 +0000)]
Major enhancements to pipe memory usage:

- pipespace is now able to resize non-empty pipes; this allows
  for many more resizing opportunities

- Backing is no longer pre-allocated for the reverse direction
  of pipes.  This direction is rarely (if ever) used, so this cuts the
  amount of map space allocated to a pipe in half.

- Pipe growth is now much more dynamic; a pipe will now grow when
  the total amount of data it contains and the size of the write are
  larger than the size of pipe.  Previously, only individual writes greater
  than the size of the pipe would cause growth.

- In low memory situations, pipes will now shrink during both read
  and write operations, where possible.  Once the memory shortage
  ends, the growth code will cause these pipes to grow back to an appropriate
  size.

- If the full PIPE_SIZE allocation fails when a new pipe is created, the
  allocation will be retried with SMALL_PIPE_SIZE.  This helps to deal
  with the situation of a fragmented map after a low memory period has
  ended.

- Minor documentation + code changes to support the above.

In total, these changes increase the total number of pipes that
can be allocated simultaneously, drastically reducing the chances that
pipe allocation will fail.

Performance appears unchanged due to dynamic resizing.

20 years agoMFNetBSD
mbr [Mon, 16 Aug 2004 00:20:31 +0000 (00:20 +0000)]
MFNetBSD

Decrease log severity to debug if a protocol is not supported by the
kernel (rpcbind checks /etc/netconfig if a protocol is available).
This avoids "rpcbind: cannot create socket for tcp6" messages
at startup on IPv4-only kernels.

20 years agoComment out the ability to enable/disable ACPI at runtime. This appears
njl [Sun, 15 Aug 2004 23:39:37 +0000 (23:39 +0000)]
Comment out the ability to enable/disable ACPI at runtime.  This appears
to not work reliably and crash some systems.  It is not supported at all
on others.  Pending discussion, the underlying ioctls will be removed.

20 years agoTweak the compatibility macros a little so that the device printing is
imp [Sun, 15 Aug 2004 23:39:18 +0000 (23:39 +0000)]
Tweak the compatibility macros a little so that the device printing is
moved into them.

20 years agoResolve conflicts from dhclient 3.0.1 final import.
mbr [Sun, 15 Aug 2004 23:37:10 +0000 (23:37 +0000)]
Resolve conflicts from dhclient 3.0.1 final import.

No functional changes in this import.

20 years agoThis commit was generated by cvs2svn to compensate for changes in r133783,
mbr [Sun, 15 Aug 2004 23:33:13 +0000 (23:33 +0000)]
This commit was generated by cvs2svn to compensate for changes in r133783,
which included commits to RCS files with non-trunk default branches.

20 years agoImport ISC DHCP 3.0.1 client.
mbr [Sun, 15 Aug 2004 23:33:13 +0000 (23:33 +0000)]
Import ISC DHCP 3.0.1 client.

20 years agoOther part of the cbb isa/pci split. This likely is causing problems
imp [Sun, 15 Aug 2004 23:17:54 +0000 (23:17 +0000)]
Other part of the cbb isa/pci split.  This likely is causing problems
wrt cbb module.

20 years agoAdd a kludge for building SBus-only kernels, i.e. kernels without support
marius [Sun, 15 Aug 2004 22:59:34 +0000 (22:59 +0000)]
Add a kludge for building SBus-only kernels, i.e. kernels without support
for EBus, ISA and PCI, by compiling ofw_isa.c and ofw_pci_if.m unconditio-
nally. The correct way is to rewrite OF_decode_addr() in ofw_machdep.c in
a bus-neutral way. That's certainly possible but we unfortunately didn't
make it for FreeBSD 5.3.

Approved by: tmm

20 years ago- Handle the '\&' mdoc(7) escape sequence.
simon [Sun, 15 Aug 2004 22:33:10 +0000 (22:33 +0000)]
- Handle the '\&' mdoc(7) escape sequence.
- Handle the .Sx macro and give a warning if it is used in the
  HARDWARE section, since that will probably produce odd text in the
  Hardware Notes.

20 years agoFix a couple of edge cases in which sb.st_size may be incorrect or
des [Sun, 15 Aug 2004 22:22:35 +0000 (22:22 +0000)]
Fix a couple of edge cases in which sb.st_size may be incorrect or
meaningless.  In particular, don't assume that it is left untouched if
stat(2) fails; that assumption happens to fail at high optimization
levels on some platforms.

MFC after: 1 week

20 years ago- Auto generate device listings for the following drivers: mpt, trm,
simon [Sun, 15 Aug 2004 22:14:29 +0000 (22:14 +0000)]
- Auto generate device listings for the following drivers: mpt, trm,
  rl, vr, dc, de, and gem.
- hme(4) is not sparc64 only anymore, so update dev.archlist.txt
  acordingly.

20 years ago- Add a HARDWARE section which lists supported devices.
simon [Sun, 15 Aug 2004 21:59:15 +0000 (21:59 +0000)]
- Add a HARDWARE section which lists supported devices.
- Fix a few mdoc(7) markup / style nits when I'm here anyway.

20 years agoRelease the vnode cache mutex when calling vgone(), since vgone() may
des [Sun, 15 Aug 2004 21:58:02 +0000 (21:58 +0000)]
Release the vnode cache mutex when calling vgone(), since vgone() may
sleep.  This makes pfs_exit() even less efficient than before, but on
the bright side, the vnode cache mutex no longer needs to be recursive.

20 years agoAdd rtld-elf to the include path for the rtld to pthread TLS interface.
dfr [Sun, 15 Aug 2004 21:51:18 +0000 (21:51 +0000)]
Add rtld-elf to the include path for the rtld to pthread TLS interface.

20 years agoCorrect some uses of the wrong members of the *min()/*max()-familiy, e.g.
marius [Sun, 15 Aug 2004 21:37:52 +0000 (21:37 +0000)]
Correct some uses of the wrong members of the *min()/*max()-familiy, e.g.
min() on unsigned long. None of these are believed to have been fatal though.

Reviewed by: tmm

20 years agoBump document date for last commit.
simon [Sun, 15 Aug 2004 21:02:48 +0000 (21:02 +0000)]
Bump document date for last commit.

20 years ago - Make pmap_emulate_reference() MP and preemption safe. Previously, it
alc [Sun, 15 Aug 2004 20:54:25 +0000 (20:54 +0000)]
 - Make pmap_emulate_reference() MP and preemption safe.  Previously, it
   contained "sanity" checks that could be violated if another CPU modified
   the pmap between the emulation trap and locking the pmap in
   pmap_emulate_reference().  As a result, the pte could be inconsistent
   with the access that caused the emulation trap.  In such cases,
   pmap_emulate_reference() now flushes the current CPU's TLB entry and
   returns.
 - Make pmap_changebit() an inline function, reducing object code size.

20 years ago- Add a HARDWARE section which lists supported devices.
simon [Sun, 15 Aug 2004 20:54:07 +0000 (20:54 +0000)]
- Add a HARDWARE section which lists supported devices.
- Add the manufacturer name to each item in the device list.
- Make the note about supporting "IBM e335" into a general list and
  change the entry to use the full product name ("IBM eServer xSeries
  335").
- Add Dell PowerEdge 1750 to the list of systems with mpt onboard.

20 years agoPreemptive anti-footshooting: cause a #error if MP_WATCHDOG is compiled
rwatson [Sun, 15 Aug 2004 20:32:40 +0000 (20:32 +0000)]
Preemptive anti-footshooting: cause a #error if MP_WATCHDOG is compiled
with SCHED_ULE.

20 years agoShorten the description of NO_OPENSSL, which was split into two lines by a
marius [Sun, 15 Aug 2004 20:32:06 +0000 (20:32 +0000)]
Shorten the description of NO_OPENSSL, which was split into two lines by a
patch of mine in revision 1.216, a bit so it fits in a single line again.

20 years agoSave on one variable in ofwo_action(). Leftover from an older version of
marius [Sun, 15 Aug 2004 20:18:54 +0000 (20:18 +0000)]
Save on one variable in ofwo_action(). Leftover from an older version of
this function which needed the handle of the /options node more than once.

20 years ago- Correct the description of the "local-mac-address?" variable. Not all NICs
marius [Sun, 15 Aug 2004 20:17:29 +0000 (20:17 +0000)]
- Correct the description of the "local-mac-address?" variable. Not all NICs
  use it, only those with FCode. Add references to dc(4), gem(4) and hme(4)
  for obtaining further information about such devices presently supported
  by FreeBSD.
- Correct the HISTORY section. There was an eeprom(8) utility in 4.4BSD and
  early versions of FreeBSD 2.x.
- Add an AUTHORS section.

20 years agoSpell MP_WATCHDIG right: I fixed the build without MP_WATCHDOG after
rwatson [Sun, 15 Aug 2004 19:57:14 +0000 (19:57 +0000)]
Spell MP_WATCHDIG right: I fixed the build without MP_WATCHDOG after
testing MP_WATCHDOG, and used an incorrect ifdef.

20 years agoChange the HARDWARE section to have the same style as in the other
simon [Sun, 15 Aug 2004 19:48:27 +0000 (19:48 +0000)]
Change the HARDWARE section to have the same style as in the other
FreeBSD section 4 manual pages.

20 years agoAdd a useful snippet on limitations of inductive reasoning by Bertrand
rwatson [Sun, 15 Aug 2004 19:24:17 +0000 (19:24 +0000)]
Add a useful snippet on limitations of inductive reasoning by Bertrand
Russell.

20 years agoYet another tweak to the shutdown messages in boot():
truckman [Sun, 15 Aug 2004 19:17:23 +0000 (19:17 +0000)]
Yet another tweak to the shutdown messages in boot():

  Don't count busy buffers before the initial call to sync() and
  don't skip the initial sync() if no busy buffers were called.
  Always call sync() at least once if syncing is requested.  This
  defers the "Syncing disks, buffers remaining..." message until
  after the initial sync() call and the first count of busy
  buffers.  This backs out changes in kern_shutdown 1.162.

  Print a different message when there are no busy buffers after the
  initial sync(), which is now the expected situation.

  Print an additional message when syncing has completed successfully
  in the unusual situation where the work of syncing was done by
  boot().

  Uppercase one message to make it consistent with all of the other
  kernel shutdown messages.

Discussed with: bde (in a much earlier form, prior to 1.162)
Reviewed by: njl (in an earlier form)

20 years agoAdd a "fillchar" command line argument to dd(1) that permits the user
rwatson [Sun, 15 Aug 2004 19:10:05 +0000 (19:10 +0000)]
Add a "fillchar" command line argument to dd(1) that permits the user
to specify an alternative padding character when using a conversion
mode, or when using noerror with sync and an input error occurs.  This
facilities reading old and error-prone media by allowing the user to
more effectively mark error blocks in the output stream.

20 years ago- Add a HARDWARE section which lists supported devices.
simon [Sun, 15 Aug 2004 18:13:08 +0000 (18:13 +0000)]
- Add a HARDWARE section which lists supported devices.

20 years ago- Add a HARDWARE section which lists supported devices.
simon [Sun, 15 Aug 2004 18:09:47 +0000 (18:09 +0000)]
- Add a HARDWARE section which lists supported devices.
- Remove reference to the NOTES section in the entry for Sun DMFE,
  since ot doesn't work well with the auto generated Hardware Notes. [1]

OK'ed by: marius [1]

20 years agoAdd an "options MP_WATCHDOG" to i386. This option allows one of the
rwatson [Sun, 15 Aug 2004 18:02:09 +0000 (18:02 +0000)]
Add an "options MP_WATCHDOG" to i386.  This option allows one of the
logical CPUs on a system to be used as a dedicated watchdog to cause a
drop to the debugger and/or generate an NMI to the boot processor if
the kernel ceases to respond.  A sysctl enables the watchdog running
out of the processor's idle thread; a callout is launched to reset a
timer in the watchdog.  If the callout fails to reset the timer for ten
seconds, the watchdog will fire.  The sysctl allows you to select which
CPU will run the watchdog.

A sample "debug.leak_schedlock" is included, which causes a sysctl to
spin holding sched_lock in order to trigger the watchdog.  On my Xeons,
the watchdog is able to detect this failure mode and break into the
debugger, which cannot otherwise be done without an NMI button.

This option does not currently work with sched_ule due to ule's push
notion of scheduling, similar to machdep.hlt_logical_cpus failing to
work with that scheduler.

On face value, this might seem somewhat inefficient, but there are a
lot of dual-processor Xeons with HTT around, so using one as a watchdog
for testing is not as inefficient as one might fear.