]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoFixed misplacement of __FBSDID(). Backed out editing of vendor id lines.
bde [Thu, 11 Dec 2003 10:58:29 +0000 (10:58 +0000)]
Fixed misplacement of __FBSDID().  Backed out editing of vendor id lines.
Just wrap them in #if 0...#endif.

20 years agoFixed English error in previous commit. Fixed some older English errors.
bde [Thu, 11 Dec 2003 09:53:25 +0000 (09:53 +0000)]
Fixed English error in previous commit.  Fixed some older English errors.
Removed a redundant clause.

20 years agoRemove uses of GIANT_LOCK and replace with appropriate thread
mtm [Thu, 11 Dec 2003 08:34:07 +0000 (08:34 +0000)]
Remove uses of GIANT_LOCK and replace with appropriate thread
and thread list locks.

20 years agoRevert part of revision 1.74 after bde reminded me of a detail I'd
des [Thu, 11 Dec 2003 07:46:08 +0000 (07:46 +0000)]
Revert part of revision 1.74 after bde reminded me of a detail I'd
forgotten about how sysctl works.  This removes a potential (though
not very likely) race that 1.74 introduced.

20 years agoOnly define CACHE_LINE_SIZE in one place..
peter [Thu, 11 Dec 2003 04:49:37 +0000 (04:49 +0000)]
Only define CACHE_LINE_SIZE in one place..

20 years agoCACHE_LINE_SIZE is 64 on athlon and amd64 chips, not 32. This should
peter [Thu, 11 Dec 2003 04:47:53 +0000 (04:47 +0000)]
CACHE_LINE_SIZE is 64 on athlon and amd64 chips, not 32.  This should
probably be 128 since that is what the hardware prefetch fill size is
on both the p3, p4 and athlon* cpus.

20 years ago - Don't let the pctcpu rate limiter throttle us if we have recorded over
jeff [Thu, 11 Dec 2003 04:23:39 +0000 (04:23 +0000)]
 - Don't let the pctcpu rate limiter throttle us if we have recorded over
   SCHED_CPU_TICKS ticks.  This was allowing processes to display
   (1/SCHED_CPU_TIME * 100) % more cpu than they had used.

20 years ago - In sched_switch(), if a thread has been assigned, don't touch the runqueues
jeff [Thu, 11 Dec 2003 04:00:49 +0000 (04:00 +0000)]
 - In sched_switch(), if a thread has been assigned, don't touch the runqueues
   or load.  These things have already been taken care of in sched_bind()
   which should be the only place that we're switching in an assigned thread.

20 years ago - Add support for CPU groups to ule. All SMT cores on the same physical
jeff [Thu, 11 Dec 2003 03:57:10 +0000 (03:57 +0000)]
 - Add support for CPU groups to ule.  All SMT cores on the same physical
   cpu are added to a group.
 - Don't place a cpu into the kseq_idle bitmask until all cpus in that group
   have idled.
 - Prefer idle groups over idle group members in the new kseq_transfer()
   function.  In this way we will prefer to balance load across full cores
   rather than add further load a partial core.
 - Before a cpu goes idle, check the other group members for threads.  Since
   SMT cpus may freely share threads, this is cheap.
 - SMT cores may be individually pinned and bound to now.  This contrasts the
   old mechanism where binding or pinning would have allowed a thread to run
   on any available cpu.
 - Remove some unnecessary logic from sched_switch().  Priority propagation
   should be properly taken care of in sched_prio() now.

20 years ago - Call mp_topology() after all CPUs have been probed.
jeff [Thu, 11 Dec 2003 03:49:02 +0000 (03:49 +0000)]
 - Call mp_topology() after all CPUs have been probed.

20 years ago - Add the mp_topology() function to mp_machdep.c. This function builds up
jeff [Thu, 11 Dec 2003 03:48:31 +0000 (03:48 +0000)]
 - Add the mp_topology() function to mp_machdep.c.  This function builds up
   the smp_topology structure to reflect the layout of HTT enabled machines.
 - Add a prototype for mp_topology() in smp.h

20 years agoMFp4:
simokawa [Thu, 11 Dec 2003 03:42:40 +0000 (03:42 +0000)]
MFp4:
- Introduce sbp_targ_login instead of sbp_targ_istate.
- Implement reconnection and logout.
- Freeze simq while bus reset.

20 years agoMFi386: remove APIC_IRQ* defines that are no longer used.
peter [Thu, 11 Dec 2003 02:48:25 +0000 (02:48 +0000)]
MFi386: remove APIC_IRQ* defines that are no longer used.

20 years agoMFi386: (jhb): Deal with MAXCPU etc correctly
peter [Thu, 11 Dec 2003 02:47:35 +0000 (02:47 +0000)]
MFi386: (jhb): Deal with MAXCPU etc correctly

20 years agoregen
peter [Thu, 11 Dec 2003 02:36:37 +0000 (02:36 +0000)]
regen

20 years agoMark freebsd32_gettimeofday() as mpsafe
peter [Thu, 11 Dec 2003 02:36:07 +0000 (02:36 +0000)]
Mark freebsd32_gettimeofday() as mpsafe

20 years agoJust implementing a 32 bit version of gettimeofday() was smaller than
peter [Thu, 11 Dec 2003 02:34:49 +0000 (02:34 +0000)]
Just implementing a 32 bit version of gettimeofday() was smaller than
the wrapper code.  And it doesn't use the stackgap as a bonus.

20 years agoMove the ia32_sigtramp.S file back under amd64/. This interfaces closely
peter [Thu, 11 Dec 2003 01:09:51 +0000 (01:09 +0000)]
Move the ia32_sigtramp.S file back under amd64/.  This interfaces closely
with the sendsig code in the MD area.  It is not safe to assume that all
the register conventions will be the same.  Also, the way of producing
32 bit code (.code32 directives) in this file is amd64 specific.

20 years agoAssimilate ia64 back into the fold with the common freebsd32/ia32 code.
peter [Thu, 11 Dec 2003 01:05:09 +0000 (01:05 +0000)]
Assimilate ia64 back into the fold with the common freebsd32/ia32 code.
The split-up code is derived from the ia64 code originally.

Note that I have only compile-tested this, not actually run-tested it.
The ia64 side of the force is missing some significant chunks of signal
delivery code.

20 years agoUse the correct syscall table limit
peter [Wed, 10 Dec 2003 23:16:32 +0000 (23:16 +0000)]
Use the correct syscall table limit

20 years agoFinetune last commit, we need to take care of empty channels.
sos [Wed, 10 Dec 2003 23:06:24 +0000 (23:06 +0000)]
Finetune last commit, we need to take care of empty channels.

20 years agoFix last second typo.
peter [Wed, 10 Dec 2003 22:59:03 +0000 (22:59 +0000)]
Fix last second typo.

20 years agoUse gcc's superior ffs() builtin.
peter [Wed, 10 Dec 2003 22:51:40 +0000 (22:51 +0000)]
Use gcc's superior ffs() builtin.

20 years agoUse ffs(x) == popcnt(x ^ (x - 1)) to implement 64 bit ffsl(). gcc's
peter [Wed, 10 Dec 2003 22:47:02 +0000 (22:47 +0000)]
Use ffs(x) == popcnt(x ^ (x - 1)) to implement 64 bit ffsl().  gcc's
ffs() builtin uses this already but truncates the upper 32 bits.

20 years agoRegen
peter [Wed, 10 Dec 2003 22:33:45 +0000 (22:33 +0000)]
Regen

20 years agoAdd missing extattr_list_fd(), extattr_list_file(), extattr_list_link()
peter [Wed, 10 Dec 2003 22:33:27 +0000 (22:33 +0000)]
Add missing extattr_list_fd(), extattr_list_file(), extattr_list_link()
and kse_switchin() syscall slots.

20 years agoThe osigpending, oaccept, orecvfrom and ogetdirentries entries were
peter [Wed, 10 Dec 2003 22:31:46 +0000 (22:31 +0000)]
The osigpending, oaccept, orecvfrom and ogetdirentries entries were
accidently being compiled in as standard.  These are part of the
set of unimplemented COMPAT_43 syscall set.

20 years agoWhitespace cleanup.
des [Wed, 10 Dec 2003 22:19:19 +0000 (22:19 +0000)]
Whitespace cleanup.

20 years agoRegen
peter [Wed, 10 Dec 2003 22:18:54 +0000 (22:18 +0000)]
Regen

20 years agoPrevious commit also changed the sendmsg prototype to something more
peter [Wed, 10 Dec 2003 22:16:42 +0000 (22:16 +0000)]
Previous commit also changed the sendmsg prototype to something more
closely matching reality.   I did not actually mean to commit that yet.

20 years agoRemove debugging printf
des [Wed, 10 Dec 2003 22:14:32 +0000 (22:14 +0000)]
Remove debugging printf

20 years agoDocument the simpler -a semantics.
des [Wed, 10 Dec 2003 22:12:06 +0000 (22:12 +0000)]
Document the simpler -a semantics.

20 years agoFix a couple of issues in the interrupt code:
des [Wed, 10 Dec 2003 22:11:51 +0000 (22:11 +0000)]
Fix a couple of issues in the interrupt code:

 - Replace overly-complicated (and buggy) -a logic with a much simpler
   version: -a causes all interrupts to be displayed, otherwise only
   those that have occurred are displayed.  This removes the need for
   any MD code.

 - Instead of just making sure intrcnt is large enough, figure out the
   exact size it needs to be.  We derive nintr from this number, and we
   don't want to risk printing garbage.  Note that on sparc64, we end up
   printing garbage anyway because the names of non-existent interrupts
   are left uninitialized by the kernel.

Tested on: alpha, i386, sparc64

20 years agoUpdate file locations for syscall tables to copy to.
peter [Wed, 10 Dec 2003 22:08:37 +0000 (22:08 +0000)]
Update file locations for syscall tables to copy to.

20 years agos/u_(int|long)\>/unsigned \1/
des [Wed, 10 Dec 2003 21:43:10 +0000 (21:43 +0000)]
s/u_(int|long)\>/unsigned \1/

20 years agoDelete old and unused APIC_IRQ defines.
jhb [Wed, 10 Dec 2003 21:33:07 +0000 (21:33 +0000)]
Delete old and unused APIC_IRQ defines.

20 years agoStill-born. These landed in dev/kbtr in the end.
obrien [Wed, 10 Dec 2003 19:30:52 +0000 (19:30 +0000)]
Still-born.  These landed in dev/kbtr in the end.

20 years agoThe DEV_ACPI option is no longer used, so g/c it.
jhb [Wed, 10 Dec 2003 19:30:42 +0000 (19:30 +0000)]
The DEV_ACPI option is no longer used, so g/c it.

20 years agoUse NAPICID for the maximum number of local APICs rather than MAXCPU when
jhb [Wed, 10 Dec 2003 19:29:39 +0000 (19:29 +0000)]
Use NAPICID for the maximum number of local APICs rather than MAXCPU when
doing the HTT fixup.   This is a step closer to possibly having an apic.ko
module someday.

20 years agoCorrect usage of MAXCPU. The MAXCPU value itself is not a valid CPU ID
jhb [Wed, 10 Dec 2003 19:28:37 +0000 (19:28 +0000)]
Correct usage of MAXCPU.  The MAXCPU value itself is not a valid CPU ID
value as it is a count of maximum values.

Reported by: bde

20 years agoFix a couple of warnings (const a function parameter and change some
dwmalone [Wed, 10 Dec 2003 19:24:11 +0000 (19:24 +0000)]
Fix a couple of warnings (const a function parameter and change some
ints to size_ts to better match the types of variables they are used
with).

Glanced at by: des

20 years agoTrim trailing whitespace.
jhb [Wed, 10 Dec 2003 19:10:27 +0000 (19:10 +0000)]
Trim trailing whitespace.

20 years agoFix typo in comment.
jhb [Wed, 10 Dec 2003 19:08:09 +0000 (19:08 +0000)]
Fix typo in comment.

20 years agointerpvnodelabel can be NULL in mac_test_execve_transition(). This
rwatson [Wed, 10 Dec 2003 18:48:05 +0000 (18:48 +0000)]
interpvnodelabel can be NULL in mac_test_execve_transition().  This
only turned up when running mac_test side by side with a transitioning
policy such as SEBSD.  Make the NULL testing match
mac_test_execve_will_transition(), which already tested the vnode
label pointer for NULL.

Obtained from: TrustedBSD Project
Sponsored by: DARPA, Network Associates Laboratories

20 years agovalidate the argument for multicast routing socket options
ume [Wed, 10 Dec 2003 18:17:05 +0000 (18:17 +0000)]
validate the argument for multicast routing socket options
correctly.

Obtained from: KAME
MFC after: 3 days

20 years agoThis commit was generated by cvs2svn to compensate for changes in r123394,
des [Wed, 10 Dec 2003 16:46:08 +0000 (16:46 +0000)]
This commit was generated by cvs2svn to compensate for changes in r123394,
which included commits to RCS files with non-trunk default branches.

20 years agoMFP4: really fix resp issue, and don't trim trailing spaces (#43727)
des [Wed, 10 Dec 2003 16:46:08 +0000 (16:46 +0000)]
MFP4: really fix resp issue, and don't trim trailing spaces (#43727)

20 years agoAdd reference to standards/55112 for next time SHLIB_MAJOR is bumped.
mikeh [Wed, 10 Dec 2003 16:33:17 +0000 (16:33 +0000)]
Add reference to standards/55112 for next time SHLIB_MAJOR is bumped.

Suggested by: wollman

20 years agoImplement seeking to earlier offsets in gzipfs. This allows my loader
green [Wed, 10 Dec 2003 16:10:34 +0000 (16:10 +0000)]
Implement seeking to earlier offsets in gzipfs.  This allows my loader
to e.g. correctly load all .ko.gz's I've tried, as opposed to messing
up trying to read section headers on some of them.

20 years agoAdd a note about MPTABLE_FORCE_HTT.
jhb [Wed, 10 Dec 2003 15:59:20 +0000 (15:59 +0000)]
Add a note about MPTABLE_FORCE_HTT.

Prompted by: John Stockdale <jstockdale@stanford.edu>

20 years agoChange rc.d/syscons to not need /usr/sbin/ispcvt to exist. This should
peter [Wed, 10 Dec 2003 15:48:49 +0000 (15:48 +0000)]
Change rc.d/syscons to not need /usr/sbin/ispcvt to exist.  This should
make it work for all non-i386 syscons platforms.

20 years agoNursemaid: Fix compilation.
bms [Wed, 10 Dec 2003 15:09:38 +0000 (15:09 +0000)]
Nursemaid: Fix compilation.

Sponsored by: Weyland-Yutani Corporation

20 years agoAdd sticky(8) cross reference.
dds [Wed, 10 Dec 2003 15:08:41 +0000 (15:08 +0000)]
Add sticky(8) cross reference.

PR: docs/60068
Submitted by: Ken Stailey
MFC after: 2 weeks

20 years agoPowerPC-conditional changes:
grehan [Wed, 10 Dec 2003 13:11:03 +0000 (13:11 +0000)]
PowerPC-conditional changes:
 - include ashldi3.c/ashrdi3.c from libc for UFS 64-bit arith
 - pull in syncicache.c from libc, and define _STANDALONE for
   loader usage

20 years ago- Put manh/manl bitfields into correct-sized integral type.
grehan [Wed, 10 Dec 2003 13:04:42 +0000 (13:04 +0000)]
- Put manh/manl bitfields into correct-sized integral type.
- remove XXX normalization bit comment, code is correct.

20 years agoUse .St -susv2 rather than "The Single UNIX specification".
maxim [Wed, 10 Dec 2003 10:52:48 +0000 (10:52 +0000)]
Use .St -susv2 rather than "The Single UNIX specification".

Submitted by: osa

20 years agoGrammar.
ru [Wed, 10 Dec 2003 09:43:53 +0000 (09:43 +0000)]
Grammar.

20 years agoEnable FICL build on powerpc
grehan [Wed, 10 Dec 2003 09:18:42 +0000 (09:18 +0000)]
Enable FICL build on powerpc

20 years agoUpdate libstand filesystems to be in-line with tier-1 platforms.
grehan [Wed, 10 Dec 2003 09:17:01 +0000 (09:17 +0000)]
Update libstand filesystems to be in-line with tier-1 platforms.

20 years ago- Bring Makefile up to rev with sparc64 in terms of config options and rules
grehan [Wed, 10 Dec 2003 09:16:22 +0000 (09:16 +0000)]
- Bring Makefile up to rev with sparc64 in terms of config options and rules
- Move loader relocation up to 0x1C00000. This is in line with OSX bootx,
  and allows more space for boot-time modules/ramdisks without conflicting
  with OpenFirmware's use of RAM

20 years agoDisable floating point on PowerPC for the ficl library.
grehan [Wed, 10 Dec 2003 09:10:54 +0000 (09:10 +0000)]
Disable floating point on PowerPC for the ficl library.

20 years agoDefine 'arch-powerpc'
grehan [Wed, 10 Dec 2003 09:09:38 +0000 (09:09 +0000)]
Define 'arch-powerpc'

20 years agoFICL system-dependent files for powerpc. Taken from the i386 versions,
grehan [Wed, 10 Dec 2003 09:05:08 +0000 (09:05 +0000)]
FICL system-dependent files for powerpc. Taken from the i386 versions,
which were the most relevant.

20 years agoimac revA-D and beige G3 OpenFirmware uses the "ide" string for
grehan [Wed, 10 Dec 2003 08:58:48 +0000 (08:58 +0000)]
imac revA-D and beige G3 OpenFirmware uses the "ide" string for
ATA drives.

20 years ago- removed obsolete ppc_exit/ppc_boot functions
grehan [Wed, 10 Dec 2003 08:55:53 +0000 (08:55 +0000)]
- removed obsolete ppc_exit/ppc_boot functions
- OpenFirmware returns overlapping memory regions. Use a simple
  brute force algorithm to merge these into non-overlapping
  regions. This fixes bugs in reporting of available memory
  and also prevents pages from being added twice in the VM system.

20 years agodon't touch after free.
ume [Wed, 10 Dec 2003 05:01:41 +0000 (05:01 +0000)]
don't touch after free.

20 years agoAdd just enough of i386/include/pcvt_ioctl.h to amd64/include/pcvt_ioctl.h
obrien [Wed, 10 Dec 2003 02:49:17 +0000 (02:49 +0000)]
Add just enough of i386/include/pcvt_ioctl.h to amd64/include/pcvt_ioctl.h
such that 'ispcvt' can build.  Unforunately 'ispcvt' is needed in order for
/etc/rc.d/syscons to run.  This fixes the bug where I could not get my
keymap effective at boot.

20 years agoAdd a short description of the kse_switchin(2) syscall to the kse
marcel [Wed, 10 Dec 2003 02:38:51 +0000 (02:38 +0000)]
Add a short description of the kse_switchin(2) syscall to the kse
manpage and add a kse_switchin link. While here, list kse_thr_interrupt
before kse_wakeup in the MLINKS variable and the synopsis.

20 years agoWrite the thread pointer (val) in the kse mailbox (loc) before we
marcel [Wed, 10 Dec 2003 01:59:23 +0000 (01:59 +0000)]
Write the thread pointer (val) in the kse mailbox (loc) before we
set the new context in kse_switchin(2). This allows us to return
an error to the calling context when the suword() fails.

20 years agoaccept() returns a file descriptor when it succeeds which is very
deischen [Tue, 9 Dec 2003 23:40:27 +0000 (23:40 +0000)]
accept() returns a file descriptor when it succeeds which is very
likely to be non-zero.  When leaving the cancellation point, check
the return value against -1 to see if cancellation should be
checked.  While I'm here, make the same change to connect() just
to be consisitent.

Pointed out by: davidxu

20 years agoAdjust an assertion for the TDF_TSNOBLOCK race handling in
jhb [Tue, 9 Dec 2003 21:14:31 +0000 (21:14 +0000)]
Adjust an assertion for the TDF_TSNOBLOCK race handling in
turnstile_unpend().  A racing thread that does not have TDI_LOCK set may
either be running on another CPU or it may be sitting on a run queue if it
was preempted during the very small window in turnstile_wait() between
unlocking the turnstile chain lock and locking sched_lock.

20 years agoAssert that the we never give a thread a NULL turnstile when waking it up.
jhb [Tue, 9 Dec 2003 21:09:54 +0000 (21:09 +0000)]
Assert that the we never give a thread a NULL turnstile when waking it up.

20 years agoRevert the previous race fix and replace it with a more general fix. The
jhb [Tue, 9 Dec 2003 21:09:04 +0000 (21:09 +0000)]
Revert the previous race fix and replace it with a more general fix.  The
case of a turnstile having no threads is just one instance of the more
general case where the thread we are examining has been partially awakened
already in that it has been removed from the turnstile's blocked list but
still has TDI_LOCK set.  We detect that case by checking to see if the
thread has already had a turnstile reassigned to it.

20 years agoAdd back the national support, this was removed by accident earlier.
sos [Tue, 9 Dec 2003 19:13:50 +0000 (19:13 +0000)]
Add back the national support, this was removed by accident earlier.

20 years agoAdd support for Moxa Technologies' Smartio CP-104UL/PCI 4S RS232 card.
obrien [Tue, 9 Dec 2003 17:31:25 +0000 (17:31 +0000)]
Add support for Moxa Technologies' Smartio CP-104UL/PCI 4S RS232 card.

Submitted by: Guido Falsi <mad@madpilot.net>

20 years agoAdd two additional headers for amd64 and x86.
kan [Tue, 9 Dec 2003 16:31:30 +0000 (16:31 +0000)]
Add two additional headers for amd64 and x86.

Reported by: Ollivier Robert

20 years agoRemove duplicate script for locore.o
gallatin [Tue, 9 Dec 2003 15:48:20 +0000 (15:48 +0000)]
Remove duplicate script for locore.o

20 years agoRemove redundant declaration of ddb_trap
gallatin [Tue, 9 Dec 2003 15:46:33 +0000 (15:46 +0000)]
Remove redundant declaration of  ddb_trap

20 years agoRemove an unused struct definition.
deischen [Tue, 9 Dec 2003 15:18:40 +0000 (15:18 +0000)]
Remove an unused struct definition.

20 years agoAdd cancellation points for accept() and connect().
deischen [Tue, 9 Dec 2003 15:16:27 +0000 (15:16 +0000)]
Add cancellation points for accept() and connect().

20 years agopmap_query_bit() should return false if the bit is not set.
gallatin [Tue, 9 Dec 2003 14:47:33 +0000 (14:47 +0000)]
pmap_query_bit() should return false if the bit is not set.

Reviewed by: grehan

20 years agoUse the "shut-down" and "reset-all" Forth procedures to halt and
gallatin [Tue, 9 Dec 2003 14:45:39 +0000 (14:45 +0000)]
Use the "shut-down" and "reset-all" Forth procedures to halt and
reboot, as calling OF_exit() just hangs a mac.

FreeBSD on my G4 800Mhz mac behaves identically to OSX for halt
and reboot now.

Reviewed by: grehan  (who also supplied the concept and sample code)

20 years agoMake breakpoint() actually break into ddb.
gallatin [Tue, 9 Dec 2003 13:37:31 +0000 (13:37 +0000)]
Make breakpoint() actually break into ddb.

Reviewed by: grehan

20 years agoTake a stab at fixing some of the macro-nightmare.
mtm [Tue, 9 Dec 2003 11:20:01 +0000 (11:20 +0000)]
Take a stab at fixing some of the macro-nightmare.
PTHREAD_NEW_STATE should work as expected now: a thread
marked PS_RUNNING will get sent a SIGTHR.
Still more cleanups necessary.

20 years agoFix the wrapper function around signals so that a signal handling
mtm [Tue, 9 Dec 2003 11:12:11 +0000 (11:12 +0000)]
Fix the wrapper function around signals so that a signal handling
thread on one of the mutex or condition variable queues is removed
from those queues before the real signal handler is called.

20 years agoUgghh, cvs add the functions necessary to lock the global signal action
mtm [Tue, 9 Dec 2003 11:06:55 +0000 (11:06 +0000)]
Ugghh, cvs add the functions necessary to lock the global signal action
table.

20 years agoo Add a wrapper around sigaction(2), so we can insert our own wrapper
mtm [Tue, 9 Dec 2003 11:04:36 +0000 (11:04 +0000)]
o Add a wrapper around sigaction(2), so we can insert our own wrapper
  around signals.
o Lock the process global signal action table.

20 years agoEnable cancellation points around some syscalls.
mtm [Tue, 9 Dec 2003 11:01:09 +0000 (11:01 +0000)]
Enable cancellation points around some syscalls.

20 years agoDon't panic for misalignment traps when the onfault handler is set.
marcel [Tue, 9 Dec 2003 09:52:14 +0000 (09:52 +0000)]
Don't panic for misalignment traps when the onfault handler is set.
Not all transfers between kernel and user space are byte oriented
and thus alignment safe. Especially fuword*() and suword*() are
sensitive to alignment but in general more optimal than block copies.
By catching the misalignment trap we avoid pessimizing the common
case of properly aligned memory accesses which we would do if we
were to use byte copies or adding tests for proper alignment.

Note that the expectation that the kernel produces aligned pointers
is unchanged. This change therefore relates to possible unaligned
pointers generated in userland.

20 years agoGrammar
mtm [Tue, 9 Dec 2003 09:04:45 +0000 (09:04 +0000)]
Grammar

20 years agoo Rename devfs_link() to make_symlink() and turn it into a generic
mtm [Tue, 9 Dec 2003 08:51:11 +0000 (08:51 +0000)]
o Rename devfs_link() to make_symlink() and turn it into a generic
  symlinking routine.
o Modify rc.d/jail to create its own symlink relative to the jail's
  filesystem

20 years agoFix the build of libski now that we use the "official" MADT table
marcel [Tue, 9 Dec 2003 08:35:17 +0000 (08:35 +0000)]
Fix the build of libski now that we use the "official" MADT table
definitions. Those are slightly different than the ones we used
before ACPI-CA 20031203 got imported. No structural or functional
change.

20 years agoIf fdescfs(5) is mounted in a jail unmount it before devfs(5).
mtm [Tue, 9 Dec 2003 08:32:43 +0000 (08:32 +0000)]
If fdescfs(5) is mounted in a jail unmount it before devfs(5).

Submitted by: "Scot W. Hetzel" <hetzelsw@westbend.net>

20 years agoIf rc.d/mountcritlocal is unable to mount local filesystems,
mtm [Tue, 9 Dec 2003 08:17:33 +0000 (08:17 +0000)]
If rc.d/mountcritlocal is unable to mount local filesystems,
then immediately terminate the shell (during boot this
also terminates the parent rc(8) shell). This was the pre-rcNG behaviour.

Also, remove an extraneous mount /.

PR: conf/57659
Submitted by: yar (with modifications)

20 years agoRedirect rc(8) output to a file in the jail's /var/log directory.
mtm [Tue, 9 Dec 2003 08:09:04 +0000 (08:09 +0000)]
Redirect rc(8) output to a file in the jail's /var/log directory.
Previously it was redirected to /dev/null.

Submitted by: phk

20 years agoOlder versions of the intersil firmware is a lot slower than newer
imp [Tue, 9 Dec 2003 07:41:07 +0000 (07:41 +0000)]
Older versions of the intersil firmware is a lot slower than newer
versions of the firmware.  It responds more slowly to commands, and we
bogusly failed them.  We assume that all versions of the intersil
firmware before 1.0 are 10 times slower and will give it 10x the time
to finish.

# for 5.2 we should always just assume 5s.

20 years agoDeclare gre(4) as being of IFT_TUNNEL, Like God Intended.
bms [Tue, 9 Dec 2003 06:39:25 +0000 (06:39 +0000)]
Declare gre(4) as being of IFT_TUNNEL, Like God Intended.

Suggested by: fenner

20 years agoUse sbufs instead of snprintf for parsing debug strings.
njl [Tue, 9 Dec 2003 06:29:57 +0000 (06:29 +0000)]
Use sbufs instead of snprintf for parsing debug strings.

20 years agoSync these two files.
obrien [Tue, 9 Dec 2003 05:40:06 +0000 (05:40 +0000)]
Sync these two files.

20 years agoDelete madt.h since definitions are now used from actbl.h
njl [Tue, 9 Dec 2003 03:18:06 +0000 (03:18 +0000)]
Delete madt.h since definitions are now used from actbl.h