]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoAdd a couple of ioctls for modifying MBR's and the rest of sector0.
phk [Sun, 29 Dec 2002 14:56:04 +0000 (14:56 +0000)]
Add a couple of ioctls for modifying MBR's and the rest of sector0.

21 years agoUse a timeout of one second while we wait for the vnode washer,
phk [Sun, 29 Dec 2002 11:18:25 +0000 (11:18 +0000)]
Use a timeout of one second while we wait for the vnode washer,
this prevents a potential race and makes the system a little bit
less jerky under extreme loads.

21 years agoSave 16 bytes per mutex if MUTEX_PROFILING is not defined.
phk [Sun, 29 Dec 2002 11:14:41 +0000 (11:14 +0000)]
Save 16 bytes per mutex if MUTEX_PROFILING is not defined.

MUTEX_PROFILING is in opt_global.h, so this does not introduce a risk of
variant structure sizes unless foreign kernel modules are used.

This saved 16 bytes per vnode and 16 bytes per vm object for a total of
4MB on a 2GB machine.

Idea from: alc

21 years agoVnodes pull in 800-900 bytes these days, all things counted, so we need
phk [Sun, 29 Dec 2002 10:39:05 +0000 (10:39 +0000)]
Vnodes pull in 800-900 bytes these days, all things counted, so we need
to treat desiredvnodes much more like a limit than as a vague concept.

On a 2GB RAM machine where desired vnodes is 130k, we run out of
kmem_map space when we hit about 190k vnodes.

If we wake up the vnode washer in getnewvnode(), sleep until it is done,
so that it has a chance to offer us a washed vnode.  If we don't sleep
here we'll just race ahead and allocate yet a vnode which will never
get freed.

In the vnodewasher, instead of doing 10 vnodes per mountpoint per
rotation, do 10% of the vnodes distributed evenly across the
mountpoints.

21 years agoThere is some sort of race/deadlock which I have not identified
phk [Sun, 29 Dec 2002 10:32:16 +0000 (10:32 +0000)]
There is some sort of race/deadlock which I have not identified
here.  It manifests itself by sendmail hanging in "fifoow" during
boot on a diskless machine with sendmail disabled.

Giving the sleep a 1sec timout breaks the deadlock, but does not solve
the underlying problem.

XXX comment applied.

21 years agoUse memset instead of __builtin_memset. Apparently there's an inline
jake [Sun, 29 Dec 2002 08:37:11 +0000 (08:37 +0000)]
Use memset instead of __builtin_memset.  Apparently there's an inline
memset in libkern which causes problems; why that's there is beyond me.

21 years agoDiff-reduction WRT crypto telnet; replace "#if 0" code removed by an
markm [Sun, 29 Dec 2002 07:36:43 +0000 (07:36 +0000)]
Diff-reduction WRT crypto telnet; replace "#if 0" code removed by an
earlier invocation of unifdef(1).

21 years agoReduce the number of times that we acquire and release the page queues
alc [Sun, 29 Dec 2002 07:17:06 +0000 (07:17 +0000)]
Reduce the number of times that we acquire and release the page queues
lock by making vm_page_rename()'s caller, rather than vm_page_rename(),
responsible for acquiring it.

21 years agoo Fix ordering of typedefs.
mike [Sun, 29 Dec 2002 03:38:48 +0000 (03:38 +0000)]
o Fix ordering of typedefs.
o Improve comment about namespace pollution.
o Improve CPP style.

Submitted by: bde

21 years agosa_handler is in the POSIX namespace (5.0-R candidate).
mike [Sun, 29 Dec 2002 01:50:22 +0000 (01:50 +0000)]
sa_handler is in the POSIX namespace (5.0-R candidate).

21 years agoo Add typedef for off_t, pid_t, and useconds_t.
mike [Sun, 29 Dec 2002 01:07:55 +0000 (01:07 +0000)]
o Add typedef for off_t, pid_t, and useconds_t.
o Use useconds_t where appropriate.
o Fix a bug in typedef for uid_t (5.0-R candidate).

21 years agoUse `useconds_t' instead of `u_int' or `unsigned int' where appropriate.
mike [Sun, 29 Dec 2002 00:59:09 +0000 (00:59 +0000)]
Use `useconds_t' instead of `u_int' or `unsigned int' where appropriate.

21 years agoUse the meaningful mnemonics for ancillary state registers now that gas
jake [Sun, 29 Dec 2002 00:23:48 +0000 (00:23 +0000)]
Use the meaningful mnemonics for ancillary state registers now that gas
is invoked properly to understand them.

%asr19 -> %gsr
%asr20 -> %set_softint
%asr21 -> %clear_softint

21 years agoForgot this file in previous commit.
jake [Sat, 28 Dec 2002 23:58:18 +0000 (23:58 +0000)]
Forgot this file in previous commit.

21 years ago- Moved storing %g1-%g5 in the trapframe until after interrupts are enabled.
jake [Sat, 28 Dec 2002 23:57:52 +0000 (23:57 +0000)]
- Moved storing %g1-%g5 in the trapframe until after interrupts are enabled.
- Restore %g6 and %g7 for kernel traps if we are returning to prom code.
  This allows complex traps (ones that call into C code) to be handled from
  the prom.

21 years agoMap VAPPEND to VWRITE since the ugidfw rule syntax doesn't specifically
rwatson [Sat, 28 Dec 2002 23:41:18 +0000 (23:41 +0000)]
Map VAPPEND to VWRITE since the ugidfw rule syntax doesn't specifically
address the append access mode.

Reported by: "Daniel C. Sobral" <dcs@newsguy.com>
Obtained from: TrustedBSD Project
Sponosred by: DARPA, Network Associates Laboratories

MFC Candidate.

21 years agoAdd 'swapctl' - as a hardlink to swapon/swapoff, and augment swapon with
dillon [Sat, 28 Dec 2002 23:39:47 +0000 (23:39 +0000)]
Add 'swapctl' - as a hardlink to swapon/swapoff, and augment swapon with
swapctl functionality.  The idea is to create a swapctl command that is
fairly close to the OpenBSD and NetBSD version.  FreeBSD does not implement
swap priority (and it would be a mistake if we did) so we didn't bother with
that part of it.

Submitted by: Eirik Nygaard <eirikn@bluezone.no>
Augmented by: dillon (extensively)
Reviewed by: David Schultz <dschultz@uclink.Berkeley.EDU>

21 years agoPass 0 in %o1 to tl0_trap for all non-interrupt traps. This will be used
jake [Sat, 28 Dec 2002 23:34:21 +0000 (23:34 +0000)]
Pass 0 in %o1 to tl0_trap for all non-interrupt traps.  This will be used
to pass the pil when tl0_trap also handles interrupts.

21 years agoSince our default boot block now supports UFS1 and UFS2 even on
rwatson [Sat, 28 Dec 2002 23:33:09 +0000 (23:33 +0000)]
Since our default boot block now supports UFS1 and UFS2 even on
i386, remove the seatbelt preventing users from setting the UFS2 flag
on the root file system on i386.  This seatbelt did not exist on
other platforms.

MFC candidate.

21 years agoDon't put a newline in KTR traces.
jake [Sat, 28 Dec 2002 23:22:22 +0000 (23:22 +0000)]
Don't put a newline in KTR traces.

21 years agoAdd a tunable kern.smp.disabled for disabling explicitly smp on an smp
jake [Sat, 28 Dec 2002 23:21:13 +0000 (23:21 +0000)]
Add a tunable kern.smp.disabled for disabling explicitly smp on an smp
kernel.

21 years agoAssert that the page queues lock rather than Giant is held in
alc [Sat, 28 Dec 2002 22:49:37 +0000 (22:49 +0000)]
Assert that the page queues lock rather than Giant is held in
vm_page_flag_clear().

21 years ago - Use VM_ALLOC_WIRED.
alc [Sat, 28 Dec 2002 22:47:45 +0000 (22:47 +0000)]
 - Use VM_ALLOC_WIRED.
 - Hold the page queues lock around calls to vm_page_flag_clear() and
   vm_page_wakeup().

21 years agoAllow lint-like tools to override DROP_GIANT and friends.
phk [Sat, 28 Dec 2002 22:29:10 +0000 (22:29 +0000)]
Allow lint-like tools to override DROP_GIANT and friends.

Apply parens around macro arguments.

21 years agoKASSERT that vop_revoke() gets a VCHR.
phk [Sat, 28 Dec 2002 22:27:14 +0000 (22:27 +0000)]
KASSERT that vop_revoke() gets a VCHR.

21 years agoIt is bad style to define the same structure in multiple header
phk [Sat, 28 Dec 2002 22:17:29 +0000 (22:17 +0000)]
It is bad style to define the same structure in multiple header
files which might be included together.

Things like debuggers and lint-like programs get their knickers in
a twist (rightly so one might add) when they find different locations
for the same named struct depending on which .h file were included
first.

This is a stellar example of Very Bad Thinking on the part of the
standards dudes who wrote that both sys/uio.h and sys/socket.h
should define struct iovec the same way.

Fix this by putting struct iovec into its own miniature sys/_iovec.h
file and #include that from sys/socket.h and sys/uio.h.

Sensible people could just put iovec into sys/_types.h but there
is probably some standard or other which will be violated if we
did something that horrible.

21 years agoRemove unused cdevsw_ALLOCSTART macro.
phk [Sat, 28 Dec 2002 21:47:43 +0000 (21:47 +0000)]
Remove unused cdevsw_ALLOCSTART macro.

21 years agoRemove cdevw_add() calls, they are deprecated.
phk [Sat, 28 Dec 2002 21:40:20 +0000 (21:40 +0000)]
Remove cdevw_add() calls, they are deprecated.

21 years agoRemove cdevsw_add calls, they are deprecated.
phk [Sat, 28 Dec 2002 21:39:46 +0000 (21:39 +0000)]
Remove cdevsw_add calls, they are deprecated.

21 years agoRemove calls to cdevsw_add(), they are dreprecated on this branch now.
phk [Sat, 28 Dec 2002 21:39:04 +0000 (21:39 +0000)]
Remove calls to cdevsw_add(), they are dreprecated on this branch now.

21 years agovm_pager_put_pages() takes VM_PAGER_* flags, not OBJPC_* flags. It just
dillon [Sat, 28 Dec 2002 21:15:39 +0000 (21:15 +0000)]
vm_pager_put_pages() takes VM_PAGER_* flags, not OBJPC_* flags.   It just
so happens that OBJPC_SYNC has the same value as VM_PAGER_PUT_SYNC so no
harm done.  But fix it :-)

No operational changes.

MFC after: 1 day

21 years agoHold the page queues lock around calls to vm_page_flag_clear() and
alc [Sat, 28 Dec 2002 21:14:44 +0000 (21:14 +0000)]
Hold the page queues lock around calls to vm_page_flag_clear() and
vm_page_wakeup().

21 years agoIndirectly pull in declaration for M_IFADDR.
hsu [Sat, 28 Dec 2002 21:09:38 +0000 (21:09 +0000)]
Indirectly pull in declaration for M_IFADDR.

21 years agoAllow the VM object flushing code to cluster. When the filesystem syncer
dillon [Sat, 28 Dec 2002 21:03:42 +0000 (21:03 +0000)]
Allow the VM object flushing code to cluster.  When the filesystem syncer
comes along and flushes a file which has been mmap()'d SHARED/RW, with
dirty pages, it was flushing the underlying VM object asynchronously,
resulting in thousands of 8K writes.  With this change the VM Object flushing
code will cluster dirty pages in 64K blocks.

Note that until the low memory deadlock issue is reviewed, it is not safe
to allow the pageout daemon to use this feature.  Forced pageouts still
use fs block size'd ops for the moment.

MFC after: 3 days

21 years agoAbstract-out the constants for the sequential heuristic.
dillon [Sat, 28 Dec 2002 20:37:50 +0000 (20:37 +0000)]
Abstract-out the constants for the sequential heuristic.

No operational changes.

MFC after: 1 day

21 years agoAbstract-out the constants for the sequential heuristic.
dillon [Sat, 28 Dec 2002 20:28:10 +0000 (20:28 +0000)]
Abstract-out the constants for the sequential heuristic.

No operational changes.

MFC after: 1 day

21 years agoSome Pa fixes from Tom, plus a little rearranging of the paragraph by me.
dillon [Sat, 28 Dec 2002 19:08:11 +0000 (19:08 +0000)]
Some Pa fixes from Tom, plus a little rearranging of the paragraph by me.

Submitted by: Tom Rhodes

21 years agoTwo changes to kmem_malloc():
alc [Sat, 28 Dec 2002 19:03:54 +0000 (19:03 +0000)]
Two changes to kmem_malloc():
 - Use VM_ALLOC_WIRED.
 - Perform vm_page_wakeup() after pmap_enter(), like we do everywhere else.

21 years agoSynchronize old rc.diskless scripts from new rc.d/ scripts
dillon [Sat, 28 Dec 2002 18:47:53 +0000 (18:47 +0000)]
Synchronize old rc.diskless scripts from new rc.d/ scripts

21 years agoChange 'mfsroot' filename in mfsroot-small floppies from 'mfsroot-small'
nyan [Sat, 28 Dec 2002 16:23:42 +0000 (16:23 +0000)]
Change 'mfsroot' filename in mfsroot-small floppies from 'mfsroot-small'
to 'mfsroot'.

This is MFC candidate.

Noticed by: Toru Morimoto <too@os.gulf.or.jp>

21 years ago(forced)
sobomax [Sat, 28 Dec 2002 16:03:46 +0000 (16:03 +0000)]
(forced)

Previous delta was a subject of:

MFC after: 1 week

21 years agoAdd support for IC Book Labs Dreadnought x16 Lite.
sobomax [Sat, 28 Dec 2002 15:45:38 +0000 (15:45 +0000)]
Add support for IC Book Labs Dreadnought x16 Lite.

Approved by: re

21 years agoChange ACPI make_dev() calls to use UID_ and GID_ constants rather
rwatson [Sat, 28 Dec 2002 14:58:50 +0000 (14:58 +0000)]
Change ACPI make_dev() calls to use UID_ and GID_ constants rather
than hard-coded uids and gids.

Switch the device to a group of wheel instead of operator.

Narrow down the permissions on the device to require root privilege
to manipulate the system power state.  It may be that we can broaden
access to the device after review of the access control in ACPI.

Submitted by: kris
Reviewed by: takawata

21 years agoUse VM_ALLOC_WIRED in pmap_pinit().
alc [Sat, 28 Dec 2002 08:10:29 +0000 (08:10 +0000)]
Use VM_ALLOC_WIRED in pmap_pinit().

21 years agoPut the if_bge driver on the second floppy
scottl [Sat, 28 Dec 2002 06:23:44 +0000 (06:23 +0000)]
Put the if_bge driver on the second floppy

21 years agoAdd the if_bge driver. I can't find any reason why it's not here, and it's
scottl [Sat, 28 Dec 2002 06:22:17 +0000 (06:22 +0000)]
Add the if_bge driver.  I can't find any reason why it's not here, and it's
pretty common on Dell servers and other high-end boxes.

21 years agoTrim left-over and unused vop_refreshlabel() bits from devfs.
rwatson [Sat, 28 Dec 2002 05:39:25 +0000 (05:39 +0000)]
Trim left-over and unused vop_refreshlabel() bits from devfs.

Reported by: bde

21 years agoFix typos in comment.
deischen [Sat, 28 Dec 2002 05:20:03 +0000 (05:20 +0000)]
Fix typos in comment.

Submitted by: Craig Rodrigues <rodrigc@attbi.com>

21 years agoRemove useless local variable from raw_input().
kbyanc [Sat, 28 Dec 2002 02:29:19 +0000 (02:29 +0000)]
Remove useless local variable from raw_input().

Sponsored by: NTT Multimedia Communications Labs
MFC after: 3 days

21 years agoAdd code to ddb to allow backtracing an arbitrary thread.
julian [Sat, 28 Dec 2002 01:23:07 +0000 (01:23 +0000)]
Add code to ddb to allow backtracing an arbitrary thread.
(show thread {address})

Remove the IDLE kse state and replace it with a change in
the way threads sahre KSEs. Every KSE now has a thread, which is
considered its "owner" however a KSE may also be lent to other
threads in the same group to allow completion of in-kernel work.
n this case the owner remains the same and the KSE will revert to the
owner when the other work has been completed.

All creations of upcalls etc. is now done from
kse_reassign() which in turn is called from mi_switch or
thread_exit(). This means that special code can be removed from
msleep() and cv_wait().

kse_release() does not leave a KSE with no thread any more but
converts the existing thread into teh KSE's owner, and sets it up
for doing an upcall. It is just inhibitted from being scheduled until
there is some reason to do an upcall.

Remove all trace of the kse_idle queue since it is no-longer needed.
"Idle" KSEs are now on the loanable queue.

21 years agoAssert that the page queues lock is held in pmap_testbit().
alc [Sat, 28 Dec 2002 00:19:40 +0000 (00:19 +0000)]
Assert that the page queues lock is held in pmap_testbit().

21 years ago - Change vm_object_page_collect_flush() to assert rather than
alc [Fri, 27 Dec 2002 20:16:13 +0000 (20:16 +0000)]
 - Change vm_object_page_collect_flush() to assert rather than
   acquire the page queues lock.
 - Acquire the page queues lock in vm_object_page_clean().

21 years agoDefine UMA_MD_SMALL_ALLOC so that uma_small_alloc and uma_small_free will
jake [Fri, 27 Dec 2002 19:31:26 +0000 (19:31 +0000)]
Define UMA_MD_SMALL_ALLOC so that uma_small_alloc and uma_small_free will
be used for zones that allocate objects of less 1 page.  The biggest advantage
of this is that all of a sudden the majority of kernel malloc-ed data doesn't
need kva allocated for it.  Besides microbenchmarks I haven't seen a measurable
performance improvement from doing this.

21 years agoTeach /dev/kmem about direct mapped addresses.
jake [Fri, 27 Dec 2002 19:18:04 +0000 (19:18 +0000)]
Teach /dev/kmem about direct mapped addresses.

Note that a better solution for how to make kernacc work for direct mapped
addresses is needed for all platforms that use them.

21 years agoRe-add MNT_ACLS to the list of "updateable" mount flags, per our
rwatson [Fri, 27 Dec 2002 18:20:16 +0000 (18:20 +0000)]
Re-add MNT_ACLS to the list of "updateable" mount flags, per our
documentation.  Generally, you really shouldn't twiddle the flag,
but there are sensible scenarios where one might.

Obtained from: TrustedBSD Project

21 years agoUse OPERATOR instead of WHEEL for make_dev(), use 0640 instead of 0644
rwatson [Fri, 27 Dec 2002 17:52:16 +0000 (17:52 +0000)]
Use OPERATOR instead of WHEEL for make_dev(), use 0640 instead of 0644
for disk devices to be consistent with other storage devices.

Submitted by: kris
Reviewed by: scottl

21 years agoUse UID_ and GID_ constants instead of hard-coded numeric values
rwatson [Fri, 27 Dec 2002 17:50:39 +0000 (17:50 +0000)]
Use UID_ and GID_ constants instead of hard-coded numeric values
with make_dev().  Use OPERATOR instead of implicit WHEEL to match
other storage devices.  Use a mode of 0640 to be consistent
with other storage devices.

Submitted by: kris
Reviewed by: scottl

21 years agoBridged packets are supplied to the firewall with their IP header
iedowse [Fri, 27 Dec 2002 17:43:25 +0000 (17:43 +0000)]
Bridged packets are supplied to the firewall with their IP header
in network byte order, but icmp_error() expects the IP header to
be in host order and the code here did not perform the necessary
swapping for the bridged case. This bug causes an "icmp_error: bad
length" panic when certain length IP packets (e.g. ip_len == 0x100)
are rejected by the firewall with an ICMP response.

MFC after: 3 days

21 years agoOops, I misread the purpose of the NULL check in EH_RESTORE() in
iedowse [Fri, 27 Dec 2002 17:15:16 +0000 (17:15 +0000)]
Oops, I misread the purpose of the NULL check in EH_RESTORE() in
revision 1.62. It was checking for M_PREPEND() failing, not for the
case of a NULL mbuf pointer being supplied to the macro. Back out
that revision, and fix the NULL dereference by not calling EH_RESTORE()
in the case where the mbuf pointer is NULL because the firewall
rejected the packet.

21 years agoImprove consistency between devfs and MAKEDEV: use UID_ROOT and
rwatson [Fri, 27 Dec 2002 16:54:44 +0000 (16:54 +0000)]
Improve consistency between devfs and MAKEDEV: use UID_ROOT and
GID_WHEEL instead of UID_BIN and GID_BIN for /dev/fd/* entries.

Submitted by: kris

21 years agoUse UID_ROOT and GID_WHEEL for uid/gid argments to make_dev().
rwatson [Fri, 27 Dec 2002 16:44:11 +0000 (16:44 +0000)]
Use UID_ROOT and GID_WHEEL for uid/gid argments to make_dev().
Remove the setgid bit from the tga device (?).
Synchronize mode with modes used for related frame buffer devices
in MAKEDEV (tga doesn't appear in MAKEDEV).

Submitted by: kris

21 years agoMake use of UID_ROOT, GID_WHEEL for make_dev() arguments.
rwatson [Fri, 27 Dec 2002 16:40:54 +0000 (16:40 +0000)]
Make use of UID_ROOT, GID_WHEEL for make_dev() arguments.
Remove the setgid bit from the 3dfx device (?).
Synchronize permissions with the values in MAKEDEV for consistency.

Submitted by: kris

21 years agoSynchronize mode for ppsX devices to that found previously in MAKEDEV
rwatson [Fri, 27 Dec 2002 16:34:12 +0000 (16:34 +0000)]
Synchronize mode for ppsX devices to that found previously in MAKEDEV
for consistency.

Submitted by: kris

21 years agoSynchronize make_dev() for pcfclock devices to the values in MAKEDEV
rwatson [Fri, 27 Dec 2002 16:32:10 +0000 (16:32 +0000)]
Synchronize make_dev() for pcfclock devices to the values in MAKEDEV
for consistency.

Submitted by: kris

21 years agoRemove S_IROTH from the make_dev() lines for iir-related devices. This
rwatson [Fri, 27 Dec 2002 16:28:31 +0000 (16:28 +0000)]
Remove S_IROTH from the make_dev() lines for iir-related devices.  This
improves protection consistency with other storage devices (generally
root:operator,660).  This driver appears not to have an active
maintainer.

Submitted by: kris

21 years agoFix a bug introduced by revision 1.59 that would cause an immediate
iedowse [Fri, 27 Dec 2002 16:27:45 +0000 (16:27 +0000)]
Fix a bug introduced by revision 1.59 that would cause an immediate
NULL dereference if a bridged packet was rejected by ipfw.

21 years agoMFbed: translation updates:
ue [Fri, 27 Dec 2002 15:00:06 +0000 (15:00 +0000)]
MFbed: translation updates:
hardware/alpha/proc-alpha.sgml: 1.49  -> 1.51
installation/common/install.sgml: 1.16  -> 1.17
installation/common/trouble.sgml: 1.8   -> 1.10
relnotes/common/new.sgml: 1.466 -> 1.467

21 years agoenglish(4) police.
schweikh [Fri, 27 Dec 2002 12:15:40 +0000 (12:15 +0000)]
english(4) police.

21 years agoUse three UMA zones for FFS/UFS inodes instead of malloc space.
phk [Fri, 27 Dec 2002 11:05:05 +0000 (11:05 +0000)]
Use three UMA zones for FFS/UFS inodes instead of malloc space.
Since inodes are currently 144 bytes, this will save 112 bytes per
inode.  This can amount to up to 10MByte on large systems.

21 years agoMove the allocation of the inode contents into ffs_vfsops.c rather than
phk [Fri, 27 Dec 2002 10:23:03 +0000 (10:23 +0000)]
Move the allocation of the inode contents into ffs_vfsops.c rather than
passing malloc types around.

21 years agoPOLA dictates that in the file designated with the -f option
ru [Fri, 27 Dec 2002 10:09:04 +0000 (10:09 +0000)]
POLA dictates that in the file designated with the -f option
argument, leading whitespace and empty lines be ignored, and
the `#' character marks the rest of the line as a comment.

PR: bin/45958
MFC after: 3 days

21 years agoMake ffs_mountfs() static.
phk [Fri, 27 Dec 2002 10:06:37 +0000 (10:06 +0000)]
Make ffs_mountfs() static.

Remove the malloctype from the ufs mount structure, instead add a callback
to the storage method for freeing inodes: UFS_IFREE().

Add vfs_ifree() method function which frees an inode.

Unvariablelize the malloc type used for allocating inodes.

21 years agoSilence casting away constness warnings.
alfred [Fri, 27 Dec 2002 08:43:40 +0000 (08:43 +0000)]
Silence casting away constness warnings.

Make cgetmatch's locals const.
Make cdbget take a const string and copy it into a buffer.

21 years agoLock filedesc while performing a range check on the file descriptor.
alfred [Fri, 27 Dec 2002 08:39:42 +0000 (08:39 +0000)]
Lock filedesc while performing a range check on the file descriptor.

Reviewed by: alc

21 years agoFix a small typo.
tjr [Fri, 27 Dec 2002 08:28:53 +0000 (08:28 +0000)]
Fix a small typo.

21 years agoSlight tuning if teh KSE man page to indicate some functionality is
julian [Fri, 27 Dec 2002 08:21:15 +0000 (08:21 +0000)]
Slight tuning if teh KSE man page to indicate some functionality is
not yet inplemented and to clear up some wording.

21 years agoRemove the "ascii" attribute from the sysctls so that "sysctl -a" will
phk [Fri, 27 Dec 2002 07:58:59 +0000 (07:58 +0000)]
Remove the "ascii" attribute from the sysctls so that "sysctl -a" will
skip them.

21 years agoHold the page queues lock when calling vm_page_flag_clear().
alc [Fri, 27 Dec 2002 06:52:32 +0000 (06:52 +0000)]
Hold the page queues lock when calling vm_page_flag_clear().

21 years agoIncrease the scope of the page queues lock in phys_pager_getpages().
alc [Fri, 27 Dec 2002 06:09:56 +0000 (06:09 +0000)]
Increase the scope of the page queues lock in phys_pager_getpages().

21 years agomount_msdos(8) > mount_msdosfs(8) as they have different names in 4.X/5.X.
trhodes [Fri, 27 Dec 2002 04:06:04 +0000 (04:06 +0000)]
mount_msdos(8) > mount_msdosfs(8) as they have different names in 4.X/5.X.
Correct link to the handbook (typo).

Pointed out by: chris

21 years agoWhitespace created in Makefile from previous commit.
trhodes [Fri, 27 Dec 2002 03:17:39 +0000 (03:17 +0000)]
Whitespace created in Makefile from previous commit.

21 years agoAdd a manual page for msdosfs (msdos.5) and connect it to the build.
trhodes [Fri, 27 Dec 2002 03:17:02 +0000 (03:17 +0000)]
Add a manual page for msdosfs (msdos.5) and connect it to the build.

Quick review by: chris (older version)

21 years agoImplement uma_small_alloc and uma_small_free. Not yet used.
jake [Fri, 27 Dec 2002 03:11:29 +0000 (03:11 +0000)]
Implement uma_small_alloc and uma_small_free.  Not yet used.

21 years ago- Use direct mapped addresses for the message buffer, for the crash dump
jake [Fri, 27 Dec 2002 01:50:29 +0000 (01:50 +0000)]
- Use direct mapped addresses for the message buffer, for the crash dump
  mappings, and for pmap_map which is used to map the vm_page structures.
- Don't allocate kva space for any of the above.

21 years agoTeach libkvm to deal with direct mapped addresses.
jake [Fri, 27 Dec 2002 01:45:05 +0000 (01:45 +0000)]
Teach libkvm to deal with direct mapped addresses.

21 years agoBe more consistent with "static".
tjr [Fri, 27 Dec 2002 01:01:03 +0000 (01:01 +0000)]
Be more consistent with "static".

21 years agoLong chain of calls starting with bridge_on(), going through IPv6, and
hsu [Fri, 27 Dec 2002 00:24:35 +0000 (00:24 +0000)]
Long chain of calls starting with bridge_on(), going through IPv6, and
ending up at ifa_ifwithdstaddr() could lead to a recursive lock of
the ifnet list mutex.

21 years agowhite-space changes
phk [Thu, 26 Dec 2002 21:02:50 +0000 (21:02 +0000)]
white-space changes

21 years agoUse a mutex assert to document our locking circumstances.
phk [Thu, 26 Dec 2002 20:48:22 +0000 (20:48 +0000)]
Use a mutex assert to document our locking circumstances.

21 years agoWe should not need to hold Giant for sbuf operations any more.
phk [Thu, 26 Dec 2002 20:46:30 +0000 (20:46 +0000)]
We should not need to hold Giant for sbuf operations any more.

21 years agoAdd an XXX comment to explain the predicament.
phk [Thu, 26 Dec 2002 20:45:37 +0000 (20:45 +0000)]
Add an XXX comment to explain the predicament.

21 years agos/pseudo-device/device/
ru [Thu, 26 Dec 2002 19:56:29 +0000 (19:56 +0000)]
s/pseudo-device/device/

21 years agoReturn an error if the size of the sector is zero. This is for removable
nyan [Thu, 26 Dec 2002 15:50:45 +0000 (15:50 +0000)]
Return an error if the size of the sector is zero.  This is for removable
devices that is not inserted any media.

This is MFC candidate.

Submitted by: ISAKA Yoji <isaka@cory.jp>

21 years agoUnbreak for the !__XSI_VISIBLE case when another header that defines
tjr [Thu, 26 Dec 2002 14:44:45 +0000 (14:44 +0000)]
Unbreak for the !__XSI_VISIBLE case when another header that defines
size_t has not been included first.

21 years agoAdd "restrict" qualifiers to the prototype for wordexp().
tjr [Thu, 26 Dec 2002 14:38:27 +0000 (14:38 +0000)]
Add "restrict" qualifiers to the prototype for wordexp().

21 years agoAdd an implementation of the POSIX wordexp() and wordfree() functions,
tjr [Thu, 26 Dec 2002 14:34:18 +0000 (14:34 +0000)]
Add an implementation of the POSIX wordexp() and wordfree() functions,
which perform shell-style word expansion on strings. This is still a
little rough around the edges.

PR: 13420

21 years agoAdd the POSIX <wordexp.h> header file.
tjr [Thu, 26 Dec 2002 14:30:55 +0000 (14:30 +0000)]
Add the POSIX <wordexp.h> header file.

PR: 13420

21 years agoAdd the "wordexp" shell built-in command which will be used to implement
tjr [Thu, 26 Dec 2002 14:28:54 +0000 (14:28 +0000)]
Add the "wordexp" shell built-in command which will be used to implement
the POSIX wordexp() function.

21 years agoWhen resuming after a system suspend, re-issue the UHCI_CMD_MAXP
iedowse [Thu, 26 Dec 2002 13:25:57 +0000 (13:25 +0000)]
When resuming after a system suspend, re-issue the UHCI_CMD_MAXP
command in case this setting was not saved. Since bandwidth reclamation
(-current only) often results in bus activity continuing to the end
of every frame, most transfers would fail with IOERROR if this
setting is missed.

Reviewed by: n_hibma
MFC after: 1 week

21 years agoSet %bx register instead of %di register, see vbe3.0, page 45.
davidxu [Thu, 26 Dec 2002 13:20:58 +0000 (13:20 +0000)]
Set %bx register instead of %di register, see vbe3.0, page 45.