]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoIn original kern_execve() code, at the start of the function, it forces
davidxu [Wed, 6 Oct 2004 00:40:41 +0000 (00:40 +0000)]
In original kern_execve() code, at the start of the function, it forces
all other threads to suicide, problem is execve() could be failed, and
a failed execve() would change threaded process to unthreaded, this side
effect is unexpected.
The new code introduces a new single threading mode SINGLE_BOUNDARY, in
the mode, all threads should suspend themself at user boundary except
the singler. we can not use SINGLE_NO_EXIT because we want to start from
a clean state if execve() is successful, suspending other threads at unknown
point and later resuming them from there and forcing them to exit at user
boundary may cause the process to start from a dirty state. If execve() is
successful, current thread upgrades to SINGLE_EXIT mode and forces other
threads to suicide at user boundary, otherwise, other threads will be resumed
and their interrupted syscall will be restarted.

Reviewed by: julian

19 years agoForced commit to note the submitter.
green [Tue, 5 Oct 2004 23:50:10 +0000 (23:50 +0000)]
Forced commit to note the submitter.

Submitted by: stefanf

19 years agoMake sure to return 0 from kernel_getnfile() since if there were an
green [Tue, 5 Oct 2004 23:49:27 +0000 (23:49 +0000)]
Make sure to return 0 from kernel_getnfile() since if there were an
error, it would exit() (and it needs to return a value).

19 years agoDon't prepend the directory specified by -d when the file is a relative
brooks [Tue, 5 Oct 2004 22:16:31 +0000 (22:16 +0000)]
Don't prepend the directory specified by -d when the file is a relative
path.  Doing so makes no sense.  I'm not sure allowing relative paths
makes sense either, but I'm not going to break that now.

19 years agoFix whitespace botch that only showed up in the commit message diff :-/
julian [Tue, 5 Oct 2004 22:14:02 +0000 (22:14 +0000)]
Fix whitespace botch that only showed up in the commit message diff :-/

MFC after: 4 days

19 years agoBump __FreeBSD_version for addition of newsyslog -d.
brooks [Tue, 5 Oct 2004 22:09:12 +0000 (22:09 +0000)]
Bump __FreeBSD_version for addition of newsyslog -d.

19 years agoSlight cleanup in the single threading code.
julian [Tue, 5 Oct 2004 22:05:25 +0000 (22:05 +0000)]
Slight cleanup in the single threading code.

MFC after: 4 days

19 years agoWhen preempting a thread, put it back on the HEAD of its run queue.
julian [Tue, 5 Oct 2004 22:03:10 +0000 (22:03 +0000)]
When preempting a thread, put it back on the HEAD of its run queue.
(Only really implemented in 4bsd)

MFC after: 4 days

19 years agoOops. left out part of the diff.
julian [Tue, 5 Oct 2004 21:26:27 +0000 (21:26 +0000)]
Oops. left out part of the diff.

MFC after: 4 days

19 years agoUpdate a comment that was outdated.
njl [Tue, 5 Oct 2004 21:24:20 +0000 (21:24 +0000)]
Update a comment that was outdated.

19 years agoUse some macros to trach available scheduler slots to allow
julian [Tue, 5 Oct 2004 21:10:44 +0000 (21:10 +0000)]
Use some macros to trach available scheduler slots to allow
easier debugging.

MFC after: 4 days

19 years agolight rearrangement of some code to get some locking
julian [Tue, 5 Oct 2004 20:48:16 +0000 (20:48 +0000)]
light rearrangement of some code to get some locking
more correct

MFC after: 4 days

19 years agoBack out the -s flag and go back to dumping the SSDTs by default.
njl [Tue, 5 Oct 2004 20:45:05 +0000 (20:45 +0000)]
Back out the -s flag and go back to dumping the SSDTs by default.

19 years agoUpdate the man page with new info about overriding your DSDT.
njl [Tue, 5 Oct 2004 20:42:16 +0000 (20:42 +0000)]
Update the man page with new info about overriding your DSDT.

MFC after: 3 days

19 years agoWhen the user overrides the DSDT, replace any SSDTs with a simple no-op
njl [Tue, 5 Oct 2004 20:41:44 +0000 (20:41 +0000)]
When the user overrides the DSDT, replace any SSDTs with a simple no-op
table.  acpidump(8) concatenates the body of the DSDT and SSDTs so an
edited ASL will contain all the necessary information.  We can't use a
completely empty table since ACPI-CA reports this as a problem.

MFC after: 3 days

19 years agoBreak out to a separate function, the code to revert a multithreaded
julian [Tue, 5 Oct 2004 20:39:26 +0000 (20:39 +0000)]
Break out to a separate function, the code to revert a multithreaded
process back to officially being a non-threaded program.

MFC after: 4 days

19 years agoRegen:
sheldonh [Tue, 5 Oct 2004 20:10:35 +0000 (20:10 +0000)]
Regen:

* Hart: rev 494 of pcidevs.txt (2004-09-02)
* Boemler: vendors.txt (2004-09-30)

Approved by: re (hrs)

19 years agoadd 80211watch program
sam [Tue, 5 Oct 2004 19:53:32 +0000 (19:53 +0000)]
add 80211watch program

19 years agosimple program to watch 802.11 events through a routing socket
sam [Tue, 5 Oct 2004 19:51:34 +0000 (19:51 +0000)]
simple program to watch 802.11 events through a routing socket

19 years agoAdd 802.11-specific events that are dispatched through the routing socket.
sam [Tue, 5 Oct 2004 19:48:33 +0000 (19:48 +0000)]
Add 802.11-specific events that are dispatched through the routing socket.
This really doesn't belong here but is preferred (for the moment) over
adding yet another mechanism for sending msgs from the kernel to user apps.

Reviewed by: imp

19 years agoadd ETHERTYPE_PAE for EAPOL/802.1x
sam [Tue, 5 Oct 2004 19:28:52 +0000 (19:28 +0000)]
add ETHERTYPE_PAE for EAPOL/802.1x

19 years agoAdd note about regulatory domains to avoid confusion.
imp [Tue, 5 Oct 2004 19:27:40 +0000 (19:27 +0000)]
Add note about regulatory domains to avoid confusion.

Reviewed by: sam@

19 years agoRework how we store process times in the kernel such that we always store
jhb [Tue, 5 Oct 2004 18:51:11 +0000 (18:51 +0000)]
Rework how we store process times in the kernel such that we always store
the raw values including for child process statistics and only compute the
system and user timevals on demand.

- Fix the various kern_wait() syscall wrappers to only pass in a rusage
  pointer if they are going to use the result.
- Add a kern_getrusage() function for the ABI syscalls to use so that they
  don't have to play stackgap games to call getrusage().
- Fix the svr4_sys_times() syscall to just call calcru() to calculate the
  times it needs rather than calling getrusage() twice with associated
  stackgap, etc.
- Add a new rusage_ext structure to store raw time stats such as tick counts
  for user, system, and interrupt time as well as a bintime of the total
  runtime.  A new p_rux field in struct proc replaces the same inline fields
  from struct proc (i.e. p_[isu]ticks, p_[isu]u, and p_runtime).  A new p_crux
  field in struct proc contains the "raw" child time usage statistics.
  ruadd() has been changed to handle adding the associated rusage_ext
  structures as well as the values in rusage.  Effectively, the values in
  rusage_ext replace the ru_utime and ru_stime values in struct rusage.  These
  two fields in struct rusage are no longer used in the kernel.
- calcru() has been split into a static worker function calcru1() that
  calculates appropriate timevals for user and system time as well as updating
  the rux_[isu]u fields of a passed in rusage_ext structure.  calcru() uses a
  copy of the process' p_rux structure to compute the timevals after updating
  the runtime appropriately if any of the threads in that process are
  currently executing.  It also now only locks sched_lock internally while
  doing the rux_runtime fixup.  calcru() now only requires the caller to
  hold the proc lock and calcru1() only requires the proc lock internally.
  calcru() also no longer allows callers to ask for an interrupt timeval
  since none of them actually did.
- calcru() now correctly handles threads executing on other CPUs.
- A new calccru() function computes the child system and user timevals by
  calling calcru1() on p_crux.  Note that this means that any code that wants
  child times must now call this function rather than reading from p_cru
  directly.  This function also requires the proc lock.
- This finishes the locking for rusage and friends so some of the Giant locks
  in exit1() and kern_wait() are now gone.
- The locking in ttyinfo() has been tweaked so that a shared lock of the
  proctree lock is used to protect the process group rather than the process
  group lock.  By holding this lock until the end of the function we now
  ensure that the process/thread that we pick to dump info about will no
  longer vanish while we are trying to output its info to the console.

Submitted by: bde (mostly)
MFC after: 1 month

19 years ago- Estimate the amount of data in flight in sack recovery and use it
ps [Tue, 5 Oct 2004 18:36:24 +0000 (18:36 +0000)]
- Estimate the amount of data in flight in sack recovery and use it
  to control the packets injected while in sack recovery (for both
  retransmissions and new data).
- Cleanups to the sack codepaths in tcp_output.c and tcp_sack.c.
- Add a new sysctl (net.inet.tcp.sack.initburst) that controls the
  number of sack retransmissions done upon initiation of sack recovery.

Submitted by: Mohan Srinivasan <mohans@yahoo-inc.com>

19 years agoAdd a critical section in turnstile_unpend() from before dropping the
jhb [Tue, 5 Oct 2004 18:00:30 +0000 (18:00 +0000)]
Add a critical section in turnstile_unpend() from before dropping the
turnstile chain lock until after making all the awakened threads
runnable.  First, this fixes a priority inversion race.  Second, this
attempts to finish waking up all of the threads waiting on a turnstile
before doing a preemption.

Reviewed by: Stephan Uphoff (who found the priority inversion race)

19 years agoMinor Bug fix. Some file was not translated.
takawata [Tue, 5 Oct 2004 16:53:37 +0000 (16:53 +0000)]
Minor Bug fix. Some file was not translated.

19 years agoFixed symlinking of /etc/namedb.
ru [Tue, 5 Oct 2004 13:03:08 +0000 (13:03 +0000)]
Fixed symlinking of /etc/namedb.

Reported by: Jeremy Chadwick

19 years agoBack out changes which were introduced to delay mounting root file system.
pjd [Tue, 5 Oct 2004 11:26:43 +0000 (11:26 +0000)]
Back out changes which were introduced to delay mounting root file system.
Those changes were made on gmirror needs, but now gmirror handles this
by itself.

19 years agoBefore root file system is mounted, wait for mirrors in degraded state.
pjd [Tue, 5 Oct 2004 11:17:08 +0000 (11:17 +0000)]
Before root file system is mounted, wait for mirrors in degraded state.

19 years agoUse scheduler api to adjust thread priority.
davidxu [Tue, 5 Oct 2004 09:10:30 +0000 (09:10 +0000)]
Use scheduler api to adjust thread priority.

19 years agoMake pflogd cope with module unload (and the sudden disappearing of pflog0).
mlaier [Tue, 5 Oct 2004 08:26:34 +0000 (08:26 +0000)]
Make pflogd cope with module unload (and the sudden disappearing of pflog0).
Instead of eating all the available CPU we now shutdown gracefully.

Submitted by: yongari
MFC after: 3 days

19 years agoRe-enable descending into the "atm" subdir.
ru [Tue, 5 Oct 2004 07:47:46 +0000 (07:47 +0000)]
Re-enable descending into the "atm" subdir.

19 years agoUse generic tty code instead of local copy.
phk [Tue, 5 Oct 2004 07:42:19 +0000 (07:42 +0000)]
Use generic tty code instead of local copy.

Also divorce this driver from the sio driver.

19 years agoYet another case of resources:
imp [Tue, 5 Oct 2004 07:18:11 +0000 (07:18 +0000)]
Yet another case of resources:
+        * 9:   0x3f0-0x3f3,0x3f4-0x3f5,0x3f7

This requires only one change to support.  Rather than keying on the
size of the resource being 2, instead key off the end & 7 being 3.
This covers the same cases that the size of 2 would catch, but also
covers the new above case.

In addition, I think it is clearer to use the end in preference to the
size and start for case #8 as well.  Turns two tests into one, and
catches no other cases.

Make minor commentary changes to deal with new case #9.

# This change is specifically minimal to allow easy MFC.  A more
# extensive change will go into current once I've had a chance to test
# it on a lot of hardware...

19 years agoThis commit was generated by cvs2svn to compensate for changes in r136136,
ru [Tue, 5 Oct 2004 06:14:29 +0000 (06:14 +0000)]
This commit was generated by cvs2svn to compensate for changes in r136136,
which included commits to RCS files with non-trunk default branches.

19 years agoPull up the latest minor fixes.
ru [Tue, 5 Oct 2004 06:14:29 +0000 (06:14 +0000)]
Pull up the latest minor fixes.

19 years agoFix unionfs problems when a directory is mounted on other directory
takawata [Tue, 5 Oct 2004 05:59:29 +0000 (05:59 +0000)]
Fix unionfs problems when a directory is mounted on other directory
with different file systems. This may cause ill things
with my previous fix. Now it translate fsid of direct child of
mount point directory only.

Pointed out by: Uwe Doering

19 years agoRemove SWI_CAMNET since it's no longer used. Re-sort SWI priorities in its
scottl [Tue, 5 Oct 2004 04:52:41 +0000 (04:52 +0000)]
Remove SWI_CAMNET since it's no longer used.  Re-sort SWI priorities in its
absence.

19 years agoAdd PCI ID for VIA K8T800Pro chipset. Tested with agptest and X with DRI
anholt [Tue, 5 Oct 2004 04:40:32 +0000 (04:40 +0000)]
Add PCI ID for VIA K8T800Pro chipset.  Tested with agptest and X with DRI
enabled, but not 3D.

19 years agoRemove the camnet swi and CAM_PERIPH_NET. It has never been used, and given
scottl [Tue, 5 Oct 2004 04:22:20 +0000 (04:22 +0000)]
Remove the camnet swi and CAM_PERIPH_NET.  It has never been used, and given
that netowrk-over-scsi never really took off, there is little chance that
it will ever be needed.

19 years agoAdd taskqueue_drain. This waits for the specified task to finish, if
imp [Tue, 5 Oct 2004 04:16:01 +0000 (04:16 +0000)]
Add taskqueue_drain.  This waits for the specified task to finish, if
running, or returns.  The calling program is responsible for making sure
that nothing new is enqueued.

# man page coming soon.

19 years agoUse a taskqueue rather than an swi to handle deferred notifications.
scottl [Tue, 5 Oct 2004 04:03:00 +0000 (04:03 +0000)]
Use a taskqueue rather than an swi to handle deferred notifications.

19 years agoUpdate the man page for the -s flag and other changes that we've missed.
njl [Tue, 5 Oct 2004 02:19:30 +0000 (02:19 +0000)]
Update the man page for the -s flag and other changes that we've missed.

MFC after: 3 days

19 years agoAdd the -s flag to make dumping SSDTs optional (disabled by default).
njl [Tue, 5 Oct 2004 02:18:53 +0000 (02:18 +0000)]
Add the -s flag to make dumping SSDTs optional (disabled by default).
Since we can only override the DSDT, a custom ASL dumped previously that
contained SSDTs would result in lots of multiple definition errors.

A longer-term fix involves adding the ability to override SSDTs to ACPI-CA.

MFC after: 3 days

19 years agoAdd a new -d argument which is used to specify an alternate root for log
brooks [Mon, 4 Oct 2004 23:35:13 +0000 (23:35 +0000)]
Add a new -d argument which is used to specify an alternate root for log
files similar to DESTDIR in the BSD make process.  This only affects log
file paths not config file (-f) or archive directory (-a) paths.

19 years agoMake it more clear that if named is enabled, it will be chrooted by default.
dougb [Mon, 4 Oct 2004 20:11:34 +0000 (20:11 +0000)]
Make it more clear that if named is enabled, it will be chrooted by default.
Change to syslogd restart as suggested by des.

19 years agoThe previous commit added code to rm(1) to warn about and remove any
des [Mon, 4 Oct 2004 19:24:28 +0000 (19:24 +0000)]
The previous commit added code to rm(1) to warn about and remove any
occurrences of "/" in the argument list.  This corresponds to Enhancement
Request Number 5 in the Austin Group TC2 Aardvark's XCU Defects Report
(<URL:http://www.opengroup.org/austin/aardvark/finaltext/xcubug.txt>).
Further discussion is available in the Austin Group mailing list archives
(<URL:http://www.opengroup.org/austin/mailarchives/>, "Defect in XCU rm")
and for Austin Group members, in the Austin Group Interpretations archive
(<URL:http://www.opengroup.org/austin/interps/>, AI-019)

This commit makes that check conditional on !POSIXLY_CORRECT, since it
is not strictly correct according to the current version of the standard
(but is expected to be correct according to the next version, and has
already been adopted by Solaris).

19 years agoMarkup fixes.
des [Mon, 4 Oct 2004 19:03:44 +0000 (19:03 +0000)]
Markup fixes.

Pointed out by: ru

19 years agoBump the date.
yar [Mon, 4 Oct 2004 15:54:56 +0000 (15:54 +0000)]
Bump the date.

Reminded by: ru

19 years agofix typo in generated /etc/exports: s/synatx/syntax/
rse [Mon, 4 Oct 2004 15:13:51 +0000 (15:13 +0000)]
fix typo in generated /etc/exports: s/synatx/syntax/

19 years agoAdd a more verbose description for `device vlan'
yar [Mon, 4 Oct 2004 14:16:17 +0000 (14:16 +0000)]
Add a more verbose description for `device vlan'
to the above comment block devoted to such descriptions.

19 years agoHint a kernel builder that vlan needs miibus, which isn't obvious.
yar [Mon, 4 Oct 2004 14:06:01 +0000 (14:06 +0000)]
Hint a kernel builder that vlan needs miibus, which isn't obvious.

19 years agoMutually cross-reference sysexits(3) and err(3). Add text to err(3)
jkoshy [Mon, 4 Oct 2004 14:04:37 +0000 (14:04 +0000)]
Mutually cross-reference sysexits(3) and err(3).  Add text to err(3)
recommending that the standard exit codes in sysexits(3) be used.

Reviewed by:    ru

19 years agoDocument that vlan(4) requires miibus(4) now.
yar [Mon, 4 Oct 2004 13:57:44 +0000 (13:57 +0000)]
Document that vlan(4) requires miibus(4) now.

19 years agoChange the perfectly precise message
phk [Mon, 4 Oct 2004 13:13:23 +0000 (13:13 +0000)]
Change the perfectly precise message
       printf("No buffers busy after final sync");
to
       printf("All buffers synced.");
in order to not leave the users wondering if there should be.

19 years agoBump document date for last commit.
simon [Mon, 4 Oct 2004 12:06:05 +0000 (12:06 +0000)]
Bump document date for last commit.

Noticed by: ru

19 years agoFind out how flame-proof my underwear really is.
des [Mon, 4 Oct 2004 11:26:01 +0000 (11:26 +0000)]
Find out how flame-proof my underwear really is.

19 years agoWhitespace cleanup.
des [Mon, 4 Oct 2004 11:14:12 +0000 (11:14 +0000)]
Whitespace cleanup.

19 years agoUse tty layer generic code instead of local copy.
phk [Mon, 4 Oct 2004 09:38:53 +0000 (09:38 +0000)]
Use tty layer generic code instead of local copy.

Device names {cua,tty}R%r[.init,.lock] clashes with pty(4) driver
and allows for only 32 ports.  This should probably be revisited.

19 years agoPromote mediasize and physmem to uintmax_t when comparing them.
des [Mon, 4 Oct 2004 06:54:19 +0000 (06:54 +0000)]
Promote mediasize and physmem to uintmax_t when comparing them.

19 years agoAnother case where we need to guard against a partially
julian [Mon, 4 Oct 2004 06:45:48 +0000 (06:45 +0000)]
Another case where we need to guard against a partially
constructed process.

Submitted by: Stephan Uphoff ( ups at tree.com )
MFC after: 3 days

19 years agoWith the fixes to getty handling of non-existent devices a default
kensmith [Mon, 4 Oct 2004 03:39:52 +0000 (03:39 +0000)]
With the fixes to getty handling of non-existent devices a default
install now complains about ttyu0/ttyu1 not existing at boot time.
Since users wanting the uart based devices as terminals will need
to do something special to get them anyway set it up so a default
config doesn't complain.

MFC after: 3 days

19 years agoWhitespace nit.
des [Sun, 3 Oct 2004 23:53:54 +0000 (23:53 +0000)]
Whitespace nit.

19 years agodumpon builds cleanly at WARNS level 6.
des [Sun, 3 Oct 2004 23:39:14 +0000 (23:39 +0000)]
dumpon builds cleanly at WARNS level 6.

19 years agoVerify that the specified device is at least as large as hw.physmem.
des [Sun, 3 Oct 2004 23:38:49 +0000 (23:38 +0000)]
Verify that the specified device is at least as large as hw.physmem.

19 years agoWrap some long lines (no content changes)
des [Sun, 3 Oct 2004 23:35:34 +0000 (23:35 +0000)]
Wrap some long lines (no content changes)

19 years agoPAM configuration is now in /etc/pam.d/su.
simon [Sun, 3 Oct 2004 21:44:42 +0000 (21:44 +0000)]
PAM configuration is now in /etc/pam.d/su.

Submitted by: Jilles Tjoelker <jilles@stack.nl> (original version)
PR: docs/70616 (part of)
MFC after: 1 week

19 years agoUndo revision 1.251. This change was a performance pessimizing work-around
alc [Sun, 3 Oct 2004 20:14:07 +0000 (20:14 +0000)]
Undo revision 1.251.  This change was a performance pessimizing work-around
that is no longer required.  (In fact, it is not clear that it was ever
required in HEAD or RELENG_4, only RELENG_3 required a work-around.)  Now,
as before revision 1.251, if the preexisting PTE is invalid, pmap_enter()
does not call pmap_invalidate_page() to update the TLB(s).

Note: Even with this change, the handling of a copy-on-write fault is
inefficient, in such cases pmap_enter() calls pmap_invalidate_page() twice.

Discussed with: bde@
PR: kern/16568

19 years agoAlways strt out with an initilalised ksegrp structure.
julian [Sun, 3 Oct 2004 20:06:11 +0000 (20:06 +0000)]
Always strt out with an initilalised ksegrp structure.

MFC after: 3 days

19 years agoUse a proper prototype for hertz().
stefanf [Sun, 3 Oct 2004 18:22:35 +0000 (18:22 +0000)]
Use a proper prototype for hertz().

19 years agoUse the correct printf specifier.
stefanf [Sun, 3 Oct 2004 16:34:01 +0000 (16:34 +0000)]
Use the correct printf specifier.

PR: 47187

19 years agoThe macro for the function specifier inline is spelled '__inline'.
stefanf [Sun, 3 Oct 2004 16:12:29 +0000 (16:12 +0000)]
The macro for the function specifier inline is spelled '__inline'.

19 years agoUse the correct type for iop_attach().
stefanf [Sun, 3 Oct 2004 16:06:46 +0000 (16:06 +0000)]
Use the correct type for iop_attach().

19 years agoDon't add integers to void pointers.
stefanf [Sun, 3 Oct 2004 15:58:20 +0000 (15:58 +0000)]
Don't add integers to void pointers.

19 years agoDon't add integers to void pointers.
stefanf [Sun, 3 Oct 2004 15:48:32 +0000 (15:48 +0000)]
Don't add integers to void pointers.

19 years agoDirectly include <runetype.h> for _CurrentRuneLocale, <_ctype.h> doesn't
stefanf [Sun, 3 Oct 2004 15:42:59 +0000 (15:42 +0000)]
Directly include <runetype.h> for _CurrentRuneLocale, <_ctype.h> doesn't
include it in all cases.

19 years agoProperly initialise 'filename' so that random -l doesn't try to open NULL.
stefanf [Sun, 3 Oct 2004 15:34:15 +0000 (15:34 +0000)]
Properly initialise 'filename' so that random -l doesn't try to open NULL.

Const-qualify 'filename' to avoid a strdup() call due to -Wwrite-strings
silliness.

19 years agochflags compiles with WARNS?=6.
stefanf [Sun, 3 Oct 2004 15:03:15 +0000 (15:03 +0000)]
chflags compiles with WARNS?=6.

19 years agoGarbage collect a variable whose value was never read.
stefanf [Sun, 3 Oct 2004 14:55:36 +0000 (14:55 +0000)]
Garbage collect a variable whose value was never read.

19 years agoDon't rely on NULL being a pointer, add a cast before passing it to a variadic
stefanf [Sun, 3 Oct 2004 14:40:27 +0000 (14:40 +0000)]
Don't rely on NULL being a pointer, add a cast before passing it to a variadic
function.

19 years agoDon't bother to turn off other P_STOPPED bits for SIGKILL, doing
davidxu [Sun, 3 Oct 2004 13:23:49 +0000 (13:23 +0000)]
Don't bother to turn off other P_STOPPED bits for SIGKILL, doing
so would cause kernel to produce an unkillable process in some cases,
especially, P_STOPPED_SINGLE has a singling thread, turning off the
bit would mess the state.

19 years agoAdd a SYNOPSIS section for io(4), mem(4), or random(4), since they are
simon [Sun, 3 Oct 2004 13:14:20 +0000 (13:14 +0000)]
Add a SYNOPSIS section for io(4), mem(4), or random(4), since they are
now optional devices.

MFC after: 1 week

19 years agoDocument a problem with user/group filtering. With debug.mpsafenet=1 this
mlaier [Sun, 3 Oct 2004 10:42:42 +0000 (10:42 +0000)]
Document a problem with user/group filtering. With debug.mpsafenet=1 this
might result in a deadlock. The fix involves critical changes in the PF
locking strategy (which will happen after 5.3R). For now advise users to set
debug.mpsafenet=0 if they use this kind of filtering.

The same problem exists for IPFW.

mdoc help from: simon
MFC after: 2 days

19 years agoDocument how to set up libmap.conf to deal with the library version bump.
des [Sun, 3 Oct 2004 10:08:12 +0000 (10:08 +0000)]
Document how to set up libmap.conf to deal with the library version bump.

19 years agoAdd device ID for atuwi USB wlan driver,
sanpei [Sun, 3 Oct 2004 09:30:09 +0000 (09:30 +0000)]
Add device ID for atuwi USB wlan driver,
(Atmel at76c503a http://vitsch.net/bsd/atuwi)
o AINCOMM AWU2000B
o ATMEL WL1130USB

PR: kern/72195
Submitted by: Daan Vreeken [PA4DAN] <Danovitsch@Vitsch.net>
MFC after: 1 week

19 years agoAdd a HARDWARE section which lists supported devices.
sanpei [Sun, 3 Oct 2004 09:10:15 +0000 (09:10 +0000)]
Add a HARDWARE section which lists supported devices.
EPSON GT-8400UF(This product is same device id, 0x011f
as EPSON Perfection 1670 scanner)

PR: [FreeBSD-users-jp 81081]
Submitted by: TOKIO Yukiya <yukiya@ee.t-kougei.ac.jp>
MFC after: 3 days

19 years agoRemove blindly-copied extra include path.
green [Sun, 3 Oct 2004 06:32:37 +0000 (06:32 +0000)]
Remove blindly-copied extra include path.

19 years agoRemove -I from CFLAGS. This splipped in with the 3.5 import (as I was
mlaier [Sun, 3 Oct 2004 06:19:22 +0000 (06:19 +0000)]
Remove -I from CFLAGS. This splipped in with the 3.5 import (as I was
building on a box with older pfvar.h installed). Didn't intend to commit it.

Requested by: ru (on a C&P to ipfw's Makefile)

19 years agoThe print/acroread4 port seems to be gone. acroread5 doesn't run on
kensmith [Sun, 3 Oct 2004 05:20:55 +0000 (05:20 +0000)]
The print/acroread4 port seems to be gone.  acroread5 doesn't run on
Alpha so that leaves acroread (3).  Users are probably better off
with xpdf so just drop acroread completely.

MFC after: 3 days

19 years agoPrevent reentrancy of the IPv6 routing code (leading to crash with
green [Sun, 3 Oct 2004 00:49:33 +0000 (00:49 +0000)]
Prevent reentrancy of the IPv6 routing code (leading to crash with
INVARIANTS on, who knows what with it off).

19 years agoAdd support to IPFW for matching by TCP data length.
green [Sun, 3 Oct 2004 00:47:15 +0000 (00:47 +0000)]
Add support to IPFW for matching by TCP data length.

19 years agoAdd the documentation for IPFW's diverted(-loopback|-output) matches.
green [Sun, 3 Oct 2004 00:35:52 +0000 (00:35 +0000)]
Add the documentation for IPFW's diverted(-loopback|-output) matches.

19 years agoAdd support to IPFW for classification based on "diverted" status
green [Sun, 3 Oct 2004 00:26:35 +0000 (00:26 +0000)]
Add support to IPFW for classification based on "diverted" status
(that is, input via a divert socket).

19 years agoRemove accidentally-added O_DIVERTED section.
green [Sun, 3 Oct 2004 00:24:42 +0000 (00:24 +0000)]
Remove accidentally-added O_DIVERTED section.

19 years agoAdd to IPFW the ability to do ALTQ classification/tagging.
green [Sun, 3 Oct 2004 00:17:46 +0000 (00:17 +0000)]
Add to IPFW the ability to do ALTQ classification/tagging.

19 years agoThe physical address stored in the vm_page is page aligned. There is no
alc [Sun, 3 Oct 2004 00:16:43 +0000 (00:16 +0000)]
The physical address stored in the vm_page is page aligned.  There is no
need to mask off the page offset bits.  (This operation made some sense
prior to i386/i386/pmap.c revision 1.254 when we passed a physical address
rather than a vm_page pointer to pmap_enter().)

19 years agoDisable MTU feedback in IPv6 if the sender writes data that must be fragmented.
dwhite [Sat, 2 Oct 2004 23:45:02 +0000 (23:45 +0000)]
Disable MTU feedback in IPv6 if the sender writes data that must be fragmented.
Discussed extensively with KAME.  The API author's intent isn't clear at this
point, so rather than remove the code entirely, #if 0 out and put a big
comment in for now. The IPV6_RECVPATHMTU sockopt is available if the
application wants to be notified of the path MTU to optimize packet sizes.

Thanks to JINMEI Tatuya <jinmei@isl.rdc.toshiba.co.jp> for putting up
with my incessant badgering on this issue, and fenner for pointing out
the API issue and suggesting solutions.

19 years ago* Use two cdevsw's for ugen(4): one for control endpoints, and one for
green [Sat, 2 Oct 2004 22:49:54 +0000 (22:49 +0000)]
* Use two cdevsw's for ugen(4): one for control endpoints, and one for
  data endpoints.  The control endpoint doesn't need read/write/poll
  operations, and more importantly, the thread counts should be
  separate so that the control endpoint can properly reference itself
  while deleting and recreating the data endpoints.
* Add some macros that handle referencing/releasing devices, and use them
  for sleeping/woken-up and open/close operations as apppropriate.
* Use d_purge for FreeBSD, and a loop testing the open status for all
  the endpoints for NetBSD and OpenBSD, so that when the device is
  detached, the right thing always happens.

19 years ago* When toggling short transfers on a bulk transfer endpoint, cancel and
green [Sat, 2 Oct 2004 22:33:26 +0000 (22:33 +0000)]
* When toggling short transfers on a bulk transfer endpoint, cancel and
  restart the current waiting transfer.  If this isn't done, the device's
  next transfer (that we would like to do a short read on) is going to
  return an error -- for short transfer.
* For bulk transfer endpoints, restore the maximum transfer length each
  time a transfer is done, or the first short transfer will make all the
  rest that size or smaller.
* Remove impossibilities (malloc(M_WAITOK) == NULL, &var == NULL).

19 years agoDon't allow to create a drive that already exists.
le [Sat, 2 Oct 2004 20:50:21 +0000 (20:50 +0000)]
Don't allow to create a drive that already exists.