]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoTruncate long file names in stat_display(), as was originally intended.
des [Mon, 5 Jul 2004 11:21:50 +0000 (11:21 +0000)]
Truncate long file names in stat_display(), as was originally intended.

MFC after: 1 week

20 years agoMake libalias WARNS?=6-clean. This mostly involves renaming variables
des [Mon, 5 Jul 2004 11:10:57 +0000 (11:10 +0000)]
Make libalias WARNS?=6-clean.  This mostly involves renaming variables
named link, foo_link or link_foo to lnk, foo_lnk or lnk_foo, fixing
signed / unsigned comparisons, and shoving unused function arguments
under the carpet.

I was hoping WARNS?=6 might reveal more serious problems, and perhaps
the source of the -O2 breakage, but found no smoking gun.

20 years agoParenthesize return values.
des [Mon, 5 Jul 2004 10:55:23 +0000 (10:55 +0000)]
Parenthesize return values.

20 years agoMechanical whitespace cleanup.
des [Mon, 5 Jul 2004 10:53:28 +0000 (10:53 +0000)]
Mechanical whitespace cleanup.

20 years agoMarkup fixes.
ru [Mon, 5 Jul 2004 07:21:55 +0000 (07:21 +0000)]
Markup fixes.

20 years agoMarkup fixes.
ru [Mon, 5 Jul 2004 06:59:17 +0000 (06:59 +0000)]
Markup fixes.

20 years agoMarkup fixes.
ru [Mon, 5 Jul 2004 06:53:34 +0000 (06:53 +0000)]
Markup fixes.

20 years agoMarkup nits.
ru [Mon, 5 Jul 2004 06:39:03 +0000 (06:39 +0000)]
Markup nits.

20 years agoLocking cleanup for rl(4).
bms [Mon, 5 Jul 2004 02:51:32 +0000 (02:51 +0000)]
Locking cleanup for rl(4).
 - Eliminate the use of a recursive mutex.
 - Mark the driver INTR_MPSAFE.

This work is incomplete and will be refined in a future commit.
 - Most notably, _locked() variants of entry points need to be introduced.
 - The mii upcall/downcall may still be racy.
 - Add a stubbed-out guard against racing rl_detach() for the time being.

Tested on: UP, debug.mpsafenet && !debug.mpsafenet
Reviewed by: silence on -net

20 years agostyle(9) and whitespace cleanup.
bms [Mon, 5 Jul 2004 02:46:42 +0000 (02:46 +0000)]
style(9) and whitespace cleanup.
Use C99 types. Use ANSI function definitions. Sort prototypes.
Split long lines correctly. Punctuate/wordsmith comments.
Use device_printf()/if_printf() where possible.

Reviewed by: -net (silence)

20 years agoReintroduce and clean up locking in xl(4).
bms [Mon, 5 Jul 2004 02:34:35 +0000 (02:34 +0000)]
Reintroduce and clean up locking in xl(4).

 - Eliminate the use of a recursive mutex.
 - Mark the driver as INTR_MPSAFE.
 - Split the default media choice code out into xl_choose_media() to
   avoid making poor assumptions about the state of the lock during attach.
 - The miibus upcall/downcall paths may still be racy.
   Change to commented-out locking assertions there for now.
 - Tested with nfsclient, routed, ssh, ntp, dhclient and quagga bgpd.
 - This needs SMP test coverage. I do not have such resources.

Tested on: UP, !debug.mpsafenet && debug.mpsafenet
Hardware: 3C905B-TX (0x905510b7)

20 years agoMFp4: gmirror(8) is coming soon, and we need g_metadata_read() there.
pjd [Mon, 5 Jul 2004 02:06:44 +0000 (02:06 +0000)]
MFp4: gmirror(8) is coming soon, and we need g_metadata_read() there.

20 years agoRework syncer termination code:
truckman [Mon, 5 Jul 2004 01:07:33 +0000 (01:07 +0000)]
Rework syncer termination code:

    Speed up the syncer when shutting down by sleeping for a shorter
    period of time instead of cranking up rushjob and using the
    normal one second sleep.

    Skip empty worklist slots when shutting down to avoid lengthy
    intervals of inactivity.

    Give I/O more time to complete between steps by not speeding the
    syncer quite as much.

    Terminate the syncer after one full pass through the worklist
    plus one second with the worklist containing nothing but syncer
    vnodes.

    Print an indication of shutdown progress to the console.

Add a sysctl, vfs.worklist_len, to allow the size of the syncer worklist
to be monitored.

20 years agoUse if_printf() and device_printf() where appropriate, i.e.:
bms [Mon, 5 Jul 2004 00:15:23 +0000 (00:15 +0000)]
Use if_printf() and device_printf() where appropriate, i.e.:
 - Use device_printf() during device probe/attach.
 - Move if_xname initialization to before xl_reset() is called.
 - Use if_printf() at all other times after struct ifnet has been
   initialized.

20 years agoANSIfy function definitions.
bms [Sun, 4 Jul 2004 23:31:28 +0000 (23:31 +0000)]
ANSIfy function definitions.
Remove unnecessary return keywords.
Other minor stylistic changes.

20 years agoMark our interrupt shareable and don't insist on DMA. This gets us closer
phk [Sun, 4 Jul 2004 23:16:04 +0000 (23:16 +0000)]
Mark our interrupt shareable and don't insist on DMA.  This gets us closer
to working with a secondary floppy controller on a PC.

20 years agoGive synthetic root filesystem device vnodes a v_bsize of DEV_BSIZE.
phk [Sun, 4 Jul 2004 22:33:22 +0000 (22:33 +0000)]
Give synthetic root filesystem device vnodes a v_bsize of DEV_BSIZE.

20 years agoFix whitespace, indentation, long line wrapping and comments.
bms [Sun, 4 Jul 2004 22:20:52 +0000 (22:20 +0000)]
Fix whitespace, indentation, long line wrapping and comments.

20 years agoFix whitespace and comments.
bms [Sun, 4 Jul 2004 22:19:50 +0000 (22:19 +0000)]
Fix whitespace and comments.

20 years agoFixed cross-references in SEE ALSO.
ru [Sun, 4 Jul 2004 21:15:37 +0000 (21:15 +0000)]
Fixed cross-references in SEE ALSO.
Emininated double space and hard sentence breaks.

20 years agoSort SEE ALSO references (in dictionary order, ignoring case).
ru [Sun, 4 Jul 2004 20:55:50 +0000 (20:55 +0000)]
Sort SEE ALSO references (in dictionary order, ignoring case).

20 years agoPass the operation in with the fsidctl.
alfred [Sun, 4 Jul 2004 20:21:58 +0000 (20:21 +0000)]
Pass the operation in with the fsidctl.
Remove some fsidctls that we will not be using.
Correct prototypes for fs sysctls.

20 years agoAdd commentary explaining why we return EBADF upon attempts to fflush() a
cperciva [Sun, 4 Jul 2004 20:17:00 +0000 (20:17 +0000)]
Add commentary explaining why we return EBADF upon attempts to fflush() a
read-only file.

Discussed on: -current

20 years agoFix dofsync() declaration.
cperciva [Sun, 4 Jul 2004 19:52:48 +0000 (19:52 +0000)]
Fix dofsync() declaration.

Pointed out by: dwmalone

20 years agoMake the last commit handle non-phk root devices better.
phk [Sun, 4 Jul 2004 19:42:25 +0000 (19:42 +0000)]
Make the last commit handle non-phk root devices better.

20 years agoWhere syslogd would have fsync()ed a file in the past, instead set a flag
cperciva [Sun, 4 Jul 2004 19:13:58 +0000 (19:13 +0000)]
Where syslogd would have fsync()ed a file in the past, instead set a flag
FFLAG_NEEDSYNC and fsync the file when select() next returns zero.  This
dramatically speeds up the process of logging large amounts of data, while
leaving the essential semantics (that data can be expected to be on disk
if we crash) unchanged.

In my tests, this speeds up the rc phase of booting by 18-20%. [1]

YES PLEASE! by: phk [1]

20 years agoRecord the offset of thr_id in the thread structure. Required for
marcel [Sun, 4 Jul 2004 19:07:07 +0000 (19:07 +0000)]
Record the offset of thr_id in the thread structure. Required for
debugging.

20 years agoCheck the return value of bus_dmamem_alloc() correctly.
bms [Sun, 4 Jul 2004 18:40:36 +0000 (18:40 +0000)]
Check the return value of bus_dmamem_alloc() correctly.

Submitted by: Darron Broad (with cleanups)

20 years agoWorkaround a locking problem in vlan(4). vlan_setmulti() may be called
bms [Sun, 4 Jul 2004 18:32:54 +0000 (18:32 +0000)]
Workaround a locking problem in vlan(4). vlan_setmulti() may be called
with sleepable locks held from further up in the network stack, and
attempts to allocate memory to hold multicast group membership information
with M_WAITOK.

This panic was triggered specifically when an exiting routing daemon
process closes its raw sockets after joining multicast groups on them.

While we're here, comment some possible locking badness.

PR: kern/48560

20 years agoYet another pointy hat: When restoring file flags, it's okay to use the
kientzle [Sun, 4 Jul 2004 18:28:56 +0000 (18:28 +0000)]
Yet another pointy hat: When restoring file flags, it's okay to use the
shared stat buffer, but don't try to access it through an uninitialized
pointer.

20 years agoPass the pointy hat, please:
kientzle [Sun, 4 Jul 2004 18:13:01 +0000 (18:13 +0000)]
Pass the pointy hat, please:
All of --help should go to stdout, not some to stdout and some to stderr.

20 years agoAdd documentation for the ieee80211_node.c functions.
bms [Sun, 4 Jul 2004 18:07:58 +0000 (18:07 +0000)]
Add documentation for the ieee80211_node.c functions.

Submitted by: Darron Broad

20 years agoNothing says that /var/log can't be not a directory but a symbolic link
sobomax [Sun, 4 Jul 2004 17:24:12 +0000 (17:24 +0000)]
Nothing says that /var/log can't be not a directory but a symbolic link
to a directory. Therefore, use stat(2) instead of lstat(2) to check if
/var/log exists.

MFC after: 7 days

20 years agostyle(9)/whitespace cleanup while I'm in this file.
bms [Sun, 4 Jul 2004 16:43:24 +0000 (16:43 +0000)]
style(9)/whitespace cleanup while I'm in this file.

20 years agoAdd modemcontrol support for DTR and DCD.
phk [Sun, 4 Jul 2004 16:27:58 +0000 (16:27 +0000)]
Add modemcontrol support for DTR and DCD.

Tested by: ambrisko

20 years agoFix bug causing `[' to be wrongly included in character class expressions
tjr [Sun, 4 Jul 2004 16:25:41 +0000 (16:25 +0000)]
Fix bug causing `[' to be wrongly included in character class expressions
in some multibyte locales (Red Hat bug #108484).

Obtained from: Fedora (Tim Waugh)

20 years agoMake the default memory range in the top 2GB of ram in the hopes that
imp [Sun, 4 Jul 2004 16:23:25 +0000 (16:23 +0000)]
Make the default memory range in the top 2GB of ram in the hopes that
this more accurately reflects what the underlying hardware of most
acpi machines that don't have children pci busses.

We still need a better way to get this information from acpi/hardware.

20 years agoMake grep run much (~10x) faster in multibyte locales by caching the wide
tjr [Sun, 4 Jul 2004 16:16:59 +0000 (16:16 +0000)]
Make grep run much (~10x) faster in multibyte locales by caching the wide
character representation of input data across calls to dfaexec(), and by
caching the lengths of character across calls to check_multibyte_string().

Obtained from: Fedora (Tim Waugh)

20 years agoConsistently use __inline instead of __inline__ as the former is an empty macro
stefanf [Sun, 4 Jul 2004 16:11:03 +0000 (16:11 +0000)]
Consistently use __inline instead of __inline__ as the former is an empty macro
in <sys/cdefs.h> for compilers without support for inline.

20 years agometa_p is a void *, so a variable that's of type void * can't be
imp [Sun, 4 Jul 2004 16:10:36 +0000 (16:10 +0000)]
meta_p is a void *, so a variable that's of type void * can't be
dereferenced directly.  Toss an ifdef around it for the moment and
allow this to compile.  This likely means that priority packets aren't
queued to the special high priority queue.  The maintainer of this
should look into the problem.

This is likely fallout from the netgraph migration to using a more
generic meta tag from the mbug recently.

Fixes: pc98 tinerbox
20 years agoIntroduce debug.nosleepwithlocks sysctl, 0 by default. If set to 1
bmilekic [Sun, 4 Jul 2004 16:07:44 +0000 (16:07 +0000)]
Introduce debug.nosleepwithlocks sysctl, 0 by default.  If set to 1
and WITNESS is not built, then force all M_WAITOK allocations to
M_NOWAIT behavior (transparently).  This is to be used temporarily
if wierd deadlocks are reported because we still have code paths
that perform M_WAITOK allocations with lock(s) held, which can
lead to deadlock.  If WITNESS is compiled, then the sysctl is ignored
and we ask witness to tell us wether we have locks held, converting
to M_NOWAIT behavior only if it tells us that we do.

Note this removes the previous mbuf.h inclusion as well (only needed
by last revision), and cleans up unneeded [artificial] comparisons
to just the mbuf zones.  The problem described above has nothing to
do with previous mbuf wait behavior; it is a general problem.

20 years agoReextend the M_WAITOK-disabling-hack to all three of the mbuf-related
green [Sun, 4 Jul 2004 15:59:25 +0000 (15:59 +0000)]
Reextend the M_WAITOK-disabling-hack to all three of the mbuf-related
zones, and do it by direct comparison of uma_zone_t instead of strcmp.

The mbuf subsystem used to provide M_TRYWAIT/M_DONTWAIT semantics, but
this is mostly no longer the case.  M_WAITOK has taken over the spot
M_TRYWAIT used to have, and for mbuf things, still may return NULL if
the code path is incorrectly holding a mutex going into mbuf allocation
functions.

The M_WAITOK/M_NOWAIT semantics are absolute; though it may deadlock
the system to try to malloc or uma_zalloc something with a mutex held
and M_WAITOK specified, it is absolutely required to not return NULL
and will result in instability and/or security breaches otherwise.
There is still room to add the WITNESS_WARN() to all cases so that
we are notified of the possibility of deadlocks, but it cannot change
the value of the "badness" variable and allow allocation to actually
fail except for the specialized cases which used to be M_TRYWAIT.

20 years agoThe net.link.ether.bridge.enable sysctl MIB variable enables bridge
bms [Sun, 4 Jul 2004 15:53:28 +0000 (15:53 +0000)]
The net.link.ether.bridge.enable sysctl MIB variable enables bridge
functionality by setting to a non-zero value. This is an integer, but
is treated as a boolean by the code, so clamp it to a boolean value
when set so as to avoid unnecessary bridge reinitialization if it's
changed to another value.

PR: kern/61174
Requested by: Bruce Cran

20 years agoAdd a HARDWARE section which lists supported devices. The actual
simon [Sun, 4 Jul 2004 14:17:41 +0000 (14:17 +0000)]
Add a HARDWARE section which lists supported devices.  The actual
device listings has been moved (and in some cases more or less
rewritten) from the DESCRIPTION section.

This will be used later for automatically generating device listings
in the Hardware Notes, by parsing the manual pages.

Reviewed in principle by: ru, hrs, trhodes
No objections: -doc, re
Section name inspired by: NetBSD

20 years agoAvoid accessing accessing memory past the end of mb_properties in the
tjr [Sun, 4 Jul 2004 14:00:25 +0000 (14:00 +0000)]
Avoid accessing accessing memory past the end of mb_properties in the
degenerate case of fgrep with an empty pattern in a multibyte locale.
Found by phkmalloc.

20 years agoWe only need to check for overlaps if we increasing access counts.
phk [Sun, 4 Jul 2004 13:44:48 +0000 (13:44 +0000)]
We only need to check for overlaps if we increasing access counts.

20 years agoGive natd multi-instance capabilities.
phk [Sun, 4 Jul 2004 12:53:54 +0000 (12:53 +0000)]
Give natd multi-instance capabilities.

This makes it possible to do load-sharing on two xDSL lines etc.

20 years agoAdd LibAliasOutTry() which checks a packet for a hit in the tables, but
phk [Sun, 4 Jul 2004 12:53:07 +0000 (12:53 +0000)]
Add LibAliasOutTry() which checks a packet for a hit in the tables, but
does not create a new entry if none is found.

20 years agoBlocksize for I/O should be a property of the vnode and not found by groping
phk [Sun, 4 Jul 2004 12:49:04 +0000 (12:49 +0000)]
Blocksize for I/O should be a property of the vnode and not found by groping
around in the vnodes surroundings when we allocate a block.

Assign a blocksize when we create a vnode, and yell a warning (and ignore it)
if we got the wrong size.

Please email all such warnings to me.

20 years agoImprove case-insensitive matching in multibyte locales.
tjr [Sun, 4 Jul 2004 11:58:10 +0000 (11:58 +0000)]
Improve case-insensitive matching in multibyte locales.

Obtained from: Isamu Hasegawa (IBM) via Fedora

20 years agoFix failure of fgrep to report some matches (Red Hat bug #116909).
tjr [Sun, 4 Jul 2004 11:33:49 +0000 (11:33 +0000)]
Fix failure of fgrep to report some matches (Red Hat bug #116909).

Obtained from: Fedora (Tim Waugh)

20 years agoIntroduce a new kevent filter. EVFILT_FS that will be used to signal
alfred [Sun, 4 Jul 2004 10:52:54 +0000 (10:52 +0000)]
Introduce a new kevent filter.  EVFILT_FS that will be used to signal
generic filesystem events to userspace.  Currently only mount and unmount
of filesystems are signalled.  Soon to be added, up/down status of NFS.

Introduce a sysctl node used to route requests to/from filesystems
based on filesystem ids.

Introduce a new vfsop, vfs_sysctl(mp, req) that is used as the callback/
entrypoint by the sysctl code to change individual filesystems.

20 years agoUse hard_locale() to check whether collating order is "hard" regardless
tjr [Sun, 4 Jul 2004 10:24:48 +0000 (10:24 +0000)]
Use hard_locale() to check whether collating order is "hard" regardless
of whether NLS is enabled.

20 years agoIgnore ENABLE_NLS when deciding whether we can use setlocale().
tjr [Sun, 4 Jul 2004 10:22:50 +0000 (10:22 +0000)]
Ignore ENABLE_NLS when deciding whether we can use setlocale().
Use the same shortcut as glibc to test for the "C" or "POSIX" locale.

20 years agoRevision 1.496 would not boot on my system due to
alfred [Sun, 4 Jul 2004 10:19:15 +0000 (10:19 +0000)]
Revision 1.496 would not boot on my system due to
ffs_mount -> bdevvp -> getnewvnode(..., mp = NULL, ...) ->
 insmntqueue(vp, mp = NULL) -> KASSERT -> panic

Make getnewvnode() only call insmntqueue() if the mountpoint parameter
is not NULL.

20 years agoUpdate for grep 2.5.1.
tjr [Sun, 4 Jul 2004 10:05:37 +0000 (10:05 +0000)]
Update for grep 2.5.1.

20 years agoMerge local changes.
tjr [Sun, 4 Jul 2004 10:02:03 +0000 (10:02 +0000)]
Merge local changes.

20 years agoThis commit was generated by cvs2svn to compensate for changes in r131554,
tjr [Sun, 4 Jul 2004 09:52:08 +0000 (09:52 +0000)]
This commit was generated by cvs2svn to compensate for changes in r131554,
which included commits to RCS files with non-trunk default branches.

20 years agoImport GNU grep 2.5.1 (trimmed)
tjr [Sun, 4 Jul 2004 09:52:08 +0000 (09:52 +0000)]
Import GNU grep 2.5.1 (trimmed)

20 years agoAdd device ID for atuwi USB wlan driver, Linksys WUSB11 v2.8 wireless adapter.
sanpei [Sun, 4 Jul 2004 09:13:53 +0000 (09:13 +0000)]
Add device ID for atuwi USB wlan driver, Linksys WUSB11 v2.8 wireless adapter.
(Atmel at76c503a http://vitsch.net/bsd/atuwi)

PR: kern/68465
Submitted by: Suihong Liang <s2liang@uwaterloo.ca>

20 years agoWhen we traverse the vnodes on a mountpoint we need to look out for
phk [Sun, 4 Jul 2004 08:52:35 +0000 (08:52 +0000)]
When we traverse the vnodes on a mountpoint we need to look out for
our cached 'next vnode' being removed from this mountpoint.  If we
find that it was recycled, we restart our traversal from the start
of the list.

Code to do that is in all local disk filesystems (and a few other
places) and looks roughly like this:

MNT_ILOCK(mp);
loop:
for (vp = TAILQ_FIRST(&mp...);
    (vp = nvp) != NULL;
    nvp = TAILQ_NEXT(vp,...)) {
if (vp->v_mount != mp)
goto loop;
MNT_IUNLOCK(mp);
...
MNT_ILOCK(mp);
}
MNT_IUNLOCK(mp);

The code which takes vnodes off a mountpoint looks like this:

MNT_ILOCK(vp->v_mount);
...
TAILQ_REMOVE(&vp->v_mount->mnt_nvnodelist, vp, v_nmntvnodes);
...
MNT_IUNLOCK(vp->v_mount);
...
vp->v_mount = something;

(Take a moment and try to spot the locking error before you read on.)

On a SMP system, one CPU could have removed nvp from our mountlist
but not yet gotten to assign a new value to vp->v_mount while another
CPU simultaneously get to the top of the traversal loop where it
finds that (vp->v_mount != mp) is not true despite the fact that
the vnode has indeed been removed from our mountpoint.

Fix:

Introduce the macro MNT_VNODE_FOREACH() to traverse the list of
vnodes on a mountpoint while taking into account that vnodes may
be removed from the list as we go.  This saves approx 65 lines of
duplicated code.

Split the insmntque() which potentially moves a vnode from one mount
point to another into delmntque() and insmntque() which does just
what the names say.

Fix delmntque() to set vp->v_mount to NULL while holding the
mountpoint lock.

20 years agoWrap rc.subr inside
cperciva [Sun, 4 Jul 2004 07:21:18 +0000 (07:21 +0000)]
Wrap rc.subr inside

if [ -z "${_rc_subr_loaded}" ]; then
_rc_subr_loaded="YES"
...
fi

in order to avoid re-interpreting rc.subr every time an rc.d
script is run.  In my tests, this speeds up rc time by about
8-10%.

20 years agoDocument that boundaries cannot be smaller than the max segment size.
scottl [Sun, 4 Jul 2004 04:50:00 +0000 (04:50 +0000)]
Document that boundaries cannot be smaller than the max segment size.

20 years agoYes, NgRecvAsciiMsg has the same results as NgRecvAsciiMsg, but it's
jmallett [Sun, 4 Jul 2004 04:03:37 +0000 (04:03 +0000)]
Yes, NgRecvAsciiMsg has the same results as NgRecvAsciiMsg, but it's
much more apt to note that it has the same result as NgRecvMsg.  Make
the manual page less circular in its reference to this fact.

20 years agoFix regression in new version of GNU regex code: bracket expressions
tjr [Sun, 4 Jul 2004 02:46:55 +0000 (02:46 +0000)]
Fix regression in new version of GNU regex code: bracket expressions
like [X-Y] should match all characters between X-Y according to the
locale's collating order, not by binary value. For now, this only fixes
the !MBS_SUPPORT case (which is the default).

20 years agoMerge local changes (addition of $FreeBSD$).
tjr [Sun, 4 Jul 2004 01:07:38 +0000 (01:07 +0000)]
Merge local changes (addition of $FreeBSD$).

20 years agoThis commit was generated by cvs2svn to compensate for changes in r131543,
tjr [Sun, 4 Jul 2004 01:03:02 +0000 (01:03 +0000)]
This commit was generated by cvs2svn to compensate for changes in r131543,
which included commits to RCS files with non-trunk default branches.

20 years agoImport of GNU regex from GNU grep 2.5.1 (trimmed).
tjr [Sun, 4 Jul 2004 01:03:02 +0000 (01:03 +0000)]
Import of GNU regex from GNU grep 2.5.1 (trimmed).

20 years agoFixed markup.
ru [Sat, 3 Jul 2004 23:14:34 +0000 (23:14 +0000)]
Fixed markup.

20 years agoOne more hard sentence break.
ru [Sat, 3 Jul 2004 23:01:44 +0000 (23:01 +0000)]
One more hard sentence break.

20 years agoMerge junk.
ru [Sat, 3 Jul 2004 22:53:50 +0000 (22:53 +0000)]
Merge junk.

20 years agoEliminate double whitespace.
ru [Sat, 3 Jul 2004 22:30:10 +0000 (22:30 +0000)]
Eliminate double whitespace.

20 years agoDocument /var/run/dmesg.boot, which is created by the rc scripts. Many
imp [Sat, 3 Jul 2004 21:01:18 +0000 (21:01 +0000)]
Document /var/run/dmesg.boot, which is created by the rc scripts.  Many
people have suggested that we document this somewhere, and this was a common
suggestion.

20 years agoThese don't need RMAN_RESOURCE_VISIBLE now that rman is visible
imp [Sat, 3 Jul 2004 20:56:16 +0000 (20:56 +0000)]
These don't need RMAN_RESOURCE_VISIBLE now that rman is visible

20 years agoReally remove __RMAN_RESORUCE_VISIBLE
imp [Sat, 3 Jul 2004 20:49:00 +0000 (20:49 +0000)]
Really remove __RMAN_RESORUCE_VISIBLE

20 years agoUse the rman_* functions in preference to reaching into struct resource.
imp [Sat, 3 Jul 2004 20:48:01 +0000 (20:48 +0000)]
Use the rman_* functions in preference to reaching into struct resource.
Remove __RMAN_RESOURCE_VISIBLE after compilation confirms it is now not
needed.

20 years agoDon't define __RMAN_RESOURCE_VISISBLE. They aren't needed here after
imp [Sat, 3 Jul 2004 20:11:49 +0000 (20:11 +0000)]
Don't define __RMAN_RESOURCE_VISISBLE.  They aren't needed here after
I've converted the direct accessing of struct resource members to the
preferred interface.

20 years agoRemove stale comment
phk [Sat, 3 Jul 2004 19:37:06 +0000 (19:37 +0000)]
Remove stale comment

20 years agoChange M_WAITOK argument to sodupsockaddr() to M_NOWAIT. When the call
rwatson [Sat, 3 Jul 2004 19:17:06 +0000 (19:17 +0000)]
Change M_WAITOK argument to sodupsockaddr() to M_NOWAIT.  When the call
to dup_sockaddr() was renamed to sodupsockaddr(), the argument was
changed from '1' to 'M_WAITOK', which changed the semantics.  This
resulted in a WITNESS warning about a potential sleep while holding the
NFS server mutex.  Now this will no longer happen, restoring a possible
bug present in the original code (setting RC_NAM even though the malloc
to copy the addres may fail).  bde observes that the flag names here
should probably not be the same as the malloc flags for name space
reasons.

Bumped into by: kuriyama

20 years agoEliminated double whitespace.
ru [Sat, 3 Jul 2004 18:35:53 +0000 (18:35 +0000)]
Eliminated double whitespace.

20 years agoMechanically kill hard sentence breaks and double whitespaces.
ru [Sat, 3 Jul 2004 18:29:24 +0000 (18:29 +0000)]
Mechanically kill hard sentence breaks and double whitespaces.

20 years agoCommit the first of half of changes that allow busdma to transparently
scottl [Sat, 3 Jul 2004 18:18:36 +0000 (18:18 +0000)]
Commit the first of half of changes that allow busdma to transparently
honor the alignment and boundary constraints in the dma tag when loading
buffers.  Previously, these constraints were only honored when allocating
memory via bus_dmamem_alloc().  Now, bus_dmamap_load() will automatically
use bounce buffers when needed.

Also add a set of sysctls to monitor the global busdma stats.  These are:

hw.busdma.free_bpages
hw.busdma.reserved_bpages
hw.busdma.active_bpages
hw.busdma.total_bpages
hw.busdma.total_bounced
hw.busdma.total_deferred

20 years agoLimit mbuma damage. Suddenly ALL allocations with M_WAITOK are subject
green [Sat, 3 Jul 2004 18:11:41 +0000 (18:11 +0000)]
Limit mbuma damage.  Suddenly ALL allocations with M_WAITOK are subject
to failing -- that is, allocations via malloc(M_WAITOK) that are required
to never fail -- if WITNESS is not defined.  While everyone should be
running WITNESS, in any case, zone "Mbuf" allocations are really the only
ones that should be screwed with by this hack.

This hack is crashing people, and would continue to do so with or without
WITNESS.  Things shouldn't be allocating with M_WAITOK with locks held,
but it's not okay just to always remove M_WAITOK when !WITNESS.

Reported by: Bernd Walter <ticso@cicely5.cicely.de>

20 years agoAdd NULL arg to mi_switch() call to stop kernel compiles from breaking.
phk [Sat, 3 Jul 2004 16:57:51 +0000 (16:57 +0000)]
Add NULL arg to mi_switch() call to stop kernel compiles from breaking.

20 years agoRemove "register" keyword and trailing white space.
phk [Sat, 3 Jul 2004 16:56:45 +0000 (16:56 +0000)]
Remove "register" keyword and trailing white space.

20 years agoUse C99 conforming designated initialisers rather than the obsolete GCC syntax.
stefanf [Sat, 3 Jul 2004 16:52:57 +0000 (16:52 +0000)]
Use C99 conforming designated initialisers rather than the obsolete GCC syntax.

20 years agoBy popular request, add a workaround that allows large (>128GB or so)
tjr [Sat, 3 Jul 2004 13:22:38 +0000 (13:22 +0000)]
By popular request, add a workaround that allows large (>128GB or so)
FAT32 filesystems to be mounted, subject to some fairly serious limitations.

This works by extending the internal pseudo-inode-numbers generated from
the file's starting cluster number to 64-bits, then creating a table
mapping these into arbitrary 32-bit inode numbers, which can fit in
struct dirent's d_fileno and struct vattr's va_fileid fields. The mappings
do not persist across unmounts or reboots, so it's not possible to export
these filesystems through NFS. The mapping table may grow to be rather
large, and may grow large enough to exhaust kernel memory on filesystems
with millions of files.

Don't enable this option unless you understand the consequences.

20 years agoThe -O2 bugs are in libalias(3), not ppp(8).
des [Sat, 3 Jul 2004 09:41:58 +0000 (09:41 +0000)]
The -O2 bugs are in libalias(3), not ppp(8).

20 years agoAdd FreeBSD-SA-04:13.linux and the device.hints related
josef [Sat, 3 Jul 2004 08:45:20 +0000 (08:45 +0000)]
Add  FreeBSD-SA-04:13.linux and the device.hints related
expansion of pcm(4).

20 years agoGrammar nits.
pjd [Sat, 3 Jul 2004 08:36:09 +0000 (08:36 +0000)]
Grammar nits.

Submitted by: David Magda <dmagda@ee.ryerson.ca>

20 years agoFix a markup nit and a misplaced full stop in previous.
tjr [Sat, 3 Jul 2004 07:07:11 +0000 (07:07 +0000)]
Fix a markup nit and a misplaced full stop in previous.

Noticed by: ru

20 years agoSMPng locking cleanup for vr(4).
bms [Sat, 3 Jul 2004 02:59:02 +0000 (02:59 +0000)]
SMPng locking cleanup for vr(4).
 - Remove recursive locking situations. Remove the MTX_RECURSE bit.
 - Take the lock for any routine which is not called from within if_vr.c
   itself; this includes entry points called by newbus, ifnet, callout,
   ifmedia, and polling subsystems.
 - Remove spl references from the code added to miibus callbacks in rev 1.60.
 - Add the INTR_MPSAFE bit.
 - Tidy up some assignments; locks are not needed for taking the address
   of something at a known offset, for example.
 - Tested on the machine this was committed from.

Tested on: UP only, !debug.mpsafenet && debug.mpsafenet
Reviewed by: rwatson

20 years ago - Another whitespace pass; make locking calls more obvious.
bms [Sat, 3 Jul 2004 02:52:32 +0000 (02:52 +0000)]
 - Another whitespace pass; make locking calls more obvious.
 - Use C99 types for vr_miibus_readreg().

20 years agoAdd a NULL param to an mi_switch() that I missed.
jhb [Sat, 3 Jul 2004 02:38:03 +0000 (02:38 +0000)]
Add a NULL param to an mi_switch() that I missed.

Reported by: Jung-uk Kim jkim at niksun dot com

20 years agoUpdate for coreutils.
tjr [Sat, 3 Jul 2004 02:37:42 +0000 (02:37 +0000)]
Update for coreutils.

20 years agoUnbreak -O2 build: initialize nstatus to avoid uninitialized warning.
marcel [Sat, 3 Jul 2004 02:10:52 +0000 (02:10 +0000)]
Unbreak -O2 build: initialize nstatus to avoid uninitialized warning.

20 years agoDocument missing multibyte character handling in utilities specified
tjr [Sat, 3 Jul 2004 02:03:44 +0000 (02:03 +0000)]
Document missing multibyte character handling in utilities specified
by POSIX.

20 years agoClarify: test compares strings by binary value, not "ASCII value".
tjr [Sat, 3 Jul 2004 01:49:17 +0000 (01:49 +0000)]
Clarify: test compares strings by binary value, not "ASCII value".

20 years agoDocument missing multibyte character support in utilities specified
tjr [Sat, 3 Jul 2004 01:28:21 +0000 (01:28 +0000)]
Document missing multibyte character support in utilities specified
by POSIX.

20 years agoFix SCHED_ULE build on SMP. The previous revision (1.110)
bmilekic [Sat, 3 Jul 2004 01:19:46 +0000 (01:19 +0000)]
Fix SCHED_ULE build on SMP.  The previous revision (1.110)
introduced a KSE_CAN_MIGRATE() invocation with one argument
missing (class).  Either this is a genuine forget or it crept
in from JHB's repo where he may have modified it.  If it's
the latter then it may require more attention.  For now fix
the make depend.