]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
20 years ago- put the list of supported adapters into a HARDWARE section
brueffer [Wed, 25 Aug 2004 22:39:05 +0000 (22:39 +0000)]
- put the list of supported adapters into a HARDWARE section
- make the Buffalo entry look better
- properly capitalize Ethernet and Fast Ethernet
- bump document date

MFC after: 3 days

20 years agoFix grammar nit.
simon [Wed, 25 Aug 2004 22:36:47 +0000 (22:36 +0000)]
Fix grammar nit.

Submitted by: brueffer
MFC after: 2 days

20 years agoDon't call uart_bus_probe() for non-matching PnP-devices. Trying to probe
marius [Wed, 25 Aug 2004 22:15:33 +0000 (22:15 +0000)]
Don't call uart_bus_probe() for non-matching PnP-devices. Trying to probe
the keyboard controller with uart_bus_probe() caused a hang here on an i386
machine.

Approved by: marcel

20 years agoFix "make world DESTDIR=/mnt" to work again. A recent change
ru [Wed, 25 Aug 2004 22:06:29 +0000 (22:06 +0000)]
Fix "make world DESTDIR=/mnt" to work again.  A recent change
to make(1) that causes command-line variables to be passed as
command-line variables to sub-processes that make(1) executes
broke it.  By changing the type of all DESTDIR variables used
internally in Makefile.inc1, from environment to command-line
variables of the highest priority, I was able to "make world"
with success, with the command-line variable DESTDIR set.

20 years agoDon't hold the UNIX domain socket subsystem lock over the body of the
rwatson [Wed, 25 Aug 2004 21:24:36 +0000 (21:24 +0000)]
Don't hold the UNIX domain socket subsystem lock over the body of the
UNIX domain socket garbage collection implementation, as that risks
holding the mutex over potentially sleeping operations (as well as
introducing some nasty lock order issues, etc).  unp_gc() will hold
the lock long enough to do necessary deferal checks and set that it's
running, but then release it until it needs to reset the gc state.

RELENG_5 candidate.

Discussed with: alfred

20 years agoRemove code to initialize the lid state at boot. It interfered with lid
njl [Wed, 25 Aug 2004 20:09:15 +0000 (20:09 +0000)]
Remove code to initialize the lid state at boot.  It interfered with lid
operation for some users with pure GPE lid switches (vs. embedded
controller.)

Tested by: Anish Mistry <mistry.7_at_osu.edu>
MFC after: 3 days

20 years agoRecommend using make kernel for current -> current upgrades. No reason
imp [Wed, 25 Aug 2004 19:39:13 +0000 (19:39 +0000)]
Recommend using make kernel for current -> current upgrades.  No reason
to have multiple commands for that.
Use relative paths into the src tree consistantly in the instructions.
Fix minor nits that have crept into things.

# is preseedrandom still necessary?

20 years agoLog verification errors at level 1.
pjd [Wed, 25 Aug 2004 19:18:07 +0000 (19:18 +0000)]
Log verification errors at level 1.

20 years agoDocument supported devices here (for lack of a better place). Users
mlaier [Wed, 25 Aug 2004 18:49:32 +0000 (18:49 +0000)]
Document supported devices here (for lack of a better place). Users
interested in ALTQ are likely to type in "man altq" at some point and maybe
they will scroll down to the bottom as well.

MFC after: 3 days

20 years agoFix a bug in in_cksum_hdr w/o -O.
obrien [Wed, 25 Aug 2004 18:28:15 +0000 (18:28 +0000)]
Fix a bug in in_cksum_hdr w/o -O.

The C code assumes that the carry bit is always kept from the previous
operation. However, the pointer indexing requires another add operation.
Thus, the carry bit from the first operation is tromped over by the
"addl" operation that ends up following it, so the "adcl" that follows
that has no effect because the carry bit is cleared before it.
The result is checksum failure on received packets.

The larger issue is that there isn't any other way of preventing the compiler
inserting arbitrary instructions between different __asm statements (and
that the commit message in revision 1.13 of in_cksum.h is wrong on
this point).  From
http://developer.apple.com/documentation/DeveloperTools/gcc-3.3/gcc/Extended-Asm.html
---8<---8<---8<---
You can't expect a sequence of volatile asm instructions to remain
perfectly consecutive. If you want consecutive output, use a single
asm.  Also, GCC will perform some optimizations across a volatile
asm instruction; GCC does not "forget everything" when it encounters
a volatile asm instruction the way some other compilers do.
---8<---8<---8<---

Also, this change also makes the ASM code much easier to read.

PR: 69257
Submitted by: Mike Bristow <mike@urgle.com>, Qing Li <qing.li@bluecoat.com>

20 years agoCorrect style nit in rev 1.17.
obrien [Wed, 25 Aug 2004 17:54:19 +0000 (17:54 +0000)]
Correct style nit in rev 1.17.

20 years agoConvert a mtx_lock(&Giant) to a mtx_unlock(&Giant) in nfsrv_link() to
rwatson [Wed, 25 Aug 2004 16:52:59 +0000 (16:52 +0000)]
Convert a mtx_lock(&Giant) to a mtx_unlock(&Giant) in nfsrv_link() to
prevent leakage of Giant.  With INVARIANTS, this results in an
assertion failure following execution of the RPC.  Without INVARIANTS,
it could result in problems if the NFS server is killed causing nfsd
to return to user space holding Giant.

Feet provided by: brueffer

20 years agoDo not display bogus entries for sockets in the TIME_WAIT or similar
roam [Wed, 25 Aug 2004 16:36:17 +0000 (16:36 +0000)]
Do not display bogus entries for sockets in the TIME_WAIT or similar
states that no longer have a corresponding file descriptor - until now,
sockstat would mostly randomly match null kern.file.*.xf_data fields
with the first mostly-closed socket.

This bugfix is a RELENG_5 candidate.

Approved by: andre

20 years agoAdd special case for the german whois nameserver. Without the
mbr [Wed, 25 Aug 2004 15:34:44 +0000 (15:34 +0000)]
Add special case for the german whois nameserver. Without the
'-T dn,ace -C US-ASCII' option one does only get:

$ whois nic.de

domain:    nic.de
status:    connect

More information available on:

http://www.denic.de/en/domains/technik/denic_whois-server/index.html

MFC: 3 days

20 years agoAdd -j flag to usage() and the man page synopsis.
maxim [Wed, 25 Aug 2004 13:15:07 +0000 (13:15 +0000)]
Add -j flag to usage() and the man page synopsis.

Inspired by: DragonFlyBSD

20 years agoDump disk number.
pjd [Wed, 25 Aug 2004 12:14:44 +0000 (12:14 +0000)]
Dump disk number.

20 years agoMove a paragraph from the HARDWARE section back into the
brueffer [Wed, 25 Aug 2004 11:57:06 +0000 (11:57 +0000)]
Move a paragraph from the HARDWARE section back into the
DESCRIPTION section.  Also adapt it to make it fit in better.

Requested by: simon

20 years agoAllocate memory when dumping pipes with M_WAITOK flag.
pjd [Wed, 25 Aug 2004 09:31:30 +0000 (09:31 +0000)]
Allocate memory when dumping pipes with M_WAITOK flag.
On a system with huge number of pipes, M_NOWAIT failes almost always,
because of memory fragmentation.
My fix is different than the patch proposed by Pawel Malachowski,
because in FreeBSD 5.x we cannot sleep while holding dummynet mutex
(in 4.x there is no such lock).
My fix is also ugly, but there is no easy way to prepare nice and clean fix.

PR: kern/46557
Submitted by: Eugene Grosbein <eugen@grosbein.pp.ru>
Reviewed by: mlaier

20 years agoGet a step closer to profiling the kernel by fixing the definitions
marcel [Wed, 25 Aug 2004 08:03:48 +0000 (08:03 +0000)]
Get a step closer to profiling the kernel by fixing the definitions
of the MCOUNT_ENTER, MCOUNT_EXIT and MCOUNT_DECL defines. Also make
sure there's a prototype of _MCOUNT_DECL(). This allows us to build
a kernel. There are still unresolved symbols, so linking fails.

20 years agoHave the linker provide btext. It's used for profiling.
marcel [Wed, 25 Aug 2004 07:43:28 +0000 (07:43 +0000)]
Have the linker provide btext. It's used for profiling.

20 years agoMake profiling actually work. The gcc compiler emits a call to the
marcel [Wed, 25 Aug 2004 07:42:34 +0000 (07:42 +0000)]
Make profiling actually work. The gcc compiler emits a call to the
_mcount() stub when profiling is enabled. Emit this code sequence
for assembly routines as welli (MCOUNT definition in <machine/asm.h>.
We do not pass the GOT entry however as the 4th argument, because it's
not used. The _mcount() stub calls __mcount(), which does the actual
work. Define _MCOUNT_DECL to define __mcount. We do not have an
implementation of mcount(), so we define MCOUNT as empty, but have a
weak alias to _mcount() in _mcount.S.
Note that the _mcount() stub in the kernel is slightly different from
the stub in userland. This is because we do not have to worry about
nested routines in the kernel.

20 years agoNo longer do special handling of Perl. FreeBSD users have all of 5.x
obrien [Wed, 25 Aug 2004 03:38:05 +0000 (03:38 +0000)]
No longer do special handling of Perl.  FreeBSD users have all of 5.x
to get used to the fact that Perl is no longer part of the base system.
It is practically impossible to install any useful package and not get
Perl automatically pulled in as a dependency.  So the typical user will
get their Perl.

This change greatly reduces the amount of manual labor in building the
miniinst.iso in release building.

20 years agoif_dc includes locking, but that locking is disabled by a #ifdef
rwatson [Wed, 25 Aug 2004 03:37:25 +0000 (03:37 +0000)]
if_dc includes locking, but that locking is disabled by a #ifdef
by default.  As such, mark if_dc as IFF_NEEDSGIANT until such
time as appropriate locking review and testing can take place,
and the locking can be enabled by default.

RELENG_5 candidate.

20 years agoBack out v1.57, the underlying problem that caused a delay between
kensmith [Wed, 25 Aug 2004 01:39:52 +0000 (01:39 +0000)]
Back out v1.57, the underlying problem that caused a delay between
mdconfig and the device name appearing in /dev seems to be fixed.

20 years agoIn nfs_timer(), pass curthread rather than &thread0 into the protocol
rwatson [Wed, 25 Aug 2004 01:23:38 +0000 (01:23 +0000)]
In nfs_timer(), pass curthread rather than &thread0 into the protocol
send routine.  In IPv6 UDP, the thread will be passed to suser(), which
asserts that if a thread is used for a super user check, it be
curthread.  Many of these protocol entry points probably need to
accept credentials instead of threads.

MT5 candidate.

Noticed/tested by: kuriyama

20 years agoBack out previous commit. Colin's fix to md makes the extra sleep
kensmith [Wed, 25 Aug 2004 00:48:35 +0000 (00:48 +0000)]
Back out previous commit.  Colin's fix to md makes the extra sleep
here unnecessary.  Thanks Colin. :-)

20 years agoAdd a CVSTAG makefile variable that can be set in /etc/make.conf that
paul [Tue, 24 Aug 2004 23:12:16 +0000 (23:12 +0000)]
Add a CVSTAG makefile variable that can be set in /etc/make.conf that
determines which CVS tag to track when running make update. This makes
it easier to configure a box to track a particular release if it does
automated updates from a cvs repository.

20 years agoClarify what the lockfunc is used for. When it will be called and
imp [Tue, 24 Aug 2004 23:09:57 +0000 (23:09 +0000)]
Clarify what the lockfunc is used for.  When it will be called and
when it won't be called.  The old wording was correct, but not
sufficiently specific to understand when and how it would be called.
The new wording describes the current implementation's usage (which
should be updated if other appropriate times are decided upon),
specifically that it is called only when the load operation is
deferred to keep the locking state consistent.  When the operation
isn't deferred, the calling routine is assumed to have a coherent
locking world.

Reviewed by: scottl

20 years agonote that it is the caller's responsibility to free any buffer passed
alfred [Tue, 24 Aug 2004 21:48:21 +0000 (21:48 +0000)]
note that it is the caller's responsibility to free any buffer passed
to setvbuf(3) and friends.

20 years agoArgh. Forced commit to note that this was
josef [Tue, 24 Aug 2004 21:41:00 +0000 (21:41 +0000)]
Argh. Forced commit to note that this was

Approved by: wes

20 years agoAutogenerate device listings for lge(4) and vx(4)
brueffer [Tue, 24 Aug 2004 21:38:13 +0000 (21:38 +0000)]
Autogenerate device listings for lge(4) and vx(4)

20 years agoAdd snapshots.se.freebsd.org to list of available ftp mirrors.
josef [Tue, 24 Aug 2004 21:37:46 +0000 (21:37 +0000)]
Add snapshots.se.freebsd.org to list of available ftp mirrors.
Dont forget to use X-LATEST as Release Name (where X is version major number)
when you choose a snapshot server.

20 years agoAdd lge(4) and vx(4)
brueffer [Tue, 24 Aug 2004 21:35:25 +0000 (21:35 +0000)]
Add lge(4) and vx(4)

20 years ago- move the list of supported devices into a HARDWARE section
brueffer [Tue, 24 Aug 2004 21:33:14 +0000 (21:33 +0000)]
- move the list of supported devices into a HARDWARE section
- bump document date

20 years ago- move device listing and hardware notes into a HARDWARE section
brueffer [Tue, 24 Aug 2004 21:32:03 +0000 (21:32 +0000)]
- move device listing and hardware notes into a HARDWARE section
- properly capitalize Gigabit Ethernet
- bump document date

20 years agoCorrect the arguments to kern_sigaltstack() as they were reversed.
jhb [Tue, 24 Aug 2004 20:52:52 +0000 (20:52 +0000)]
Correct the arguments to kern_sigaltstack() as they were reversed.

PR: kern/68079
Submitted by: Georg-W. Koltermann gwk at rahn-koltermann dot de

20 years agoRegenerate after fcntl() wrappers were marked MP safe.
jhb [Tue, 24 Aug 2004 20:24:34 +0000 (20:24 +0000)]
Regenerate after fcntl() wrappers were marked MP safe.

20 years agoFix the ABI wrappers to use kern_fcntl() rather than calling fcntl()
jhb [Tue, 24 Aug 2004 20:21:21 +0000 (20:21 +0000)]
Fix the ABI wrappers to use kern_fcntl() rather than calling fcntl()
directly.  This removes a few more users of the stackgap and also marks
the syscalls using these wrappers MP safe where appropriate.

Tested on: i386 with linux acroread5
Compiled on: i386, alpha LINT

20 years agoSii_reset needs to wait up to 1 second to get slow disks with it so
sos [Tue, 24 Aug 2004 20:11:26 +0000 (20:11 +0000)]
Sii_reset needs to wait up to 1 second to get slow disks with it so
use tsleep instead of DELAY.

20 years agoasr(4) is currently i386 only.
simon [Tue, 24 Aug 2004 19:24:47 +0000 (19:24 +0000)]
asr(4) is currently i386 only.

Confirmed by: scottl
MFC after: 2 days

20 years agoCatch up with i386 nexus.c rev 1.59: add bus_get_resource_list().
njl [Tue, 24 Aug 2004 19:22:54 +0000 (19:22 +0000)]
Catch up with i386 nexus.c rev 1.59: add bus_get_resource_list().

20 years agoshare/examples/worm is dead.
ru [Tue, 24 Aug 2004 19:03:55 +0000 (19:03 +0000)]
share/examples/worm is dead.

20 years agosr(4) definately won't work on IA64.
obrien [Tue, 24 Aug 2004 18:31:27 +0000 (18:31 +0000)]
sr(4) definately won't work on IA64.

20 years agoBe sure to always unlock the sx lock when exiting the sysctl function.
njl [Tue, 24 Aug 2004 17:53:25 +0000 (17:53 +0000)]
Be sure to always unlock the sx lock when exiting the sysctl function.

MFC after: 3 days

20 years agoAutogenerate device listings for sk(4)
brueffer [Tue, 24 Aug 2004 17:41:47 +0000 (17:41 +0000)]
Autogenerate device listings for sk(4)

20 years agoAdd sk(4)
brueffer [Tue, 24 Aug 2004 17:40:35 +0000 (17:40 +0000)]
Add sk(4)

20 years ago- move the list of supported adapters to a HARDWARE section
brueffer [Tue, 24 Aug 2004 17:38:53 +0000 (17:38 +0000)]
- move the list of supported adapters to a HARDWARE section
- correct entry for the Belkin card (we don't support 10GbE in sk(4)) :-)
- properly capitalize Gigabit Ethernet
- wordsmithing on a sentence
- bump document date

20 years agoUpdate for latest make(1) changes.
obrien [Tue, 24 Aug 2004 17:33:42 +0000 (17:33 +0000)]
Update for latest make(1) changes.

Submitted by: Steve Kargl <sgk@troutmask.apl.washington.edu>

20 years agoAdd the SMC EZ Card 10/100 as supported hardware
brueffer [Tue, 24 Aug 2004 16:06:57 +0000 (16:06 +0000)]
Add the SMC EZ Card 10/100 as supported hardware

20 years agoCorrect the link to the AIC-6915 programmer's manual
brueffer [Tue, 24 Aug 2004 15:56:47 +0000 (15:56 +0000)]
Correct the link to the AIC-6915 programmer's manual

20 years agoAutogenerate device listings for sf(4)
brueffer [Tue, 24 Aug 2004 15:40:57 +0000 (15:40 +0000)]
Autogenerate device listings for sf(4)

20 years agoAdd sf(4)
brueffer [Tue, 24 Aug 2004 15:39:07 +0000 (15:39 +0000)]
Add sf(4)

20 years ago- move list of supported adapters into a HARDWARE section
brueffer [Tue, 24 Aug 2004 15:37:36 +0000 (15:37 +0000)]
- move list of supported adapters into a HARDWARE section
- correctly capitalize Fast Ethernet
- correct two typos
- bump document date

20 years agoAlways pick up giant before returning from an ioctl call.
sos [Tue, 24 Aug 2004 15:09:05 +0000 (15:09 +0000)]
Always pick up giant before returning from an ioctl call.

20 years agoRevert previous revision, 1.7, as removal of GIANT_REQUIRED was made
rwatson [Tue, 24 Aug 2004 14:17:58 +0000 (14:17 +0000)]
Revert previous revision, 1.7, as removal of GIANT_REQUIRED was made
in the wrong branch (and hence to the wrong function).

20 years agoMT4 if_fwsubr.c:1.6:
rwatson [Tue, 24 Aug 2004 14:16:08 +0000 (14:16 +0000)]
MT4 if_fwsubr.c:1.6:

  date: 2004/08/22 14:48:55;  author: rwatson;  state: Exp;  lines: +0 -2
  Don't need to assert Giant in fw_output(), only in the firewire start
  routine.

Approved by: re (scottl)

20 years agoReplace the current implementations of ftw() and nftw() with the OpenBSD
tjr [Tue, 24 Aug 2004 13:00:55 +0000 (13:00 +0000)]
Replace the current implementations of ftw() and nftw() with the OpenBSD
implementations written by Todd C. Miller. These are cleaner, less buggy
and actively maintained.

20 years agoFix make search entries.
blackend [Tue, 24 Aug 2004 11:30:28 +0000 (11:30 +0000)]
Fix make search entries.

PR: docs/70899
Submitted by: Marc Silver <marcs@draenor.org>

20 years agoProperly check malloc returns.
sos [Tue, 24 Aug 2004 10:39:00 +0000 (10:39 +0000)]
Properly check malloc returns.

20 years agoFix a typo (attacked -> attached).
roam [Tue, 24 Aug 2004 08:47:15 +0000 (08:47 +0000)]
Fix a typo (attacked -> attached).

Approved by: sam

20 years agoConditional acquisition of socket buffer mutexes when testing socket
rwatson [Tue, 24 Aug 2004 05:28:18 +0000 (05:28 +0000)]
Conditional acquisition of socket buffer mutexes when testing socket
buffers with kqueue filters is no longer required: the kqueue framework
will guarantee that the mutex is held on entering the filter, either
due to a call from the socket code already holding the mutex, or by
explicitly acquiring it.  This removes the last of the conditional
socket locking.

20 years agoSet the description to NULL in the right detach routine. This should
imp [Tue, 24 Aug 2004 05:19:15 +0000 (05:19 +0000)]
Set the description to NULL in the right detach routine.  This should
keep dangling pointers to strings in loaded modules from hanging
around after the drivers are unloaded.

20 years agoMake sure to properly initialize 'size' to sizeof(sin) before passing
rwatson [Tue, 24 Aug 2004 04:59:26 +0000 (04:59 +0000)]
Make sure to properly initialize 'size' to sizeof(sin) before passing
it into accept().  Depending on the initial value in memory, it is
otherwise possible to get EINVAL.

20 years agoFixed a misspelling of the hook name.
ru [Tue, 24 Aug 2004 04:05:52 +0000 (04:05 +0000)]
Fixed a misspelling of the hook name.

Submitted by: Pawel Malachowski

20 years agoAdd a basic kqueue + UNIX domain socket pair regression test to do some
rwatson [Tue, 24 Aug 2004 04:02:41 +0000 (04:02 +0000)]
Add a basic kqueue + UNIX domain socket pair regression test to do some
elementary exercising of kqueues on datagram and stream sockets.  Note
that the datagram write kqueue case is left untested due to potentially
confusing behavior for the developer (me) that might require attention.

20 years agoPick up changes in rev 1.8 of src/sys/dev/ic/mpt_netbsd.c from NetBSD.
dwhite [Tue, 24 Aug 2004 03:47:41 +0000 (03:47 +0000)]
Pick up changes in rev 1.8 of src/sys/dev/ic/mpt_netbsd.c from NetBSD.
Set the DMA SGL length correctly if the DMA request must be chained because
it is too large to fit in one SGL.

This should fix this driver for some Dell Precision systems.
RELENG_5 candidate.

PR: kern/66479
Submitted by: HITOSHI Osada <qfh02545@nifty.com>

20 years agoIt is now an error to call pmap_unuse_pt without the paddr of the pde
peter [Tue, 24 Aug 2004 00:17:52 +0000 (00:17 +0000)]
It is now an error to call pmap_unuse_pt without the paddr of the pde
that contained the pte.

20 years agoOops, I forgot to have the idle loop call mp_grab_cpu_hlt() on the amd64
peter [Tue, 24 Aug 2004 00:16:43 +0000 (00:16 +0000)]
Oops, I forgot to have the idle loop call mp_grab_cpu_hlt() on the amd64
SMP case.

20 years agostruct tm.tm_year is listed as 'years since 1900', and is signed. On
peter [Tue, 24 Aug 2004 00:15:37 +0000 (00:15 +0000)]
struct tm.tm_year is listed as 'years since 1900', and is signed.  On
64 bit systems, years roughly -2^31 through 2^31 can be represented in
time_t without any trouble.  32 bit time_t systems only range from
roughly 1902 through 2038.  As a consequence, none of the date munging
code for all the various calendar tweaks before then is present.  There
are other problems including the fact that there was no 'year zero' and
so on.  So rather than get excited about trying to figure out when the
calendar jumped by two weeks etc, simply disallow negative (ie: prior to
1900) years.

This happens to have an important side effect.  If you bzero a 'struct
tm', it corresponds to 'Jan 0, 1900, 00:00 GMT'.  This happens to be
representable (after canonification) in 64 bit time_t space.  Zero tm
structs are generally an error and mktime normally returns -1 for them.
Interestingly, it tries to canonify the 'jan 0' to 'dec 31, 1899', ie:
year -1.  This conveniently trips the negative year test above, which
means we can trivially detect the null 'tm' struct.

This actually tripped up code at work. :-/  (Don't ask)

20 years agoEven in a 80 column, fixed point font, there's plenty of room for all
imp [Mon, 23 Aug 2004 23:28:02 +0000 (23:28 +0000)]
Even in a 80 column, fixed point font, there's plenty of room for all
the arguments to bus_dmamap_load, so don't use '...' but list the
actual args.  '...' usually means a variable number of args (cf
printf(3)), but bus_dmamap_load takes a fixed number of arguments.

20 years agoIn the SYNOPSIS section, move the bus_dmamem_alloc function prototype
imp [Mon, 23 Aug 2004 23:17:31 +0000 (23:17 +0000)]
In the SYNOPSIS section, move the bus_dmamem_alloc function prototype
to just before bus_dmamem_free, which is (a) more logical; (b) likely
what was originally intended and (c) matches the order in the NAME and
FUNCTIONS sections.

20 years agoRemove checking of single exit flag in thread_user_enter(), this is
davidxu [Mon, 23 Aug 2004 22:54:37 +0000 (22:54 +0000)]
Remove checking of single exit flag in thread_user_enter(), this is
generic code for threaded process, should not be here.

20 years agoCommit Doug White and Alan Cox's fix for the cross-ipi smp deadlock.
peter [Mon, 23 Aug 2004 21:39:29 +0000 (21:39 +0000)]
Commit Doug White and Alan Cox's fix for the cross-ipi smp deadlock.
We were obtaining different spin mutexes (which disable interrupts after
aquisition) and spin waiting for delivery.  For example, KSE processes
do LDT operations which use smp_rendezvous, while other parts of the
system are doing things like tlb shootdowns with a different mutex.

This patch uses the common smp_rendezvous mutex for all MD home-grown
IPIs that spinwait for delivery.  Having the single mutex means that
the spinloop to aquire it will enable interrupts periodically, thus
avoiding the cross-ipi deadlock.

Obtained from: dwhite, alc
Reviewed by:   jhb

20 years agoAllow to set kern.geom.mirror.timeout from /boot/loader.conf.
pjd [Mon, 23 Aug 2004 20:42:34 +0000 (20:42 +0000)]
Allow to set kern.geom.mirror.timeout from /boot/loader.conf.

20 years agoFix 'show' command for pipes and queues.
pjd [Mon, 23 Aug 2004 19:20:27 +0000 (19:20 +0000)]
Fix 'show' command for pipes and queues.

PR: bin/70311
Submitted by: Pawel Malachowski <pawmal-posting@freebsd.lublin.pl>
MFC after: 3 days

20 years agoUntil I can get a clearer architecture from PHK about why he wants
mjacob [Mon, 23 Aug 2004 19:04:19 +0000 (19:04 +0000)]
Until I can get a clearer architecture from PHK about why he wants
the geometry code to grab a mutex that prohibits any driver on the
stack below it from sleeping, it's not safe to allow anything in
the top half of isp to sleep (excepting the thread that Fibre Channel
instances use to re-scan loops/fabrics).

20 years agoAdd a blanket note about 5.x being the same as 6.0 and vice versa for
imp [Mon, 23 Aug 2004 18:51:36 +0000 (18:51 +0000)]
Add a blanket note about 5.x being the same as 6.0 and vice versa for
the time being.  Also add a note that says we are going to remove the
band-aides for 4.early -> 6.0 after 5.3-RELEASE so people get used to
the idea, even though it has been planned since before 5.0 was
released.

20 years agoTwo items:
imp [Mon, 23 Aug 2004 18:47:13 +0000 (18:47 +0000)]
Two items:
o we're 6.x now, so say so in the first few lines of the file.
o note that I'll be trimming this file around 5.3 release time.

20 years agoCompare the addresses of two RAID5 work packets directly instead
le [Mon, 23 Aug 2004 17:50:18 +0000 (17:50 +0000)]
Compare the addresses of two RAID5 work packets directly instead
of the addresses of their related bios when locking one out, since
they could share a bio and this could lead to parity corruption.

20 years agoWordsmithing in the hardware section.
brueffer [Mon, 23 Aug 2004 16:48:22 +0000 (16:48 +0000)]
Wordsmithing in the hardware section.

20 years agoAutogenerate entries for ixgb(4)
brueffer [Mon, 23 Aug 2004 16:45:11 +0000 (16:45 +0000)]
Autogenerate entries for ixgb(4)

20 years agoAdd ixgb, remove isp and mpt (work on all supported platforms)
brueffer [Mon, 23 Aug 2004 16:38:39 +0000 (16:38 +0000)]
Add ixgb, remove isp and mpt (work on all supported platforms)

20 years agoRework sysresource management. Instead of having each sysresource object
njl [Mon, 23 Aug 2004 16:28:42 +0000 (16:28 +0000)]
Rework sysresource management.  Instead of having each sysresource object
hold its own values, pass them up to the parent (acpi0) and merge/uniq them
on the way.  After the namespace evaluation, acpi will reserve these
resources and manage them via rman before bus_generic_probe() and
bus_generic_attach().  This is necessary because some systems specify
conflicting resources in separate sysresource objects.  It's also cleaner
in that the interface between sysresource and acpi is now merely the parent's
resource list.  This code handles the following cases:

1. Unique resource:  add it to the parent via bus_set_resource().
2. New wholly contained in old:  discard new.
3. New tail overlaps old head:  grow old head downward.
   AND/OR
4. New head overlaps old tail:  grow old tail upward.

Tested by: Pawel Worach <sajd_at_telia.com>
Tested by: Radek Kozlowski <radek_at_raadradd.com>
MFC after: 5 days

20 years agoAdd a BUS_GET_RESOURCE_LIST method for nexus.
njl [Mon, 23 Aug 2004 16:26:16 +0000 (16:26 +0000)]
Add a BUS_GET_RESOURCE_LIST method for nexus.

MFC after: 3 days

20 years agoForced commit to document:
obrien [Mon, 23 Aug 2004 16:25:07 +0000 (16:25 +0000)]
Forced commit to document:

Doug Rabson <dfr@nlsystems.com>
Message-Id: <200408220940.18504.dfr@nlsystems.com>
Size does matter for the alpha loader.  The firmware gives it 256k
of address space which we overflowed many years ago.  I extended it
in sys/boot/alpha/common/main.c:extend_heap() by adding 512k to the
loader's mapped address space.

20 years agoMy recent measurement shows that CPU_DISABLE_CMPXCHG is no longer necessary
sobomax [Mon, 23 Aug 2004 15:55:03 +0000 (15:55 +0000)]
My recent measurement shows that CPU_DISABLE_CMPXCHG is no longer necessary
with VmWare 4.x. At least with VmWare version 4.5.2, i386 version of
atomic_cmpset_int() is about 30 times slower than non-i386 version. It
makes this delta a good 5.3 MFC candidate, since otherwise it will
mislead users who run FreeBSD under modern VmWare otherwise.

20 years agoMake sure we allocate at least enough space for the TCB for variant I TLS.
dfr [Mon, 23 Aug 2004 15:30:05 +0000 (15:30 +0000)]
Make sure we allocate at least enough space for the TCB for variant I TLS.

MFC after: 3 days

20 years agoSo here I am, also born in USSR.
danfe [Mon, 23 Aug 2004 12:53:46 +0000 (12:53 +0000)]
So here I am, also born in USSR.

Approved by: fjoe (mentor)

20 years agoDon't try to translate the control message unless we're certain it's
des [Mon, 23 Aug 2004 12:41:29 +0000 (12:41 +0000)]
Don't try to translate the control message unless we're certain it's
valid; otherwise a caller could trick us into changing any 32-bit word
in kernel memory to LINUX_SOL_SOCKET (0x00000001) if its previous value
is SOL_SOCKET (0x0000ffff).

MFC after: 3 days

20 years agoUpdate documentation to match reality: glob() sorts its result according
tjr [Mon, 23 Aug 2004 12:10:44 +0000 (12:10 +0000)]
Update documentation to match reality: glob() sorts its result according
to the current LC_COLLATE setting, not in "ASCII order".

PR: 54391
MFC after: 1 week

20 years agoRestore FD_DEBUG for pc98.
nyan [Mon, 23 Aug 2004 11:59:32 +0000 (11:59 +0000)]
Restore FD_DEBUG for pc98.

20 years agos/the on/on the/
blackend [Mon, 23 Aug 2004 11:00:20 +0000 (11:00 +0000)]
s/the on/on the/

PR: misc/70842
Submitted by: Sean C. Farley <sean-freebsd@farley.org>

20 years agoo Fix whitespace bug introduced in the previous commit.
sobomax [Mon, 23 Aug 2004 10:09:29 +0000 (10:09 +0000)]
o Fix whitespace bug introduced in the previous commit.

Submitted by: ru

o Simplify p4tcc_power_profile().

Submitted by: maxim

20 years agoo Extend boot output: print out mimimum/maximum performance value and number
sobomax [Mon, 23 Aug 2004 09:47:56 +0000 (09:47 +0000)]
o Extend boot output: print out mimimum/maximum performance value and number
  of performance steps available;

o similarly to Enhanced SpeedStep driver, export list of all available steps
  via hw.p4tcc.cpuperf_levels sysctl.

20 years agoAdd missing GEOM classes, which are aware of geom(8).
pjd [Mon, 23 Aug 2004 06:23:17 +0000 (06:23 +0000)]
Add missing GEOM classes, which are aware of geom(8).

Submitted by: kuriyama

20 years agoMake this compile again in the standalone and the MODULES_WITH_WORLD
imp [Mon, 23 Aug 2004 03:38:21 +0000 (03:38 +0000)]
Make this compile again in the standalone and the MODULES_WITH_WORLD
environments.  Chances are good that this doesn't produce a good
module, but I leave the proper defaults to the dummy opt_* files to
the author.

20 years agoRemove in6_prefix.[ch] and the contained router renumbering capability.
rwatson [Mon, 23 Aug 2004 03:00:27 +0000 (03:00 +0000)]
Remove in6_prefix.[ch] and the contained router renumbering capability.
The prefix management code currently resides in nd6, leaving only the
unused router renumbering capability in the in6_prefix files.  Removing
it will make it easier for us to provide locking for the remainder of
IPv6 by reducing the number of objects requiring synchronized access.

This functionality has also been removed from NetBSD and OpenBSD.

Submitted by: George Neville-Neil <gnn at neville-neil.com>
Discussed with/approved by: suz, keiichi at kame.net, core at kame.net

20 years agoTemporarily back out r1.74 as it seems to cause a number of regressions
kan [Mon, 23 Aug 2004 02:39:45 +0000 (02:39 +0000)]
Temporarily back out r1.74 as it seems to cause a number of regressions
accordimg to numerous reports. It  might get reintroduced some time later
when an exact failure mode is understood better.

20 years agoPass a correct lowaddr to bus_dma_tag_create(), lnc(4) cards can only
mux [Sun, 22 Aug 2004 23:01:13 +0000 (23:01 +0000)]
Pass a correct lowaddr to bus_dma_tag_create(), lnc(4) cards can only
deal with 24-bit addresses.  While the two other attachments, namely
isa and cbus, do it properly, the PCI attachment was passing
BUS_SPACE_MAXADDR instead of BUS_SPACE_MAXADDR_24BIT.  This bug
became apparent with the new contigmalloc() code.

This fixes the problem reported with lnc(4) interfaces inside VMWare,
and should theoritically also fix any user of a PCI lnc(4) card.  It
is a RELENG_5 MFC candidate.

Tested by: Florian Le Goff <madflo@beertech.org>

20 years agoStyle update: use newer style function prototypes in if_sl.c in
rwatson [Sun, 22 Aug 2004 21:32:52 +0000 (21:32 +0000)]
Style update: use newer style function prototypes in if_sl.c in
prep for merging locking.