]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
19 years agoFix so all parts of lpd, lpc, lpq, and lprm will use the same algorithm
gad [Fri, 31 Dec 2004 00:36:28 +0000 (00:36 +0000)]
Fix so all parts of lpd, lpc, lpq, and lprm will use the same algorithm
for calculating the job number for a job based on the control-file name.
We might receive cf-files named by other implementations of lpr, where
the job number shown by lpq would not match the job number that other
commands expected for the same name.

This also uses a newer algorithm for determining a job number, to avoid
problems caused when a control-file is named using an IP address, instead
of the hostname.

This also moved the declaration if isowner() from lp.h to rmjob.c.  When I
went to change the parameters, I noticed that rmjob.c was the only source
file which uses it.

MFC after: 2 weeks

19 years agoUpdate netstat(1) for recent conversion of netipx to queue(9) from
rwatson [Fri, 31 Dec 2004 00:32:50 +0000 (00:32 +0000)]
Update netstat(1) for recent conversion of netipx to queue(9) from
home-brew linked lists.  Read in the ipxpcb_list structure first in
order to find the first pcb pointer.  Then follow the chain as
before, only the termination condition is a NULL next pointer
rather than a next pointer equal to the original offset.

19 years agoDo more extensive checking of the userid field which is read in from the
gad [Fri, 31 Dec 2004 00:18:42 +0000 (00:18 +0000)]
Do more extensive checking of the userid field which is read in from the
control-file for each print job.  This is partially because the previous
checks still let through some characters which would cause trouble for
other applications which try to process the resulting userid -- such as
accounting programs.

But the main reason is to handle the case where some remote host sends a
print job where the given userid is an uppercase-version of the real userid.
For that case, lpd will now check for uppercase letters in the userid.  If
there are any, it will check to see if the given userid (with the uppercase
letters) is a valid one.  If it is *not* valid, then lpd will change the
userid to all-lowercase right when the job is received.

MFC after: 2 weeks

19 years agoRemove special case from *devs2h loop for usb
imp [Thu, 30 Dec 2004 23:22:08 +0000 (23:22 +0000)]
Remove special case from *devs2h loop for usb
Move usb out of the loop since the other devs2h programs don't grok -h
Add new rule for usbdevs_data.h

19 years agoCall usbdevs2h with -h for .h file
imp [Thu, 30 Dec 2004 23:19:40 +0000 (23:19 +0000)]
Call usbdevs2h with -h for .h file
Explicitly generate usbdevs_data.h for usb kernels with new -d switch

19 years agoAllow -d and -h as command line flags. These are used to generate
imp [Thu, 30 Dec 2004 23:18:34 +0000 (23:18 +0000)]
Allow -d and -h as command line flags.  These are used to generate
only the .h or _data.h files respectively.  Restructure code to allow
for this.

19 years agoFix a typo and two whitespace nits.
jhb [Thu, 30 Dec 2004 22:17:00 +0000 (22:17 +0000)]
Fix a typo and two whitespace nits.

19 years agoRework the interface between priority propagation (lending) and the
jhb [Thu, 30 Dec 2004 20:52:44 +0000 (20:52 +0000)]
Rework the interface between priority propagation (lending) and the
schedulers a bit to ensure more correct handling of priorities and fewer
priority inversions:
- Add two functions to the sched(9) API to handle priority lending:
  sched_lend_prio() and sched_unlend_prio().  The turnstile code uses these
  functions to ask the scheduler to lend a thread a set priority and to
  tell the scheduler when it thinks it is ok for a thread to stop borrowing
  priority.  The unlend case is slightly complex in that the turnstile code
  tells the scheduler what the minimum priority of the thread needs to be
  to satisfy the requirements of any other threads blocked on locks owned
  by the thread in question.  The scheduler then decides where the thread
  can go back to normal mode (if it's normal priority is high enough to
  satisfy the pending lock requests) or it it should continue to use the
  priority specified to the sched_unlend_prio() call.  This involves adding
  a new per-thread flag TDF_BORROWING that replaces the ULE-only kse flag
  for priority elevation.
- Schedulers now refuse to lower the priority of a thread that is currently
  borrowing another therad's priority.
- If a scheduler changes the priority of a thread that is currently sitting
  on a turnstile, it will call a new function turnstile_adjust() to inform
  the turnstile code of the change.  This function resorts the thread on
  the priority list of the turnstile if needed, and if the thread ends up
  at the head of the list (due to having the highest priority) and its
  priority was raised, then it will propagate that new priority to the
  owner of the lock it is blocked on.

Some additional fixes specific to the 4BSD scheduler include:
- Common code for updating the priority of a thread when the user priority
  of its associated kse group has been consolidated in a new static
  function resetpriority_thread().  One change to this function is that
  it will now only adjust the priority of a thread if it already has a
  time sharing priority, thus preserving any boosts from a tsleep() until
  the thread returns to userland.  Also, resetpriority() no longer calls
  maybe_resched() on each thread in the group. Instead, the code calling
  resetpriority() is responsible for calling resetpriority_thread() on
  any threads that need to be updated.
- schedcpu() now uses resetpriority_thread() instead of just calling
  sched_prio() directly after it updates a kse group's user priority.
- sched_clock() now uses resetpriority_thread() rather than writing
  directly to td_priority.
- sched_nice() now updates all the priorities of the threads after the
  group priority has been adjusted.

Discussed with: bde
Reviewed by: ups, jeffr
Tested on: 4bsd, ule
Tested on: i386, alpha, sparc64

19 years agoWhitespace fix.
jhb [Thu, 30 Dec 2004 20:30:58 +0000 (20:30 +0000)]
Whitespace fix.

19 years agoStop explicitly touching td_base_pri outside of the scheduler and simply
jhb [Thu, 30 Dec 2004 20:29:58 +0000 (20:29 +0000)]
Stop explicitly touching td_base_pri outside of the scheduler and simply
set a thread's priority via sched_prio() when that is the desired action.
The schedulers will start managing td_base_pri internally shortly.

19 years agoUse NULL instead of 0 in a few places as well as various whitespace fixes.
jhb [Thu, 30 Dec 2004 19:26:23 +0000 (19:26 +0000)]
Use NULL instead of 0 in a few places as well as various whitespace fixes.

19 years agoCall tty_close() at the very end of ttyclose() since otherwise NULL
jhb [Thu, 30 Dec 2004 19:24:49 +0000 (19:24 +0000)]
Call tty_close() at the very end of ttyclose() since otherwise NULL
deferences can occur since tty_close() may end up freeing the tty structure
if it drops the last reference to it.

Glanced at by: phk

19 years agoSmall whitespace fixes.
jhb [Thu, 30 Dec 2004 19:14:48 +0000 (19:14 +0000)]
Small whitespace fixes.

19 years agoMinor sync to i386 GENERIC in the form of comments and whitespace.
jhb [Thu, 30 Dec 2004 18:51:23 +0000 (18:51 +0000)]
Minor sync to i386 GENERIC in the form of comments and whitespace.

19 years agoMinor sync to alpha and i386:
jhb [Thu, 30 Dec 2004 18:31:53 +0000 (18:31 +0000)]
Minor sync to alpha and i386:
- Add ADAPTIVE_GIANT and AHC_REG_PRETTY_PRINT.
- Indent SMP to match other options.

19 years agoUse a global variable, ipxpcb_lport_cache, to cache the most recently
rwatson [Thu, 30 Dec 2004 17:54:53 +0000 (17:54 +0000)]
Use a global variable, ipxpcb_lport_cache, to cache the most recently
used IPX port number, rather than using the global ipxpcb list head.

19 years agoConvert netipx to use queue(9) doubly-linked lists instead of home-brew
rwatson [Thu, 30 Dec 2004 17:49:40 +0000 (17:49 +0000)]
Convert netipx to use queue(9) doubly-linked lists instead of home-brew
linked lists for ipxpcb's.

19 years agoGarbage collect unused (and incompletely implemented) functions:
rwatson [Thu, 30 Dec 2004 17:21:07 +0000 (17:21 +0000)]
Garbage collect unused (and incompletely implemented) functions:

- ipx_pcbnotify(), which is never called.
- ipx_rtchange(), which is never called, is incomplete inplemented, and
  also #ifdef notdef.
- spx_fixmtu(), which is never called, is incompletely implemented, and
  also #ifdef notdef.

19 years agoConstify ipx_zeronet, ipx_zerohost, ipx_broadnet, ipx_broadhost.
rwatson [Thu, 30 Dec 2004 16:56:07 +0000 (16:56 +0000)]
Constify ipx_zeronet, ipx_zerohost, ipx_broadnet, ipx_broadhost.
Remove 'allones' since the values of the broadcast network and
host variables are set statically.

19 years agoEnable NCP build in NOTES so that it is visible in LINT.
rwatson [Thu, 30 Dec 2004 16:53:20 +0000 (16:53 +0000)]
Enable NCP build in NOTES so that it is visible in LINT.

19 years agoMake syslog(3) thread safe.
glebius [Thu, 30 Dec 2004 16:05:11 +0000 (16:05 +0000)]
Make syslog(3) thread safe.

PR: bin/72394
Submitted by: Dan Nelson
Reviewed by: deischen
MFC after: 2 weeks

19 years agoSync with i386 GENERIC some:
jhb [Thu, 30 Dec 2004 15:32:31 +0000 (15:32 +0000)]
Sync with i386 GENERIC some:
- Update comments to newer style (space after #)
- Bring across various comment updates.
- Add AHC_REG_PRETTY_PRINT, ADAPTIVE_GIANT, and rue(4).

19 years ago- Indent the comments beside the SMP options to the same level as all the
jhb [Thu, 30 Dec 2004 15:30:23 +0000 (15:30 +0000)]
- Indent the comments beside the SMP options to the same level as all the
  other comments.  Clarify that the next two things needed for SMP are
  two lines.
- Expand mii abbreviation to miibus for clarity in the USB ethernet
  comment.

19 years agoPlug memory leak.
dds [Thu, 30 Dec 2004 14:27:57 +0000 (14:27 +0000)]
Plug memory leak.

PR: bin/75656
MFC after: 2 weeks

19 years agoMake the sysctls kern.ipc.msgmnb and kern.ipc.msgtql into tunables as
rwatson [Thu, 30 Dec 2004 13:56:34 +0000 (13:56 +0000)]
Make the sysctls kern.ipc.msgmnb and kern.ipc.msgtql into tunables as
is the case for most other sysctls in the System V IPC message queue
implementation.

PR: 75541
Submitted by: Sergiy Vyshnevetskiy <serg at vostok dot net>
MFC after: 2 weeks

19 years agomac_portacl.c no longer references vnode_if.h.
rwatson [Thu, 30 Dec 2004 13:49:01 +0000 (13:49 +0000)]
mac_portacl.c no longer references vnode_if.h.

19 years agoAdd LOG_NTP facility.
krion [Thu, 30 Dec 2004 13:09:34 +0000 (13:09 +0000)]
Add LOG_NTP facility.

Submitted by: Kurt Jaeger <pi at complx dot LF dot net>
Approved by: ru
MFC after: 3 days

19 years agoRemove a check that never returns true, because in this case we have panic
glebius [Thu, 30 Dec 2004 12:22:51 +0000 (12:22 +0000)]
Remove a check that never returns true, because in this case we have panic
before the check.

19 years agoCorrect two typos; remove a duplicate.
schweikh [Thu, 30 Dec 2004 12:18:30 +0000 (12:18 +0000)]
Correct two typos; remove a duplicate.

19 years ago[1] Remove the generic bridge support from those drivers that had it. The
anholt [Thu, 30 Dec 2004 07:18:58 +0000 (07:18 +0000)]
[1] Remove the generic bridge support from those drivers that had it.  The
generic bridge support was biting us more than it helped, whenever a new chipset
came out from a vendor and misprogramming it caused strange hangs or corruption.
[2] Add a large number of PCI IDs based on what the linux drivers support.
Note that the new PCI IDs haven't been tested, they're just *likely* to work.
In particular the VIA AGP 8x chipsets are concerning, due to lack of testing,
possible issues (kern/69953), and not having a nice "does this bridge say it
would do 8x" function.  However, this shouldn't make the situation worse, since
these chips would have probed in the past anyway.

19 years agoMake umtx_wait and umtx_wake more like linux futex does, it is
davidxu [Thu, 30 Dec 2004 02:56:17 +0000 (02:56 +0000)]
Make umtx_wait and umtx_wake more like linux futex does, it is
more general than previous. It also lets me implement cancelable point
in thread library. Also in theory, umtx_lock and umtx_unlock can
be implemented by using umtx_wait and umtx_wake, all atomic operations
can be done in userland without kernel's casuptr() function.

19 years agoFix typos.
brueffer [Thu, 30 Dec 2004 01:48:12 +0000 (01:48 +0000)]
Fix typos.

PR: 75616
Submitted by: n-kogane@syd.odn.ne.jp

19 years agoClearer va-args usage.
obrien [Thu, 30 Dec 2004 00:53:56 +0000 (00:53 +0000)]
Clearer va-args usage.

Submitted by: bde

19 years agoRewrite the code that hacks a short names to use in
kientzle [Wed, 29 Dec 2004 23:26:18 +0000 (23:26 +0000)]
Rewrite the code that hacks a short names to use in
the regular ustar entry.  The old code sometimes created
a too-long name that overflowed the ustar fields and triggered
an internal assertion failure.  This version should be more
robust.

Thanks to: Michal Listos
Fixes: bin/74385
MFC after: 15 days

19 years agoRemove vnode_if.h for rev 1.14.
obrien [Wed, 29 Dec 2004 21:50:44 +0000 (21:50 +0000)]
Remove vnode_if.h for rev 1.14.

19 years agoDon't call ugen_destroy_devnodes(). it is called from ugen_set_config()
julian [Wed, 29 Dec 2004 21:36:43 +0000 (21:36 +0000)]
Don't call ugen_destroy_devnodes(). it is called from ugen_set_config()
which is the very next line.

Submitted by: stefanf
MFC after: 3 days

19 years agoCatch up with stat.h a bit:
trhodes [Wed, 29 Dec 2004 20:00:29 +0000 (20:00 +0000)]
Catch up with stat.h a bit:

o Sticky bit is actually defined as S_ISTXT not S_ISVTX.
o Sticky bit is wrapped in if __BSD_VISIBLE not _POSIX_SOURCE.

19 years agoMark up SIGSTOP with .Dv and use .Nm in HISTORY for marking up `signal'.
keramida [Wed, 29 Dec 2004 19:40:56 +0000 (19:40 +0000)]
Mark up SIGSTOP with .Dv and use .Nm in HISTORY for marking up `signal'.

PR: docs/75615
Submitted by: n-kogane@syd.odn.ne.jp

19 years agoRemove the special sparc64 time_t support.
obrien [Wed, 29 Dec 2004 19:39:06 +0000 (19:39 +0000)]
Remove the special sparc64 time_t support.
If someone isn't running with a 64-bit time_t by now, they don't
track -CURRENT and would probably go to RELENG_5 vs. 6-CURRENT.

19 years agoA lot of people may want to submit PRs from systems that are not
keramida [Wed, 29 Dec 2004 19:32:21 +0000 (19:32 +0000)]
A lot of people may want to submit PRs from systems that are not
connected to the Internet or systems that do not have a correctly
configured email subsystem.  Now that the send-pr web interface has
antispam protection, mention that it is ok to use it for submitting
problem reports.

Ok by: linimon
Submitted by: Rob <spamrefuse@yahoo.com> on freebsd-doc

19 years agoMerge the following from the English version:
hrs [Wed, 29 Dec 2004 17:11:02 +0000 (17:11 +0000)]
Merge the following from the English version:

1.17  -> 1.21  early-adopter/article.sgml
1.261 -> 1.269 hardware/common/dev.sgml
1.6   -> 1.7   hardware/common/intro.sgml
1.13  -> 1.14  hardware/i386/proc-i386.sgml
1.7   -> 1.9   hardware/ia64/proc-ia64.sgml
1.4   -> 1.5   hardware/pc98/proc-pc98.sgml
1.7   -> 1.8   share/sgml/release.dsl

19 years agoSMC SMC2536W-AG verified as working.
brueffer [Wed, 29 Dec 2004 16:57:11 +0000 (16:57 +0000)]
SMC SMC2536W-AG verified as working.

MFC after: 3 days

19 years agoAutogenerate device listings for bktr(4).
brueffer [Wed, 29 Dec 2004 14:38:37 +0000 (14:38 +0000)]
Autogenerate device listings for bktr(4).

MFC after: 1 week

19 years ago- call the driver 'bktr' instead of 'brooktree' throughout the document
brueffer [Wed, 29 Dec 2004 14:25:41 +0000 (14:25 +0000)]
- call the driver 'bktr' instead of 'brooktree' throughout the document
- add a HARDWARE section
- add a more descriptive NAME section
- update for recently imported Pinnacle PCTV support
- add Branko Lankester, the original author of the PCTV patches,
  to the AUTHORS section
- bump .Dd

Thanks to all multimedia@ readers who sent me information about
their supported hardware!

MFC after: 1 week

19 years agoCorrectly initialise the 2nd kernel segment, and don't
grehan [Wed, 29 Dec 2004 09:41:40 +0000 (09:41 +0000)]
Correctly initialise the 2nd kernel segment, and don't
forget to actually install it in the segment register.
This may fix some of the weird panics seen when kernel VM
is heavily used.

19 years agoWe don't need to generate vnode_if.h anymore
imp [Wed, 29 Dec 2004 08:50:35 +0000 (08:50 +0000)]
We don't need to generate vnode_if.h anymore

19 years agoNo need to generate vnode_if.h anymore
imp [Wed, 29 Dec 2004 08:44:30 +0000 (08:44 +0000)]
No need to generate vnode_if.h anymore

19 years agoNO_NIS cleanup: LINKS and MLINKS
ache [Wed, 29 Dec 2004 08:16:39 +0000 (08:16 +0000)]
NO_NIS cleanup: LINKS and MLINKS

19 years agoMore NO_NIS cleanup: LINKS and MLINKS
ache [Wed, 29 Dec 2004 08:13:20 +0000 (08:13 +0000)]
More NO_NIS cleanup: LINKS and MLINKS

19 years agoDocument callout_pending() macro.
glebius [Wed, 29 Dec 2004 08:01:24 +0000 (08:01 +0000)]
Document callout_pending() macro.

MFC after: 1 week

19 years agoMove WINBOND 4 port hub to its proper location in the file. It belongs
imp [Wed, 29 Dec 2004 06:21:41 +0000 (06:21 +0000)]
Move WINBOND 4 port hub to its proper location in the file.  It belongs
in a section of its own, not arbitrarily grouped in the hawking section
just because hawking sells it...

19 years agoEliminate (now) unnecessary acquisition and release of the global page
alc [Wed, 29 Dec 2004 04:49:10 +0000 (04:49 +0000)]
Eliminate (now) unnecessary acquisition and release of the global page
queues lock.

19 years agoMake an argument name in the synopsis match that in the text. Correct
jkoshy [Wed, 29 Dec 2004 04:29:06 +0000 (04:29 +0000)]
Make an argument name in the synopsis match that in the text.  Correct
an mdoc error.

19 years agoAccess to the page's busy field is (now) synchronized by the containing
alc [Wed, 29 Dec 2004 04:18:22 +0000 (04:18 +0000)]
Access to the page's busy field is (now) synchronized by the containing
object's lock.  Therefore, the assertion that the page queues lock is held
can be removed from vm_page_io_start().

19 years agoDrop hold on patch. ENOTIME for long-stalled ideas here. Too busy elsewhere.
gad [Wed, 29 Dec 2004 03:49:58 +0000 (03:49 +0000)]
Drop hold on patch.  ENOTIME for long-stalled ideas here.  Too busy elsewhere.
Add lines for interest in 'ps' and 'pkill', where I'm more active.

19 years agoVarious markup and spelling fixes.
trhodes [Wed, 29 Dec 2004 02:18:24 +0000 (02:18 +0000)]
Various markup and spelling fixes.

PR: 75574
Submitted by: Nobuyuki Koganemaru <n-kogane@syd.odn.ne.jp> (original version)

19 years agoAdd the Winbond made Hawking UH104 4-port hub.
obrien [Wed, 29 Dec 2004 02:09:19 +0000 (02:09 +0000)]
Add the Winbond made Hawking UH104 4-port hub.
Add a note where to find the offical USB assigned numbers.

19 years agoRewrite ehci_abort_xfer() to use the method hinted at in the EHCI spec.
julian [Wed, 29 Dec 2004 01:21:18 +0000 (01:21 +0000)]
Rewrite ehci_abort_xfer() to use the method hinted at in the EHCI spec.
to remove a transaction from the async schedule. The previous method didn't
work well and led to the hardware writing to free'd buffers etc, as
it didn't always know that the transaction had been aborted.

Written after consultation with David Brownell who wrote the Linux
EHCI driver.

As part of this give the sqh structure a "previous" pointer.

MFC after: 1 week

19 years agoSome grammar, spelling, and mdoc(7) fixes. No content changes.
trhodes [Wed, 29 Dec 2004 00:11:34 +0000 (00:11 +0000)]
Some grammar, spelling, and mdoc(7) fixes.  No content changes.

PR: 75581
Submitted by: Nobuyuki Koganemaru n-kogane@syd.odn.ne.jp (original version)

19 years agoAllow ':' and '%' in hostname specifications so that we can specify IPv6
dwmalone [Tue, 28 Dec 2004 22:25:04 +0000 (22:25 +0000)]
Allow ':' and '%' in hostname specifications so that we can specify IPv6
addresses and scope IDs.

MFC after: 3 weeks

19 years agoPass an explicit pointer to the bus structure in the 'usb' device ivar
jb [Tue, 28 Dec 2004 22:17:10 +0000 (22:17 +0000)]
Pass an explicit pointer to the bus structure in the 'usb' device ivar
rather than a softc pointer (with the bus structure at the start).

This is a non-functional change. It just helps when reading the code to
know that the ehci, ohci and uhci drivers share the bus structure, not the
entire softc.

19 years agoSome syslog messages delimit the program tag with whitespace rather
dwmalone [Tue, 28 Dec 2004 21:58:54 +0000 (21:58 +0000)]
Some syslog messages delimit the program tag with whitespace rather
than a colon, so don't allow whitespace in program names. To be
consistent with hostnames, don't allow whitespace in the program
name specifiers in syslog.conf either.

(The first change is by Markus from the PR, the second is mine.)

PR: 68691
Submitted by: Markus Oestreicher <m.oe@x-trader.de>
MFC after: 3 weeks

19 years agoRemove debug code.
pjd [Tue, 28 Dec 2004 21:52:45 +0000 (21:52 +0000)]
Remove debug code.

19 years ago- Up the WITNESS_COUNT macro from 200 to 1024 to support the growing number
jhb [Tue, 28 Dec 2004 21:21:27 +0000 (21:21 +0000)]
- Up the WITNESS_COUNT macro from 200 to 1024 to support the growing number
  of lock types in the kernel.  This results in an increase of witness
  data usage from ~145k to ~280k on i386 for kernels with
  'options WITNESS'.
- Remove the unused witness malloc bucket.

Submitted by: Michal Mertl mime at traveller dot cz (1)

19 years agoDetect if class name is '-h' or 'help' and if it is, show general help
pjd [Tue, 28 Dec 2004 17:07:18 +0000 (17:07 +0000)]
Detect if class name is '-h' or 'help' and if it is, show general help
message, i.e.:
geom: usage: geom <class> <command> [options]

PR: bin/71537
Submitted by: bugghy <bugghy@phenix.rootshell.be>
MFC after: 5 days

19 years agoUpdate for FILE v4.12.
ru [Tue, 28 Dec 2004 12:29:06 +0000 (12:29 +0000)]
Update for FILE v4.12.

19 years agoUpdate for FILE v4.12.
ru [Tue, 28 Dec 2004 12:17:56 +0000 (12:17 +0000)]
Update for FILE v4.12.

19 years ago- Plug a memory leak in ng_netflow_cache_init().
glebius [Tue, 28 Dec 2004 12:11:32 +0000 (12:11 +0000)]
- Plug a memory leak in ng_netflow_cache_init().
- Initialize error to 0 in ng_netflow_flow_add() (a nop change).
- Update cache statistics holding workqueue mutex.

MFC after: 3 days

19 years agoDon't probe for a disk unless explicitly enabled by a tunable.
grehan [Tue, 28 Dec 2004 05:07:49 +0000 (05:07 +0000)]
Don't probe for a disk unless explicitly enabled by a tunable.
This allows boot to proceed on a real system until the issue
of calling back into certain OpenFirmware calls (e.g. finddevice)
in thread context is understood.

(this commit only affects psim users, of which I think I am the
only one...)

19 years agoCatch up with file 4.12 import.
obrien [Tue, 28 Dec 2004 04:35:01 +0000 (04:35 +0000)]
Catch up with file 4.12 import.

19 years agoThis commit was generated by cvs2svn to compensate for changes in r139368,
obrien [Tue, 28 Dec 2004 04:31:47 +0000 (04:31 +0000)]
This commit was generated by cvs2svn to compensate for changes in r139368,
which included commits to RCS files with non-trunk default branches.

19 years agoVirgin import of Christos Zoulas's FILE 4.12.
obrien [Tue, 28 Dec 2004 04:31:47 +0000 (04:31 +0000)]
Virgin import of Christos Zoulas's FILE 4.12.

19 years ago- Add some notes on what is required to run schedgraph.py successfully.
arr [Tue, 28 Dec 2004 00:49:02 +0000 (00:49 +0000)]
- Add some notes on what is required to run schedgraph.py successfully.
  This was pulled from Jeff Roberson's email to current@ "the other day."

Approved by: jeff@

19 years agoAdd FR support to sppp (MFCronyx).
rik [Tue, 28 Dec 2004 00:07:57 +0000 (00:07 +0000)]
Add FR support to sppp (MFCronyx).

Silence on: net@, current@, hackers@.
No objections: joerg

Requested by: by many (mostly Cronyx) users for a long long time.
MFC after: 10 days

PR: kern/21771, kern/66348

19 years agoRemove warnings and make wc WARNS=6 clean.
josef [Mon, 27 Dec 2004 22:27:56 +0000 (22:27 +0000)]
Remove warnings and make wc WARNS=6 clean.
Add FreeBSD Id to Makefile.

Approved by: ssouhlal

19 years agonos->now typo reported by njl
imp [Mon, 27 Dec 2004 18:58:57 +0000 (18:58 +0000)]
nos->now typo reported by njl

19 years agoAdd note about where major numbers are now assigned
imp [Mon, 27 Dec 2004 18:21:39 +0000 (18:21 +0000)]
Add note about where major numbers are now assigned

19 years agoFormatting nits
imp [Mon, 27 Dec 2004 18:18:38 +0000 (18:18 +0000)]
Formatting nits

19 years agoRewrite piece of code which I committed some time ago that allows to
pjd [Mon, 27 Dec 2004 17:20:06 +0000 (17:20 +0000)]
Rewrite piece of code which I committed some time ago that allows to
show file name for 'mdconfig -l -u <x>' command.
This allows to preserve API/ABI compatibility with version 0 (that's why
I changed version number back to 0) and will allow to merge this change
to RELENG_5.

MFC after: 5 days

19 years agoFix mbuf leak.
pjd [Mon, 27 Dec 2004 15:53:44 +0000 (15:53 +0000)]
Fix mbuf leak.

Submitted by: Johnny Eriksson <bygg@cafax.se>
MFC after: 5 days

19 years agoAttempt to slightly refine the print out from "show alllocks" -- list
rwatson [Mon, 27 Dec 2004 10:47:08 +0000 (10:47 +0000)]
Attempt to slightly refine the print out from "show alllocks" -- list
the process and thread numbers/names on the same line rather than on
separate lines, and print the thread pointer not just the tid.

19 years agoRemove duplicate CHPRODUCTS entry.
imp [Mon, 27 Dec 2004 06:58:05 +0000 (06:58 +0000)]
Remove duplicate CHPRODUCTS entry.
Use tab religiously in the vendor section.

19 years agoMFi386: Restore cpu_reset proxy code to enable reset from ddb on an AP.
njl [Mon, 27 Dec 2004 06:42:25 +0000 (06:42 +0000)]
MFi386: Restore cpu_reset proxy code to enable reset from ddb on an AP.

19 years agoReduce diffs to i386.
njl [Mon, 27 Dec 2004 06:35:42 +0000 (06:35 +0000)]
Reduce diffs to i386.

19 years agoRestore the cpu_reset proxy code. It is needed if you want to reset the
njl [Mon, 27 Dec 2004 06:15:03 +0000 (06:15 +0000)]
Restore the cpu_reset proxy code.  It is needed if you want to reset the
system from an AP at runtime (i.e., calling cpu_reset from ddb).  Someday,
if we move to an NMI for stopping cpus instead, we can do away with this.

Requested by: jhb

19 years agoMinor cleanup: Reduce stack usage in probe method. Outdent probe. Use
njl [Mon, 27 Dec 2004 05:42:32 +0000 (05:42 +0000)]
Minor cleanup:  Reduce stack usage in probe method.  Outdent probe.  Use
TRUE/FALSE instead of 1/0 for booleans.  Remove trailing and extra whitespace.

19 years agoRemove trailing whitespace.
njl [Mon, 27 Dec 2004 05:36:47 +0000 (05:36 +0000)]
Remove trailing whitespace.

19 years agoNote that access to the page's busy count is synchronized by the containing
alc [Mon, 27 Dec 2004 05:27:59 +0000 (05:27 +0000)]
Note that access to the page's busy count is synchronized by the containing
object's lock.

19 years agoDo not vput(9) unlocked vnode and do not VREF it with the sole purpose
kan [Mon, 27 Dec 2004 05:17:11 +0000 (05:17 +0000)]
Do not vput(9) unlocked vnode and do not VREF it with the sole purpose
of vputting it back immediately.

Complained by: DEBUG_VFS_LOCKS

19 years ago - Unintentionally checked in a debugging panic. Remove that.
jeff [Sun, 26 Dec 2004 23:21:48 +0000 (23:21 +0000)]
 - Unintentionally checked in a debugging panic.  Remove that.

19 years ago - Remove a 4BSD specific hack since this will work on ULE too.
jeff [Sun, 26 Dec 2004 22:56:51 +0000 (22:56 +0000)]
 - Remove a 4BSD specific hack since this will work on ULE too.

19 years ago - Fix a long standing problem where an ithread would not honor sched_pin().
jeff [Sun, 26 Dec 2004 22:56:08 +0000 (22:56 +0000)]
 - Fix a long standing problem where an ithread would not honor sched_pin().
 - Remove the sched_add wrapper that used sched_add_internal() as a backend.
   Its only purpose was to interpret one flag and turn it into an int.  Do
   the right thing and interpret the flag in sched_add() instead.
 - Pass the flag argument to sched_add() to kseq_runq_add() so that we can
   get the SRQ_PREEMPT optimization too.
 - Add a KEF_INTERNAL flag.  If KEF_INTERNAL is set we don't adjust the SLOT
   counts, otherwise the slot counts are adjusted as soon as we enter
   sched_add() or sched_rem() rather than when the thread is actually placed
   on the run queue.  This greatly simplifies the handling of slots.
 - Remove the explicit prevention of migration for ithreads on non-x86
   platforms.  This was never shown to have any real benefit.
 - Remove the unused class argument to KSE_CAN_MIGRATE().
 - Add ktr points for thread migration events.
 - Fix a long standing bug on platforms which don't initialize the cpu
   topology.  The ksg_maxid variable was never correctly set on these
   platforms which caused the long term load balancer to never inspect
   more than the first group or processor.
 - Fix another bug which prevented the long term load balancer from working
   properly.  If stathz != hz we can't expect sched_clock() to be called
   on the exact tick count that we're anticipating.
 - Rearrange sched_switch() a bit to reduce indentation levels.

19 years agoAdd "show alllocks" command to DDB, which dumps a list of processes
rwatson [Sun, 26 Dec 2004 22:52:24 +0000 (22:52 +0000)]
Add "show alllocks" command to DDB, which dumps a list of processes
and threads currently holding sleep mutexes (and spin mutexes for
curthread).  This can be quite useful in looking for a lock condition
summary for a system, as it avoids manually iterating through threads
and processes to find all the interesting locks.

NB: "alllocks" is up there with "lockedvnods" for a bad argument for
show.

MFC after: 2 weeks

19 years agoAssert that the vm object is locked on entry to vm_page_sleep_if_busy();
alc [Sun, 26 Dec 2004 21:46:44 +0000 (21:46 +0000)]
Assert that the vm object is locked on entry to vm_page_sleep_if_busy();
remove some unneeded code.

19 years agotsleep => msleep
rik [Sun, 26 Dec 2004 20:47:41 +0000 (20:47 +0000)]
tsleep => msleep
read sys/kern/kern_synch.c, msleep ():

KASSERT(timo != 0 || mtx_owned(&Giant) || mtx != NULL,
    ("sleeping without a mutex"));

MFC after: 3 days

19 years agoFix typo.
josef [Sun, 26 Dec 2004 18:06:05 +0000 (18:06 +0000)]
Fix typo.

Noticed by: clement

19 years agoLCA is 21066 and 21068. Add EV7 (bloody optimist.. :)
wilko [Sun, 26 Dec 2004 13:23:01 +0000 (13:23 +0000)]
LCA is 21066 and 21068. Add EV7 (bloody optimist.. :)

19 years ago* Remove some code that's in a #ifndef FreeBSD that's no longer used.
darrenr [Sun, 26 Dec 2004 09:09:29 +0000 (09:09 +0000)]
* Remove some code that's in a #ifndef FreeBSD that's no longer used.

19 years ago* The #ifdef's to cause mutex's for freebsd to be declared were meant to pick
darrenr [Sun, 26 Dec 2004 09:08:42 +0000 (09:08 +0000)]
* The #ifdef's to cause mutex's for freebsd to be declared were meant to pick
  up on USE_MUTEX being defined, but this patch
* Remove some code that's in a #ifndef FreeBSD that's no longer used.

19 years agoMore minor diff reduction against netbsd:
imp [Sun, 26 Dec 2004 07:48:46 +0000 (07:48 +0000)]
More minor diff reduction against netbsd:

ADVANCELOGIC->AVANCELOGIC (nothing in the tree uses it, so safe to do)
sort HAGIWARA vendor entry
sort ACTIONTAR vendor entry
Minor change to SYSTEMTALKS vendor entry.