]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years agoMake end of frames for KSE thread, for system scope thread, without this
David Xu [Tue, 20 Jul 2004 01:38:59 +0000 (01:38 +0000)]
Make end of frames for KSE thread, for system scope thread, without this
change, debugger will dump a weird stack backtrace.

20 years agoFix printing of long doubles to match the size that
Andrew Gallatin [Mon, 19 Jul 2004 23:56:07 +0000 (23:56 +0000)]
Fix printing of long doubles to match the size  that
gcc is using.  This fixes devstat consumers (like vmstat, iostat,
systat) so they don't print crazy zillion digit numbers for
disk transfers and bandwidth.

According to gcc, long doubles are 64-bits, rather than 128 bits
like the SVR4 ABI spec wants them to be..  Note that MacOSX also treats
long doubles as 64-bits, and not 128 bits, so we are in good company.

Reviewed by: das
Approved by: grehan

20 years agoRemove extraneous locks on the VM free page queue mutex; it is not
Brian Feldman [Mon, 19 Jul 2004 23:29:36 +0000 (23:29 +0000)]
Remove extraneous locks on the VM free page queue mutex;  it is not
meant to be recursed upon, and could cauuse a deadlock inside the
new contigmalloc (vm.old_contigmalloc=0) code.

Submitted by: alc

20 years agoIdentify VIA EHCI root hubs and at least one VIA USB2.0 controller.
Benno Rice [Mon, 19 Jul 2004 23:22:10 +0000 (23:22 +0000)]
Identify VIA EHCI root hubs and at least one VIA USB2.0 controller.

20 years agoLet IN_FASTREOCOVERY macro decide if we are in recovery mode.
Jayanth Vijayaraghavan [Mon, 19 Jul 2004 22:37:33 +0000 (22:37 +0000)]
Let IN_FASTREOCOVERY macro decide if we are in recovery mode.

Nuke sackhole_limit for now. We need to add it back to limit the total
number of sack blocks in the system.

20 years agoFix a potential panic in the SACK code that was causing
Jayanth Vijayaraghavan [Mon, 19 Jul 2004 22:06:01 +0000 (22:06 +0000)]
Fix a potential panic in the SACK code that was causing
1) data to be sent to the right of snd_recover.
2) send more data then whats in the send buffer.

The fix is to postpone sack retransmit to a subsequent recovery episode
if the current retransmit pointer is beyond snd_recover.

Thanks to Mohan Srinivasan for helping fix the bug.

Submitted by:Daniel Lang

20 years agoDiff reduction to NetBSD.
Lukas Ertl [Mon, 19 Jul 2004 20:49:02 +0000 (20:49 +0000)]
Diff reduction to NetBSD.

usbdi.c rev. 1.104, author: mycroft
   ugen_isoc_rintr() may recycle the xfer immediately.  Therefore, we
   avoid touching the xfer after calling the callback in
   usb_transfer_complete().  From PR 25960.

20 years agoDiff reduction to NetBSD.
Lukas Ertl [Mon, 19 Jul 2004 20:47:46 +0000 (20:47 +0000)]
Diff reduction to NetBSD.

ehci.c rev. 1.69, author: mycroft
uhci.c rev. 1.179, author: mycroft
   hcpriv is not actually used here.  Remove references to it.

Obtained from:   NetBSD

20 years ago - Eliminate the pte object from the pmap. Instead, page table pages are
Alan Cox [Mon, 19 Jul 2004 18:12:04 +0000 (18:12 +0000)]
 - Eliminate the pte object from the pmap.  Instead, page table pages are
   allocated as "no object" pages.  Similar changes were made to the amd64
   and i386 pmap last year.  The primary reason being that maintaining
   a pte object leads to lock order violations.  A secondary reason being
   that the pte object is redundant, i.e., the page table itself can be
   used to lookup page table pages.  (Historical note: The pte object
   predates our ability to allocate "no object" pages.  Thus, the pte
   object was a necessary evil.)
 - Unconditionally check the vm object lock's status in vm_page_remove().
   Previously, this assertion could not be made on Alpha due to its use
   of a pte object.

20 years agoYou always spot the typos after you have committed.. Start sentence
Julian Elischer [Mon, 19 Jul 2004 18:06:12 +0000 (18:06 +0000)]
You always spot the typos after you have committed.. Start sentence
with a Cap.

20 years agoAllow the user who calls doadump() from the kernel debugger
Julian Elischer [Mon, 19 Jul 2004 18:03:02 +0000 (18:03 +0000)]
Allow the user who calls doadump() from the kernel debugger
to not get a page fault if he has not defined a dump device.
Panic can often not do a dump as it can hang forever in some cases.
 The original PR was for amd64 only. This is a generalised version of
that change.

PR: amd64/67712
Submitted by: wjw@withagen.nl <Willen Jan Withagen>

20 years agoFurther function forward declaration white space tweaks.
Robert Watson [Mon, 19 Jul 2004 17:18:58 +0000 (17:18 +0000)]
Further function forward declaration white space tweaks.

20 years agoRe-style at_control.c to bring it closer to style(9), primarily with
Robert Watson [Mon, 19 Jul 2004 17:15:51 +0000 (17:15 +0000)]
Re-style at_control.c to bring it closer to style(9), primarily with
regard to function prototypes and indentation.  The lack of indentation
in if clauses and case statements made this code extremely difficult
to read.

20 years agoprgregset_t changed type from being a typedef of struct reg to an
Marcel Moolenaar [Mon, 19 Jul 2004 16:54:52 +0000 (16:54 +0000)]
prgregset_t changed type from being a typedef of struct reg to an
array of one element of type struct reg. Change the argument to
libc_r_md_getgregs() accordingly.

20 years agoAs a temporary hack, turn off deferred preemptions that are the result of
John Baldwin [Mon, 19 Jul 2004 16:37:47 +0000 (16:37 +0000)]
As a temporary hack, turn off deferred preemptions that are the result of
a fast interrupt handler doing an swi_sched().  This fixed the lockups I
saw on my laptop when using xmms in KDE and on rwatson's MySQL benchmarks
on SMP.  This will eventually be removed and/or modified when I figure out
what the root cause is and fix that.

20 years agoSince breakage of malloc(9)/uma_zalloc(9) is totally non-optional in
Brian Feldman [Mon, 19 Jul 2004 15:05:24 +0000 (15:05 +0000)]
Since breakage of malloc(9)/uma_zalloc(9) is totally non-optional in
GENERIC/for WITNESS users, make sure the sysctl to disable the behavior
is read-only and always enabled.

20 years agoGuard against argv[0] being NULL.
Tim Kientzle [Mon, 19 Jul 2004 14:54:38 +0000 (14:54 +0000)]
Guard against argv[0] being NULL.

Thanks to: Tim J Robbins

20 years agoMake it clearer what means 'won't work' for .if string == ${VAR}.
Hartmut Brandt [Mon, 19 Jul 2004 14:42:57 +0000 (14:42 +0000)]
Make it clearer what means 'won't work' for .if string == ${VAR}.

Replace the use of '=' in conditionals in the examples
by the more correct '=='.

Clarify the example explaining that .for expansion takes place before
.if handling by showing the correct code instead of saying 'the other
way around'. Change a variable name there so the example is more parseable
to the human reader.

PR: docs/65400
Submitted by: Roman Neuhauser <neuhauser@chello.cz>

20 years agoAvoid passing negative values to tolower() on machines with signed chars.
Tim J. Robbins [Mon, 19 Jul 2004 12:57:24 +0000 (12:57 +0000)]
Avoid passing negative values to tolower() on machines with signed chars.

20 years agoFix a typo that could provoke a panic or access to random memory.
Hartmut Brandt [Mon, 19 Jul 2004 12:54:00 +0000 (12:54 +0000)]
Fix a typo that could provoke a panic or access to random memory.

PR: kern/67012
Submitted by: Zhenmin <zli4@cs.uiuc.edu>

20 years agoMake kdb_backtrace() sort of work.
Olivier Houchard [Mon, 19 Jul 2004 12:26:44 +0000 (12:26 +0000)]
Make kdb_backtrace() sort of work.

20 years agoEnable libpthread build for powerpc
Peter Grehan [Mon, 19 Jul 2004 12:20:10 +0000 (12:20 +0000)]
Enable libpthread build for powerpc

20 years agoPPC MD bits for KSE. Runs test cases OK. Crippled to 1:1 mode for
Peter Grehan [Mon, 19 Jul 2004 12:19:04 +0000 (12:19 +0000)]
PPC MD bits for KSE. Runs test cases OK. Crippled to 1:1 mode for
the time being.

20 years agoAdd signalcontext, required by KSE.
Peter Grehan [Mon, 19 Jul 2004 12:08:03 +0000 (12:08 +0000)]
Add signalcontext, required by KSE.

20 years agoThe new program counter should go into the trapframe's srr0 and
Peter Grehan [Mon, 19 Jul 2004 12:05:07 +0000 (12:05 +0000)]
The new program counter should go into the trapframe's srr0 and
not the link register, which was lucky enough to work.

20 years agoSort sections.
Tim J. Robbins [Mon, 19 Jul 2004 11:21:34 +0000 (11:21 +0000)]
Sort sections.

20 years agoMove exit status information into a DIAGNOSTICS section. Add an ENVIRONMENT
Tim J. Robbins [Mon, 19 Jul 2004 11:18:56 +0000 (11:18 +0000)]
Move exit status information into a DIAGNOSTICS section. Add an ENVIRONMENT
section. Re-add a sentence from the BUGS section that went missing in
the previous commit.

20 years agoMFi386: revision 1.596.
Yoshihiro Takahashi [Mon, 19 Jul 2004 11:17:57 +0000 (11:17 +0000)]
MFi386: revision 1.596.

20 years agoAdd support for multibyte characters. While here, fix a longstanding bug in
Tim J. Robbins [Mon, 19 Jul 2004 11:12:02 +0000 (11:12 +0000)]
Add support for multibyte characters. While here, fix a longstanding bug in
the implementation of the -d option: we were skipping too many characters
when a non-alphanumeric character was encountered.

20 years agoPoint out in the BUGS section that look expects input files to have
Tim J. Robbins [Mon, 19 Jul 2004 10:03:38 +0000 (10:03 +0000)]
Point out in the BUGS section that look expects input files to have
been sorted with LC_COLLATE=C.

20 years agoAdd a skeleton makefile that runs the tests out of libc/regex/grot.
Tim J. Robbins [Mon, 19 Jul 2004 09:00:26 +0000 (09:00 +0000)]
Add a skeleton makefile that runs the tests out of libc/regex/grot.

20 years agoOops, test error behaviour of wctrans(), not wctype().
Tim J. Robbins [Mon, 19 Jul 2004 08:53:41 +0000 (08:53 +0000)]
Oops, test error behaviour of wctrans(), not wctype().

20 years agoUpdate paths to reg*.c and regex2.h. Add a target to build regex.h.
Tim J. Robbins [Mon, 19 Jul 2004 08:48:17 +0000 (08:48 +0000)]
Update paths to reg*.c and regex2.h. Add a target to build regex.h.

20 years agoUpdate for removal of cclass.h. Trim some useless targets. Invoke mkh
Tim J. Robbins [Mon, 19 Jul 2004 08:41:11 +0000 (08:41 +0000)]
Update for removal of cclass.h. Trim some useless targets. Invoke mkh
with "sh mkh" so it works if the script is not executable.

20 years agoUpdate for recent changes to struct re_guts. Disable printing the contents
Tim J. Robbins [Mon, 19 Jul 2004 08:28:53 +0000 (08:28 +0000)]
Update for recent changes to struct re_guts. Disable printing the contents
of OANYOF sets for the moment.

20 years agoRemove unused files.
Tim J. Robbins [Mon, 19 Jul 2004 08:24:21 +0000 (08:24 +0000)]
Remove unused files.

20 years agoForced commit to note that the previous commit message doesn't apply
David Schultz [Mon, 19 Jul 2004 08:24:04 +0000 (08:24 +0000)]
Forced commit to note that the previous commit message doesn't apply
to the ARM port.  We set FLT_ROUNDS to -1 (indeterminate), because the
rounding mode on ARM is static, i.e. part of the FP instruction
format.  Or at least that's my understanding.

20 years agoToday is a good day to add myself here :)
Gleb Smirnoff [Mon, 19 Jul 2004 08:21:17 +0000 (08:21 +0000)]
Today is a good day to add myself here :)

Approved by: julian (mentor)

20 years agoWhile testing fe[gs]etround(), make sure FLT_ROUNDS reflects the
David Schultz [Mon, 19 Jul 2004 08:17:47 +0000 (08:17 +0000)]
While testing fe[gs]etround(), make sure FLT_ROUNDS reflects the
proper rounding mode as well.

20 years agoMake FLT_ROUNDS correctly reflect the dynamic rounding mode.
David Schultz [Mon, 19 Jul 2004 08:17:25 +0000 (08:17 +0000)]
Make FLT_ROUNDS correctly reflect the dynamic rounding mode.

20 years agoFix two bugs in the signbit() macro, which was implemented last year:
David Schultz [Mon, 19 Jul 2004 08:16:10 +0000 (08:16 +0000)]
Fix two bugs in the signbit() macro, which was implemented last year:

- It was added to libc instead of libm.  Hopefully no programs rely
  on this mistake.

- It didn't work properly on large long doubles because its argument
  was converted to type double, resulting in undefined behavior.

20 years agoMFp4: Add two options for gnop(8)'s 'create' command:
Pawel Jakub Dawidek [Mon, 19 Jul 2004 07:52:56 +0000 (07:52 +0000)]
MFp4: Add two options for gnop(8)'s 'create' command:
-o offset - specifies where to start on the original provider
-s size - specifies size of the transparent provider

20 years agoUse the version field to identify the partial context used by
Peter Grehan [Mon, 19 Jul 2004 07:21:46 +0000 (07:21 +0000)]
Use the version field to identify the partial context used by
KSE process-scope threads.

20 years agoReimplement contigmalloc(9) with an algorithm which stands a greatly-
Brian Feldman [Mon, 19 Jul 2004 06:21:27 +0000 (06:21 +0000)]
Reimplement contigmalloc(9) with an algorithm which stands a greatly-
improved chance of working despite pressure from running programs.
Instead of trying to throw a bunch of pages out to swap and hope for
the best, only a range that can potentially fulfill contigmalloc(9)'s
request will have its contents paged out (potentially, not forcibly)
at a time.

The new contigmalloc operation still operates in three passes, but it
could potentially be tuned to more or less.  The first pass only looks
at pages in the cache and free pages, so they would be thrown out
without having to block.  If this is not enough, the subsequent passes
page out any unwired memory.  To combat memory pressure refragmenting
the section of memory being laundered, each page is removed from the
systems' free memory queue once it has been freed so that blocking
later doesn't cause the memory laundered so far to get reallocated.

The page-out operations are now blocking, as it would make little sense
to try to push out a page, then get its status immediately afterward
to remove it from the available free pages queue, if it's unlikely to
have been freed.  Another change is that if KVA allocation fails, the
allocated memory segment will be freed and not leaked.

There is a sysctl/tunable, defaulting to on, which causes the old
contigmalloc() algorithm to be used.  Nonetheless, I have been using
vm.old_contigmalloc=0 for over a month.  It is safe to switch at
run-time to see the difference it makes.

A new interface has been used which does not require mapping the
allocated pages into KVA: vm_page.h functions vm_page_alloc_contig()
and vm_page_release_contig().  These are what vm.old_contigmalloc=0
uses internally, so the sysctl/tunable does not affect their operation.

When using the contigmalloc(9) and contigfree(9) interfaces, memory
is now tracked with malloc(9) stats.  Several functions have been
exported from kern_malloc.c to allow other subsystems to use these
statistics, as well.  This invalidates the BUGS section of the
contigmalloc(9) manpage.

20 years agoAdd partial pmap locking.
Alan Cox [Mon, 19 Jul 2004 05:39:49 +0000 (05:39 +0000)]
Add partial pmap locking.

Tested by: marcel@

20 years agoFix some misspellings, document the TAPE environment
Tim Kientzle [Mon, 19 Jul 2004 05:24:41 +0000 (05:24 +0000)]
Fix some misspellings, document the TAPE environment
variable and the default tape device.

20 years agoAdd a #error requiring KDB if DDB is specified. (This can probably be
Mike Silbersack [Mon, 19 Jul 2004 02:46:34 +0000 (02:46 +0000)]
Add a #error requiring KDB if DDB is specified.  (This can probably be
relocated to a better place, if one exists.)

20 years agoEmpty GENERIC.hints file needed by make release.
Peter Grehan [Mon, 19 Jul 2004 02:08:22 +0000 (02:08 +0000)]
Empty GENERIC.hints file needed by make release.

Noticed by:  Suleiman Souhlal <refugee@segfaulted.com>

20 years agoThis file was not used to import gdb. Remove.
Marcel Moolenaar [Mon, 19 Jul 2004 00:20:33 +0000 (00:20 +0000)]
This file was not used to import gdb. Remove.

20 years agoWhen calling scheduler entrypoints for creating new threads and processes,
Julian Elischer [Sun, 18 Jul 2004 23:36:13 +0000 (23:36 +0000)]
When calling scheduler entrypoints for creating new threads and processes,
specify "us" as the thread not the process/ksegrp/kse.
You can always find the others from the thread but the converse is not true.
Theorotically this would lead to runtime being allocated to the wrong
entity in some cases though it is not clear how often this actually happenned.
(would only affect threaded processes and would probably be pretty benign,
but it WAS a bug..)

Reviewed by: peter

20 years agoNow we have NO_ADAPTIVE_MUTEXES option, so use it here too.
Pawel Jakub Dawidek [Sun, 18 Jul 2004 23:27:14 +0000 (23:27 +0000)]
Now we have NO_ADAPTIVE_MUTEXES option, so use it here too.

Missed by: scottl

20 years agoDon't forget the arguments for -M and -N in the DESCRIPTION section.
Stefan Farfeleder [Sun, 18 Jul 2004 23:05:31 +0000 (23:05 +0000)]
Don't forget the arguments for -M and -N in the DESCRIPTION section.

20 years agoReverse a lock/unlock pair that were the wrong way around in some code that
Julian Elischer [Sun, 18 Jul 2004 22:57:46 +0000 (22:57 +0000)]
Reverse a lock/unlock pair that were the wrong way around in some code that
is obviously not run a lot. (but is in some test cases).
This code is not usually run because it covers a case that doesn't
happen a lot (removing a node that has data traversing it).

20 years agoComment clarifying debug_mpsafenet.
Robert Watson [Sun, 18 Jul 2004 21:50:22 +0000 (21:50 +0000)]
Comment clarifying debug_mpsafenet.

20 years agoClarify boot.8 (i386 version) a bit, so that people know
Mike Silbersack [Sun, 18 Jul 2004 21:39:45 +0000 (21:39 +0000)]
Clarify boot.8 (i386 version) a bit, so that people know
where boot.config needs to reside.  Also change /kernel
to /boot/loader, as that is the apparent default now.  This
man page probably requires more updates.

20 years agoReplace seven nominally MD implementations of frexp() that are broken
David Schultz [Sun, 18 Jul 2004 21:23:39 +0000 (21:23 +0000)]
Replace seven nominally MD implementations of frexp() that are broken
for subnormals with one implementation that works.

20 years agoUtilize pmap_pte_quick() rather than pmap_pte() in pmap_protect(). The
Alan Cox [Sun, 18 Jul 2004 21:19:10 +0000 (21:19 +0000)]
Utilize pmap_pte_quick() rather than pmap_pte() in pmap_protect().  The
reason being that pmap_pte_quick() requires the page queues lock, which is
already held, rather than Giant.

20 years agoAfter maintaining previous behaviour in writing out the core notes, it's
Marcel Moolenaar [Sun, 18 Jul 2004 20:28:07 +0000 (20:28 +0000)]
After maintaining previous behaviour in writing out the core notes, it's
time now to break with the past: do not write the PID in the first note.
Rationale:
1.  [impact of the breakage] Process IDs in core files serve no immediate
    purpose to the debugger itself. They are only useful to relate a core
    file to a process. This can provide context to the person looking at
    the core file, provided one keeps track of this. Overall, not having
    the PID in the core file is only in very rare occasions unfortunate.
2.  [reason of the breakage] Having one PRSTATUS note contain the PID,
    while all others contain the LWPID of the corresponding kernel thread
    creates an irregularity for the debugger that cannot easily be worked
    around. This is caused by libthread_db correlating user thread IDs to
    kernel thread (aka LWP) IDs and thus aware of the actual LWPIDs.

Update comments accordingly.

20 years agoFix a possible hang which apparently occurs during a warm boot (cold boot
Gary Jennejohn [Sun, 18 Jul 2004 20:13:31 +0000 (20:13 +0000)]
Fix a possible hang which apparently occurs during a warm boot (cold boot
does not display the symptom). Evidently the ifpi2 controller needs to be
massaged more than it was.

Note that this does not close the PR since it was filed against 4.9.

MFC: 5 days
PR: kern/68756
Submitted by: Ari Suutari <ari.suutari@syncrontech.com>

20 years agoGratuitous whitespace change to un-wrap a short line.
Robert Watson [Sun, 18 Jul 2004 19:53:35 +0000 (19:53 +0000)]
Gratuitous whitespace change to un-wrap a short line.

20 years agoThis library is not WARNS=2 clean for -O2 builds, because we include
Marcel Moolenaar [Sun, 18 Jul 2004 19:33:56 +0000 (19:33 +0000)]
This library is not WARNS=2 clean for -O2 builds, because we include
headers from libpthread that are not WARNS=2 clean for -O2 builds.
Lower the WARNS level to 1. This is the highest level possible for
now.

20 years agoCorrect the mess I made by committing from the wrong tree. Most
Marcel Moolenaar [Sun, 18 Jul 2004 19:29:38 +0000 (19:29 +0000)]
Correct the mess I made by committing from the wrong tree. Most
notably, this restores some of the contents in thread_db.h as well
as David Xu's copyright notice. This also fixes the includes in
the MD libpthread files which Scott tried to provide a quick fix
for.

Pointy hat: marcel

20 years agoThe recent changes to control message passing broke some things
David Malone [Sun, 18 Jul 2004 19:10:36 +0000 (19:10 +0000)]
The recent changes to control message passing broke some things
that get certain types of control messages (ping6 and rtsol are
examples). This gets the new code closer to working:

1) Collect control mbufs for processing in the controlp ==
NULL case, so that they can be freed by externalize.

2) Loop over the list of control mbufs, as the externalize
function may not know how to deal with chains.

3) In the case where there is no externalize function,
remember to add the control mbuf to the controlp list so
that it will be returned.

4) After adding stuff to the controlp list, walk to the
end of the list of stuff that was added, incase we added
a chain.

This code can be further improved, but this is enough to get most
things working again.

Reviewed by: rwatson

20 years agoRename a preposterously long name with one that might look a bit
Mark Murray [Sun, 18 Jul 2004 18:55:31 +0000 (18:55 +0000)]
Rename a preposterously long name with one that might look a bit
more professional. While here, write a few lines of explanatory
text to explain what its for.

Discussed with: rwatson
With hat: core

20 years agoUnbreak kernel compiles by preserving an old opt_adaptive_mutexes.h file
Alexander Kabaev [Sun, 18 Jul 2004 18:21:39 +0000 (18:21 +0000)]
Unbreak kernel compiles by preserving an old opt_adaptive_mutexes.h file
name.

20 years agoFor the gbde attach script:
Simon L. B. Nielsen [Sun, 18 Jul 2004 18:01:48 +0000 (18:01 +0000)]
For the gbde attach script:
- Ask the user up to X times (3 by default) for the pass-phrase, if
  it is incorrect the first time.
- Add support for storing the lockfiles in another other directory
  than /etc.
- Document that it is possible to override the location of each single
  lockfile.

Approved by: pjd

20 years agoFix copy&paste bug.
Pawel Jakub Dawidek [Sun, 18 Jul 2004 16:51:58 +0000 (16:51 +0000)]
Fix copy&paste bug.

20 years agoAdd doxygen doc comments for most of newbus and the BUS interface.
Doug Rabson [Sun, 18 Jul 2004 16:30:31 +0000 (16:30 +0000)]
Add doxygen doc comments for most of newbus and the BUS interface.

20 years agoEnable ADAPTIVE_MUTEXES by default by changing the sense of the option to
Scott Long [Sun, 18 Jul 2004 15:59:03 +0000 (15:59 +0000)]
Enable ADAPTIVE_MUTEXES by default by changing the sense of the option to
NO_ADAPTIVE_MUTEXES.  This option has been enabled by default on amd64 for
quite some time, and has been extensively tested on i386 and sparc64.  It
shows measurable performance gains in many circumstances, and few negative
effects.  It would be nice in t he future if adaptive mutexes actually went
to sleep after a certain amount of spinning, but that will require quite a
bit more testing.

20 years agoTry to fix ia64 and alpha compiles. I don't have either equipment fired
Scott Long [Sun, 18 Jul 2004 15:24:37 +0000 (15:24 +0000)]
Try to fix ia64 and alpha compiles.  I don't have either equipment fired
up now, but it appears to be the same problem and solution as sparc64.

20 years agoAdd missing #includes so that this can compile. Obtained from the i386 version.
Scott Long [Sun, 18 Jul 2004 15:20:03 +0000 (15:20 +0000)]
Add missing #includes so that this can compile.  Obtained from the i386 version.

20 years agoFix a stupid attemp to apply host arithmetics to network byte ordered data.
Max Laier [Sun, 18 Jul 2004 14:25:48 +0000 (14:25 +0000)]
Fix a stupid attemp to apply host arithmetics to network byte ordered data.
This fixes checksum for some drivers with partial H/W ckcsum offloads.

Reported by: Simon 'corecode' Schubert, Devon H. O'Dell, hmp
Reviewed by: Pyun YongHyeon

20 years agoRemove claim of conformance to IEEE Std. 1003.2. Replace it with a list
Tim J. Robbins [Sun, 18 Jul 2004 10:11:27 +0000 (10:11 +0000)]
Remove claim of conformance to IEEE Std. 1003.2. Replace it with a list
of features required by the standard that the current implementation
does not support.

PR: 57911 (related)

20 years ago`off' is a keyword, not a parameter name.
Maxim Konovalov [Sun, 18 Jul 2004 09:57:47 +0000 (09:57 +0000)]
`off' is a keyword, not a parameter name.

Obtained from: DragonFly

20 years agoI missed two pieces of the commit to this file. Robert has already
David Malone [Sun, 18 Jul 2004 09:26:34 +0000 (09:26 +0000)]
I missed two pieces of the commit to this file. Robert has already
added one, this adds the other.

20 years agoStart the entropy device insecure/unblocked. I'll be handing over
Mark Murray [Sun, 18 Jul 2004 09:07:58 +0000 (09:07 +0000)]
Start the entropy device insecure/unblocked. I'll be handing over
responsibility for critical randomness requirements (like sshd)
to rc.d/*

Requested by: many

20 years agoIn -CURRENT pseudo devices are not statically assigned at compile time,
Maxim Konovalov [Sun, 18 Jul 2004 09:03:12 +0000 (09:03 +0000)]
In -CURRENT pseudo devices are not statically assigned at compile time,
remove a stale comment.

PR: kern/62285

20 years agoMFp4: Add 'dump' command to gconcat(8), glabel(8) and gstripe(8) which allow
Pawel Jakub Dawidek [Sun, 18 Jul 2004 08:00:30 +0000 (08:00 +0000)]
MFp4: Add 'dump' command to gconcat(8), glabel(8) and gstripe(8) which allow
      to dump metadata from given components.

20 years agoRemove claim of conformance to IEEE Std. 1003.2. Replace it with a list
Tim J. Robbins [Sun, 18 Jul 2004 06:56:40 +0000 (06:56 +0000)]
Remove claim of conformance to IEEE Std. 1003.2. Replace it with a list
of features required by the standard that the current implementation
does not support.

PR: 57911

20 years agoFix exclusive-bit leakage.
Pawel Jakub Dawidek [Sun, 18 Jul 2004 06:54:29 +0000 (06:54 +0000)]
Fix exclusive-bit leakage.

20 years agoAdd support System TALKS Inc. SGC-X2UL
MIHIRA Sanpei Yoshiro [Sun, 18 Jul 2004 06:50:58 +0000 (06:50 +0000)]
Add support System TALKS Inc. SGC-X2UL

20 years agoAdd support System TALKS Inc. SGC-X2UL
MIHIRA Sanpei Yoshiro [Sun, 18 Jul 2004 06:45:38 +0000 (06:45 +0000)]
Add support System TALKS Inc. SGC-X2UL

PR: FreeBSD-users-jp/80137
Submitted by: HORIO Shinsuke <shin@happynet.co.jp>

20 years agoQuirk for SEGRAND NP-900 USB MP3Player
MIHIRA Sanpei Yoshiro [Sun, 18 Jul 2004 05:39:14 +0000 (05:39 +0000)]
Quirk for SEGRAND NP-900 USB MP3Player

PR: kern/64563
Submitted by: Kunitada Kokubun <unix_grandy@yahoo.co.jp>
MFC after: 1 week

20 years agoOnly extract a physical address from a pte in pmap_extract() if the pte is
Alan Cox [Sun, 18 Jul 2004 05:09:28 +0000 (05:09 +0000)]
Only extract a physical address from a pte in pmap_extract() if the pte is
valid.

Implement the protection check required by the pmap_extract_and_hold()
specification.  (This enables the elimination of Giant from that function.)

20 years agoRemove GIANT_REQUIRED from vmapbuf().
Alan Cox [Sun, 18 Jul 2004 04:57:49 +0000 (04:57 +0000)]
Remove GIANT_REQUIRED from vmapbuf().

20 years agoRemove the GIANT_REQUIRED preceding pmap_remove() in
Alan Cox [Sun, 18 Jul 2004 04:38:11 +0000 (04:38 +0000)]
Remove the GIANT_REQUIRED preceding pmap_remove() in
vm_pageout_map_deactivate_pages().

20 years agoHook libthread_db into the build, except for arm and powerpc.
Marcel Moolenaar [Sun, 18 Jul 2004 04:36:22 +0000 (04:36 +0000)]
Hook libthread_db into the build, except for arm and powerpc.
Porting libthread_db to arm and/or powerpc is easy enough, but
we don't build gdb on those platforms yet.

20 years agoDefine _libthr_debug for use by libthread_db.
Marcel Moolenaar [Sun, 18 Jul 2004 04:23:30 +0000 (04:23 +0000)]
Define _libthr_debug for use by libthread_db.

20 years agoDon't include lock.h and pthread_md.h when we're being included by
Marcel Moolenaar [Sun, 18 Jul 2004 04:22:01 +0000 (04:22 +0000)]
Don't include lock.h and pthread_md.h when we're being included by
libthread_db. Both headers are included seperately.

20 years agoAdd rudimentary support and stubs for libthr and libc_r on alpha, amd64,
Marcel Moolenaar [Sun, 18 Jul 2004 04:17:15 +0000 (04:17 +0000)]
Add rudimentary support and stubs for libthr and libc_r on alpha, amd64,
i386, ia64 and sparc64. Add stubs for alpha, amd64, ia64 and sparc64 for
libpthread.

Restructure the source files to avoid unnecessary use of subdirectories
that also force us to use non-portable compilation flags to deal with
the uncommon compilation requirements (building archive libraries for
linkage into a shared library).

The libpthread support has been copied from the original local and
cleaned-up to make them WARNS=2 clean.
that also force us to use non-portable compilation flags to deal with
the uncommon compilation requirements (building archive libraries for
linkage into a shared library).

The libpthread support has been copied from the original local and
cleaned-up to make them WARNS=2 clean.

Tested on: amd64, i386, ia64

20 years agoRemove 'sg' argument to linux_sendto_hdrincl, which is what I think was
Robert Watson [Sun, 18 Jul 2004 04:09:40 +0000 (04:09 +0000)]
Remove 'sg' argument to linux_sendto_hdrincl, which is what I think was
intended.  This fixes the build, but might require revision.

20 years agoThe description of ng_mesg.cmd should call it "cmd" to match the struct
Giorgos Keramidas [Sun, 18 Jul 2004 02:42:38 +0000 (02:42 +0000)]
The description of ng_mesg.cmd should call it "cmd" to match the struct
definition and netgraph.h.

PR: docs/66539
Submitted by: Michel Lavondes <fox@vader.aacc.cc.md.us>

20 years agoRemove the dependency of the :C/regexp/replacement/ variable modifier
Giorgos Keramidas [Sun, 18 Jul 2004 02:26:30 +0000 (02:26 +0000)]
Remove the dependency of the :C/regexp/replacement/ variable modifier
from the :S modifier which follows a bit further below.  This way the
reader can read each of these two descriptions without having to jump
back and forth in the manpage.

PR: docs/26943
Submitted by: Alex Kapranoff <alex@kapran.bitmcnit.bryansk.su>

20 years agoPriner control devices are now lpt*.ctl.
David E. O'Brien [Sun, 18 Jul 2004 02:22:47 +0000 (02:22 +0000)]
Priner control devices are now lpt*.ctl.

20 years agoMention the vfs.usermount sysctl.
Christian Brueffer [Sun, 18 Jul 2004 01:51:59 +0000 (01:51 +0000)]
Mention the vfs.usermount sysctl.

Obtained from: NetBSD via DragonFly BSD
MFC after: 3 days

20 years agoDrop Giant and acquire the UNIX domain socket subsystem lock a bit
Robert Watson [Sun, 18 Jul 2004 01:29:43 +0000 (01:29 +0000)]
Drop Giant and acquire the UNIX domain socket subsystem lock a bit
earlier in unp_connect() so that vp->v_socket can't change between
our copying its value to a local variable and later use of that
variable.  This may have been responsible for a panic during
shutdown that I experienced where simultaneous closing of a listen
socket by rpcbind and a new connection being made to rpcbind by
mountd.

20 years agoAdd an "out" to the now required BS of defining DESTDIR.
David E. O'Brien [Sun, 18 Jul 2004 01:19:33 +0000 (01:19 +0000)]
Add an "out" to the now required BS of defining DESTDIR.

20 years agoThis is a forced commit.
David Xu [Sun, 18 Jul 2004 00:59:57 +0000 (00:59 +0000)]
This is a forced commit.
Clear suspension flag for debugged process when detaching.

20 years agom_tag_copy takes an additional "how" parameter in FreeBSD.
Max Laier [Sun, 18 Jul 2004 00:51:56 +0000 (00:51 +0000)]
m_tag_copy takes an additional "how" parameter in FreeBSD.

Submitted by: rwatson

20 years agoFix macro so that we don't get missing initializer warnings.
Alfred Perlstein [Sat, 17 Jul 2004 23:56:24 +0000 (23:56 +0000)]
Fix macro so that we don't get missing initializer warnings.