]> git.neil.brown.name Git - history.git/log
history.git
23 years ago[WATCHDOG] fix sc1200wdt for CONFIG_PNP=n
Dave Jones [Thu, 13 Feb 2003 09:19:05 +0000 (08:19 -0100)]
[WATCHDOG] fix sc1200wdt for CONFIG_PNP=n

23 years ago[WATCHDOG] C99 struct initialisers for sc1200wdt
Dave Jones [Thu, 13 Feb 2003 09:18:11 +0000 (08:18 -0100)]
[WATCHDOG] C99 struct initialisers for sc1200wdt

23 years ago[WATCHDOG] printk levels for alim7101_wdt.c
Dave Jones [Thu, 13 Feb 2003 09:03:09 +0000 (08:03 -0100)]
[WATCHDOG] printk levels for alim7101_wdt.c

23 years ago[WATCHDOG] missing printk level in acquirewdt
Dave Jones [Thu, 13 Feb 2003 09:00:15 +0000 (08:00 -0100)]
[WATCHDOG] missing printk level in acquirewdt

23 years ago[WATCHDOG] More panic -> shutdown replacements in pcwd.c
Dave Jones [Wed, 12 Feb 2003 22:00:45 +0000 (21:00 -0100)]
[WATCHDOG] More panic -> shutdown replacements in pcwd.c

23 years ago[WATCHDOG] printk levels for pcwd.c
Dave Jones [Wed, 12 Feb 2003 21:58:09 +0000 (20:58 -0100)]
[WATCHDOG] printk levels for pcwd.c

23 years ago[WATCHDOG] pcwd.c: if cpu has overheated, we want to shutdown, not panic.
Dave Jones [Wed, 12 Feb 2003 21:55:49 +0000 (20:55 -0100)]
[WATCHDOG] pcwd.c: if cpu has overheated, we want to shutdown, not panic.
Based on a patch from Tariq Shureih <tariq@linux.co.intel.com>

23 years ago[PATCH] fix adaptec diagnostics for ppc64
Andrew Morton [Wed, 12 Feb 2003 05:17:21 +0000 (21:17 -0800)]
[PATCH] fix adaptec diagnostics for ppc64

The return type of sizeof() is size_t.  On many architectrues size_t is
unsigned long, and may not be printed with %d.  Use %Zu instead.

23 years ago[PATCH] u14-34f fix
Andrew Morton [Wed, 12 Feb 2003 05:09:17 +0000 (21:09 -0800)]
[PATCH] u14-34f fix

Need to reenable interrupts around the call to scsi_register(), which blocks.

23 years ago[PATCH] MAP_FIXED|MAP_ANON crash fix
Andrew Morton [Wed, 12 Feb 2003 05:09:05 +0000 (21:09 -0800)]
[PATCH] MAP_FIXED|MAP_ANON crash fix

An anonymous fixed mmap with CONFIG_HUGETLB_PAGE=y will dereference NULL.  We
recur into the pagefault handler with mmap_sem held and lock up.

23 years ago[PATCH] epoll timeout and syscall return types
Andrew Morton [Wed, 12 Feb 2003 05:08:51 +0000 (21:08 -0800)]
[PATCH] epoll timeout and syscall return types

Patch from Davide Libenzi <davidel@xmailserver.org>

Changes :
 - Timeout overflow check
 - Ceil()ing of ms->jif conversion
 - Syscalls return type int->long

23 years ago[PATCH] OSS CS4232 locking fixes
Andrew Morton [Wed, 12 Feb 2003 05:08:37 +0000 (21:08 -0800)]
[PATCH] OSS CS4232 locking fixes

Patch from: Peter Waechtler <pwaechtler@mac.com>

Addresses BUGME bug #331.  "OSS CS4232 nasty spinlock printks on boot"

The locking in some OSS modules is really lousy.
Because save_flags/cli/restore_flags could be used recursivly - the
programmers pushed the locking too far the lower level.

Because on ISA cards the register sets are usually multiplexed
you had to write to an address latch and then access the data port
in an "atomic" manner.

I suggest removing the locking from ad_read/ad_write +
ad_{enter|leave}_MCE and clamping the locks wherever the functions
are called. I hope the attached patch does that correctly.

Yes, I don't like all the timeout loops while holding the locks:
high chances that a cpu is spinning in interrupt context :(

23 years ago[PATCH] fix fadvise64() return type
Andrew Morton [Wed, 12 Feb 2003 05:08:24 +0000 (21:08 -0800)]
[PATCH] fix fadvise64() return type

Patch from: David Mosberger <davidm@napali.hpl.hp.com>

Please remember to declare the return-type of syscall stubs as "long".
On 64-bit platforms, it's generally necessary to ensure that the
entire 64-bit return value is valid (and can be checked against
negative values).

23 years ago[PATCH] drivers/media/video/saa5249.c compile fix
Andrew Morton [Wed, 12 Feb 2003 05:08:11 +0000 (21:08 -0800)]
[PATCH] drivers/media/video/saa5249.c compile fix

Patch from: Frank Davis <fdavis@si.rr.com>

The following addresses bugzilla bug # 340.

23 years ago[PATCH] Make drivers/media/video/saa7110.c compile
Andrew Morton [Wed, 12 Feb 2003 05:08:00 +0000 (21:08 -0800)]
[PATCH] Make drivers/media/video/saa7110.c compile

Patch from: Frank Davis <fdavis@si.rr.com>

The following patch addresses bugzilla bug # 341.

23 years ago[PATCH] Allow summit kernels to boot on normal systems
Andrew Morton [Wed, 12 Feb 2003 05:07:46 +0000 (21:07 -0800)]
[PATCH] Allow summit kernels to boot on normal systems

Patch from: john stultz <johnstul@us.ibm.com>

This patch simply fixes the summit subarch to allow summit kernels to
boot on normal systems.

23 years ago[PATCH] make drivers/net/arlan.c compile again
Andrew Morton [Wed, 12 Feb 2003 05:07:32 +0000 (21:07 -0800)]
[PATCH] make drivers/net/arlan.c compile again

Someone typed `config' when they meant `conf'.

Also fix a function-defined-but-not-used warning.

23 years ago[PATCH] EATA driver fix
Andrew Morton [Wed, 12 Feb 2003 05:07:17 +0000 (21:07 -0800)]
[PATCH] EATA driver fix

This driver is calling down into scsi_register with local interrupts
disabled.  scsi_register performs blocking allocations, starts kernel
threads, etc.  slab debugging gets offended by someone performing blocking
operations with local interrupts disabled.

23 years ago[PATCH] jiffies wrap fixes
Andrew Morton [Wed, 12 Feb 2003 05:07:03 +0000 (21:07 -0800)]
[PATCH] jiffies wrap fixes

Patch from Tim Schmielau <tim@physik3.uni-rostock.de>

Converts many open-coded jiffy comparisons to use time_after/before/etc.

23 years ago[PATCH] sunrpc dcache cleanup
Andrew Morton [Wed, 12 Feb 2003 05:06:47 +0000 (21:06 -0800)]
[PATCH] sunrpc dcache cleanup

Patch from Dipankar Sarma <dipankar@in.ibm.com>

All fs should be using dcache APIs to manipulate dcache hash lists.  This is
in line with the dcache cleanup patch (dcache_rcu-1) from Maneesh that Linus
accepted.  This seems like a reasonable cleanup.  One change though, we don't
need to grab dcache_lock while deleting dentries from the private list and
__d_drop() should suffice here.

23 years ago[PATCH] make the adaptec driver compile
Andrew Morton [Wed, 12 Feb 2003 05:06:35 +0000 (21:06 -0800)]
[PATCH] make the adaptec driver compile

make the adaptec driver compile

23 years ago[PATCH] disassociate_ctty SMP fix
Andrew Morton [Wed, 12 Feb 2003 05:06:21 +0000 (21:06 -0800)]
[PATCH] disassociate_ctty SMP fix

Patch from Rik van Riel <riel@conectiva.com.br>

the following patch, against today's BK tree, fixes a small
SMP race in disassociate_ctty.  This function gets called
from do_exit, without the BKL held.

However, it sets the *tty variable before grabbing the bkl,
then makes decisions on what the variable was set to before
the lock was grabbed, despite the fact that another process
could modify its ->tty pointer in this same function.

23 years ago[PATCH] provide uniproc write_trylock()
Andrew Morton [Wed, 12 Feb 2003 05:06:07 +0000 (21:06 -0800)]
[PATCH] provide uniproc write_trylock()

Patch from Oleg Drokin <green@namesys.com>, Nikita Danilov <Nikita@Namesys.COM>

There is no uniprocessor definition of _raw_write_trylock(), so
write_trylock() doesn't work on UP.

23 years ago[PATCH] fix drivers/scsi/st.c warning
Andrew Morton [Wed, 12 Feb 2003 05:05:52 +0000 (21:05 -0800)]
[PATCH] fix drivers/scsi/st.c warning

kdev_t.value is not an int.  Code looks fishy.

23 years ago[PATCH] fs/reiserfs/hashes.c warning fix
Andrew Morton [Wed, 12 Feb 2003 05:05:38 +0000 (21:05 -0800)]
[PATCH] fs/reiserfs/hashes.c warning fix

printk needs kernel.h

23 years ago[PATCH] fix ppc64 nfs warning
Andrew Morton [Wed, 12 Feb 2003 05:05:25 +0000 (21:05 -0800)]
[PATCH] fix ppc64 nfs warning

Don't assume that dev_t is unsigned.  Cast it.

23 years ago[PATCH] fix ppc64 wanings in fs/partitions/check.c
Andrew Morton [Wed, 12 Feb 2003 05:05:11 +0000 (21:05 -0800)]
[PATCH] fix ppc64 wanings in fs/partitions/check.c

cast dev_t when printing

23 years ago[PATCH] kill some ppc64 warnings in knfsd
Andrew Morton [Wed, 12 Feb 2003 05:04:57 +0000 (21:04 -0800)]
[PATCH] kill some ppc64 warnings in knfsd

Don't assume that dev_t is an unsigned.  Cast it.

23 years ago[PATCH] kill warning in vmscan.c
Andrew Morton [Wed, 12 Feb 2003 05:04:43 +0000 (21:04 -0800)]
[PATCH] kill warning in vmscan.c

Make the "duplicate const" warning go away.  Arguably a compiler bug...

23 years ago[PATCH] genhd warnings fix
Andrew Morton [Wed, 12 Feb 2003 05:04:29 +0000 (21:04 -0800)]
[PATCH] genhd warnings fix

I have a whole bunch of silly compile warning fixes here, arising from
building the kernel for a 64-bit target.  Some are trivial, some are genuine
printk bugs.

assuming dev_t is unsigned generates a warning on ppc64.  Cast it.

23 years agoMerge bk://linux-dj.bkbits.net/cpufreq
Linus Torvalds [Wed, 12 Feb 2003 04:51:04 +0000 (20:51 -0800)]
Merge bk://linux-dj.bkbits.net/cpufreq
into penguin.transmeta.com:/home/penguin/torvalds/repositories/kernel/linux

23 years ago[CPUFREQ] fix longrun min/max confusion.
Dave Jones [Wed, 12 Feb 2003 19:41:23 +0000 (18:41 -0100)]
[CPUFREQ] fix longrun min/max confusion.

23 years ago[CPUFREQ] add support for cpufreq governors.
Dave Jones [Wed, 12 Feb 2003 19:23:16 +0000 (18:23 -0100)]
[CPUFREQ] add support for cpufreq governors.
More bits from Dominik.

Most cpufreq drivers (in fact, all except one, longrun) or even most
cpu frequency scaling algorithms only offer the CPU to be set to one
frequency. In order to offer dynamic frequency scaling, the cpufreq
core must be able to tell these drivers of a "target frequency". So
these specific drivers will be transformed to offer a "->target"
call instead of the existing "->setpolicy" call. For "longrun", all
stays the same, though.

How to decide what frequency within the CPUfreq policy should be used?
That's done using "cpufreq governors". Two are already in this patch
-- they're the already existing "powersave" and "performance" which
set the frequency statically to the lowest or highest frequency,
respectively. At least two more such governors will be ready for
addition in the near future, but likely many more as there are various
different theories and models about dynamic frequency scaling
around. Using such a generic interface as cpufreq offers to scaling
governors, these can be tested extensively, and the best one can be
selected for each specific use.

Basically, it's the following flow graph:

CPU can be set to switch independetly    |         CPU can only be set
      within specific "limits"           |       to specific frequencies

                                 "CPUfreq policy"
                consists of frequency limits (policy->{min,max})
                     and CPUfreq governor to be used
                         /                    \
                        /                      \
                       /                       the cpufreq governor decides
                      /                        (dynamically or statically)
                     /                         what target_freq to set within
                    /                          the limits of policy->{min,max}
                   /                                \
                  /                                  \
        Using the ->setpolicy call,              Using the ->target call,
            the limits and the                    the frequency closest
             "policy" is set.                     to target_freq is set.
                                                  It is assured that it
                                                  is within policy->{min,max}

23 years ago[CPUFREQ] Properly set memory allocated by x86 cpufreq drivers to zero.
Dave Jones [Wed, 12 Feb 2003 19:22:00 +0000 (18:22 -0100)]
[CPUFREQ] Properly set memory allocated by x86 cpufreq drivers to zero.
From Dominik Brodowski

23 years ago[PATCH] Fix stack handling in acpi_wakeup.S
Pavel Machek [Wed, 12 Feb 2003 04:50:16 +0000 (20:50 -0800)]
[PATCH] Fix stack handling in acpi_wakeup.S

This fixes stack handling in acpi_wakeup.S, and makes stack smaller so
that wakeup code actually fits inside memory allocated for it. Plus
someone renamed .L1432 to something meaningful.

23 years agoDon't wake up processes unnecessarily for ignored signals
Linus Torvalds [Wed, 12 Feb 2003 04:49:03 +0000 (20:49 -0800)]
Don't wake up processes unnecessarily for ignored signals

23 years ago[PATCH] alpha: Add missing sighand bits
Ivan Kokshaysky [Wed, 12 Feb 2003 01:49:42 +0000 (17:49 -0800)]
[PATCH] alpha: Add missing sighand bits

23 years ago[PATCH] export allow_signal()
Tomas Szepe [Wed, 12 Feb 2003 01:25:32 +0000 (17:25 -0800)]
[PATCH] export allow_signal()

Export allow_signal().  It's needed by lockd, sunrpc and other modules.

23 years ago[PATCH] x86_64 compatibility layer update
Stephen Rothwell [Tue, 11 Feb 2003 13:37:16 +0000 (05:37 -0800)]
[PATCH] x86_64 compatibility layer update

Andi has asked that I send these straight forward compatibility patches
to you and he will fix up any merge problems later.  These are the
outstanding patches for x86_64 against 2.5.60.

23 years ago[PATCH] parisc compatibility layer update
Stephen Rothwell [Tue, 11 Feb 2003 13:37:08 +0000 (05:37 -0800)]
[PATCH] parisc compatibility layer update

At Linux Conf AU, Willy asked me to send any further parisc compatibility
changes directly to you, so this is what I have outstanding.  Basically,
it is just the uses of compat_sigset_t that seemed to have been missed in
the previous merges.

23 years ago[PATCH] x86-64 merge
Andi Kleen [Tue, 11 Feb 2003 13:20:56 +0000 (05:20 -0800)]
[PATCH] x86-64 merge

This brings the x86-64 port uptodate in 2.5.60. Unfortunately I cannot
test too much because i constantly get deadlocks in exit/wait in initscripts
on SMP bootup. The kernel seems to still lose a lot of SIGCHLD. 2.5.59/SMP
had the same problem. Uniprocessor and SMP kernel on UP seems to work.

This patch only touches x86-64 specific files. It requires a few simple
changes to arch independent files that I will send separately.

 - Fixed a lot of obsolete/misleading configure help texts.
 - Remove old bootblock disk loader and support fdimage target for syslinux
   instead (H. Peter Anvin)
 - Fix potential fpu signal restore problem on 32bit emulation.
 - Merge with 2.5.60 i386 (hugetlbfs, acpi etc.)
 - Some fixes for local apic disabled modus.
 - Beginngs of S3 ACPI wakeup from real-mode (not working yet, don't use)
 - Beginnings of NUMA/CONFIG_DISCONTIGMEM support for AMD K8 (work in progress,
   port from 2.4): clean up memory mapping at bootup, generalize bootmem etc.
 - Fix 64bit GS base reload problem and reenable (Karsten Keil)
 - Fix race with vmalloc accesses from interrupt handlers disturbing page fault/
   similar race for the debug handler (thanks to Andrew Morton)
 - Merge cpu access primitives with i386
 - Revert to private module list for now because putting modules
   nto vmlist triggered too many problems.
 - Some cleanups, removal of unneeded code.
 - Let early __get_free_pages see consistent pda
 - Preempt disabled for now because it is too broken right now
 - Signal handler fixes
 - Fix do_gettimeofday to be completely lockless and reenable vsyscalls
 - Optimize context switch path a bit (should be ported to i386)
 - Get thread_info via stack for better code
 - Don't leak pmd pages
 - Clean up hardcoded task stack sizes.

23 years ago[PATCH] apm daemonize
Stephen Rothwell [Tue, 11 Feb 2003 12:22:21 +0000 (04:22 -0800)]
[PATCH] apm daemonize

Since daemonize now blocks all signals, this simplification is trivially
correct.

23 years agoMerge bk://kernel.bkbits.net/davem/sparc-2.5
Linus Torvalds [Tue, 11 Feb 2003 10:58:50 +0000 (02:58 -0800)]
Merge bk://kernel.bkbits.net/davem/sparc-2.5
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoMerge bk://kernel.bkbits.net/davem/net-2.5
Linus Torvalds [Tue, 11 Feb 2003 10:48:21 +0000 (02:48 -0800)]
Merge bk://kernel.bkbits.net/davem/net-2.5
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoMerge http://linux-isdn.bkbits.net/linux-2.5.make
Linus Torvalds [Tue, 11 Feb 2003 10:47:46 +0000 (02:47 -0800)]
Merge http://linux-isdn.bkbits.net/linux-2.5.make
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years agoMerge tp1.ruhr-uni-bochum.de:/scratch/kai/kernel/v2.5/linux-2.5
Kai Germaschewski [Tue, 11 Feb 2003 11:41:41 +0000 (05:41 -0600)]
Merge tp1.ruhr-uni-bochum.de:/scratch/kai/kernel/v2.5/linux-2.5
into tp1.ruhr-uni-bochum.de:/scratch/kai/kernel/v2.5/linux-2.5.make

23 years ago[PATCH] nfs fix
Andries E. Brouwer [Tue, 11 Feb 2003 10:46:39 +0000 (02:46 -0800)]
[PATCH] nfs fix

nfs must not use MINORBITS - that fails with 32-bit dev_t

23 years agoMerge davem@nuts.ninka.net:/home/davem/src/BK/net-2.5
David S. Miller [Tue, 11 Feb 2003 09:42:51 +0000 (01:42 -0800)]
Merge davem@nuts.ninka.net:/home/davem/src/BK/net-2.5
into kernel.bkbits.net:/home/davem/net-2.5

23 years agoMerge davem@nuts.ninka.net:/home/davem/src/BK/sparc-2.5
David S. Miller [Tue, 11 Feb 2003 09:33:57 +0000 (01:33 -0800)]
Merge davem@nuts.ninka.net:/home/davem/src/BK/sparc-2.5
into kernel.bkbits.net:/home/davem/sparc-2.5

23 years agokbuild: scripts/fixdep.c doesn't close files when finished
Kai Germaschewski [Tue, 11 Feb 2003 08:54:42 +0000 (02:54 -0600)]
kbuild: scripts/fixdep.c doesn't close files when finished

From Chris Wedgwood:

> fixdep doesn't close files when finished with them... normally this
> doesn't matter unless you have strict ulimits in place.
>
> Trivial fix is:

23 years agoAdd macthing 'va_end()' to the 'va_start()' in daemonize().
Linus Torvalds [Tue, 11 Feb 2003 08:10:57 +0000 (00:10 -0800)]
Add macthing 'va_end()' to the 'va_start()' in daemonize().

Not that it makes any difference on x86, but there may be
architectures that actually need it.

23 years agoSanitize kernel daemon signal handling and process naming.
Linus Torvalds [Tue, 11 Feb 2003 06:58:04 +0000 (22:58 -0800)]
Sanitize kernel daemon signal handling and process naming.

Add a name argument to daemonize() (va_arg) to avoid all the
kernel threads having to duplicate the name setting over and
over again.

Make daemonize() disable all signals by default, and add a
"allow_signal()" function to let daemons say they explicitly
want to support a signal.

Make flush_signal() take the signal lock, so that callers do
not need to.

23 years agoReport shared pending signals in /proc/<pid>/status
Linus Torvalds [Tue, 11 Feb 2003 06:04:23 +0000 (22:04 -0800)]
Report shared pending signals in /proc/<pid>/status

Patch from Roland McGrath.

23 years ago[PATCH] oprofile: kernel/user addresses fix
John Levon [Tue, 11 Feb 2003 04:26:06 +0000 (20:26 -0800)]
[PATCH] oprofile: kernel/user addresses fix

This patch replaces the assumption that > PAGE_OFFSET == kernel address
with testing for user_mode(regs) and inserting switch codes instead.

23 years ago[PATCH] oprofile: fix oprofilefs integer files base
John Levon [Tue, 11 Feb 2003 04:25:51 +0000 (20:25 -0800)]
[PATCH] oprofile: fix oprofilefs integer files base

This patch allows the oprofilefs files to take entry in any base
instead of just base 10

23 years ago[PATCH] oprofile: CPU type as string
John Levon [Tue, 11 Feb 2003 04:25:38 +0000 (20:25 -0800)]
[PATCH] oprofile: CPU type as string

This patch updates the horrible enum for the logical CPU type with a
string instead.

23 years ago[PATCH] oprofile: Pentium IV support
John Levon [Tue, 11 Feb 2003 04:25:24 +0000 (20:25 -0800)]
[PATCH] oprofile: Pentium IV support

The below patch implements a P4 driver for OProfile, mostly written
by Graydon Hoare.

23 years agoMerge bk://linux-dj.bkbits.net/watchdog
Linus Torvalds [Tue, 11 Feb 2003 03:15:38 +0000 (19:15 -0800)]
Merge bk://linux-dj.bkbits.net/watchdog
into home.transmeta.com:/home/torvalds/v2.5/linux

23 years ago[WATCHDOG] PNP API conversion.
Dave Jones [Tue, 11 Feb 2003 03:36:51 +0000 (02:36 -0100)]
[WATCHDOG] PNP API conversion.
Done by Adam Belay.

23 years agoIf we set TIF_SIGPENDING for SIGCONT, we have to wake up any sleeping
Linus Torvalds [Tue, 11 Feb 2003 03:05:34 +0000 (19:05 -0800)]
If we set TIF_SIGPENDING for SIGCONT, we have to wake up any sleeping
tasks (even if we don't otherwise need to wake anything up), since
otherwise later signals would see that signals are already pending and
wouldn't cause wakeups.

23 years ago[PATCH] genhd device unregistration fix
Andries E. Brouwer [Tue, 11 Feb 2003 01:23:47 +0000 (17:23 -0800)]
[PATCH] genhd device unregistration fix

23 years ago[PATCH] signal error return fix
Andries E. Brouwer [Tue, 11 Feb 2003 01:23:38 +0000 (17:23 -0800)]
[PATCH] signal error return fix

23 years ago[PATCH] sched_init enables interrupts too early
Andrew Morton [Tue, 11 Feb 2003 01:16:52 +0000 (17:16 -0800)]
[PATCH] sched_init enables interrupts too early

wake_up_forked_process() unconditionally enables interrupts.  It is called
from sched_init().  Enabling interrupts that early makes Anton's ppc64
machine lock up.

23 years ago[PATCH] bounds/limits fixes (Stanford Checker)
Randy Dunlap [Tue, 11 Feb 2003 01:12:57 +0000 (17:12 -0800)]
[PATCH] bounds/limits fixes (Stanford Checker)

This fixes several trivial bounds/limits errors that were pointed out by
the Stanford Checker.

23 years ago[PATCH] ftape divide-by-zero found by Stanford Checker
Randy Dunlap [Tue, 11 Feb 2003 01:12:49 +0000 (17:12 -0800)]
[PATCH] ftape divide-by-zero found by Stanford Checker

This fixes a potential divide-by-zero found by the Stanford Checker.

23 years agoMerge nuts.ninka.net:/home/davem/src/BK/sparcwork-2.5
David S. Miller [Mon, 10 Feb 2003 20:35:58 +0000 (12:35 -0800)]
Merge nuts.ninka.net:/home/davem/src/BK/sparcwork-2.5
into nuts.ninka.net:/home/davem/src/BK/sparc-2.5

23 years ago[SIGNAL]: Allow more platforms to use generic get_signal_to_deliver.
David S. Miller [Mon, 10 Feb 2003 20:27:26 +0000 (12:27 -0800)]
[SIGNAL]: Allow more platforms to use generic get_signal_to_deliver.

The few platforms that cannot use the generic
get_signal_to_deliver implementation cannot do
so because they do special things for ptraced
children.  This can be easily avoided and thus
all of the signal handling code duplication can
be eliminated.

This is the first part, which adds a platform hook
right before the parent of the ptraced child is woken.
Data can be passed in via a cookie argument.

The next part will be dealing with platforms
that need to muck with breakpoints in the child
in this same code block.

23 years ago[TCP]: Do not bump backoff too high during 0-window probes.
David S. Miller [Mon, 10 Feb 2003 20:09:47 +0000 (12:09 -0800)]
[TCP]: Do not bump backoff too high during 0-window probes.

23 years ago[TULIP DE4X5]: Cannot use initdata before including linux/init.h
David S. Miller [Mon, 10 Feb 2003 20:06:51 +0000 (12:06 -0800)]
[TULIP DE4X5]: Cannot use initdata before including linux/init.h

23 years ago[SPARC]: ADd init_sighand.
Rob Radez [Mon, 10 Feb 2003 19:42:31 +0000 (11:42 -0800)]
[SPARC]: ADd init_sighand.

23 years ago[SPARC]: HEAD --> HEAD_Y
Rob Radez [Mon, 10 Feb 2003 19:41:03 +0000 (11:41 -0800)]
[SPARC]: HEAD --> HEAD_Y

23 years ago[ECONET]: Add comment to point out a bug spotted by Joern Engel.
Rusty Russell [Mon, 10 Feb 2003 19:38:29 +0000 (11:38 -0800)]
[ECONET]: Add comment to point out a bug spotted by Joern Engel.

23 years ago[X25]: Fix improper | precendence, pointed out by Joern Engel.
Rusty Russell [Mon, 10 Feb 2003 19:30:44 +0000 (11:30 -0800)]
[X25]: Fix improper | precendence, pointed out by Joern Engel.

23 years ago[AF_UNIX] Cleanup forall_unix_sockets.
Rusty Russell [Mon, 10 Feb 2003 19:22:57 +0000 (11:22 -0800)]
[AF_UNIX] Cleanup forall_unix_sockets.

23 years ago[PATCH] Get 3c59x to compile on non-PCI systems
Andrew Morton [Mon, 10 Feb 2003 15:37:27 +0000 (07:37 -0800)]
[PATCH] Get 3c59x to compile on non-PCI systems

Patch from Marc Zyngier <mzyngier@freesurf.fr>

My previous round of EISA hacking left the 3c59x driver unable to
compile on non-PCI systems (that is, EISA only...).

This small patch fixes it.

23 years ago[PATCH] 3c509 compile fix
Andrew Morton [Mon, 10 Feb 2003 15:37:20 +0000 (07:37 -0800)]
[PATCH] 3c509 compile fix

Patch from Marc Zyngier <mzyngier@freesurf.fr>

Fix compilation of 3c509.c

23 years ago[PATCH] fix current->user->processes leak
Andrew Morton [Mon, 10 Feb 2003 15:37:12 +0000 (07:37 -0800)]
[PATCH] fix current->user->processes leak

Patch from: Eric Lammerts <eric@lammerts.org>

Every time you do a loop mount, a kernel thread is started (those
processes are called "loop0", "loop1", etc.). The problem is that when
it starts, it's counted as one of your processes. Then, it's
changed to be a root-owned process without correcting that count.

Patch below fixes the problem. It moves the bookkeeping of changing
current->user to a new function switch_uid() (which is now also used
by exec_usermodehelper() in kmod.c). The patch is tested.

23 years ago[PATCH] remove the buffer_head mempool
Andrew Morton [Mon, 10 Feb 2003 15:37:03 +0000 (07:37 -0800)]
[PATCH] remove the buffer_head mempool

mempools have the wrong semantics for use by buffer_heads.  The problem
scenario:

- Process A calls mempool_alloc(), asking for a buffer_head.

- While process A sleeps, process B frees up a ton of memory.

That's it.  There is no longer any memory pressure, so nobody frees any
buffer_heads, so process A does not get woken up.  I managed to trigger this
in some testing recently.

One approach would be to use a schedule_timeout(2) in mempool_alloc().

Anyway, the importance of buffer_head allocation was lessened when swapout
stopped using them, so let's just drop the mempool out of it for now.

23 years ago[PATCH] Don't run unlock_super() in ext3_fill_super()
Andrew Morton [Mon, 10 Feb 2003 15:36:54 +0000 (07:36 -0800)]
[PATCH] Don't run unlock_super() in ext3_fill_super()

A very long time ago ext3_fill_super() was beng called under the VFS's
lock_super(), and it had to drop that lock for deadlock avoidance reasons.

But the VFS was later changed to not take lock_super() across ->fill_super(),
so we're upping a non-downed semaphore.

Just take it out.

23 years ago[PATCH] ext3 commenting cleanup
Andrew Morton [Mon, 10 Feb 2003 15:36:46 +0000 (07:36 -0800)]
[PATCH] ext3 commenting cleanup

Redo the (strange) layout of some of Roger's comments.

23 years ago[PATCH] remove unneeded test from radix_tree_extend()
Andrew Morton [Mon, 10 Feb 2003 15:36:39 +0000 (07:36 -0800)]
[PATCH] remove unneeded test from radix_tree_extend()

Patch from: Szabolcs Berecz <szabi@mplayerhq.hu>

No need to check root->rnode in the loop. In the loop it remains nonzero.

23 years ago[PATCH] hugetlbpage MAP_FIXED fix
Andrew Morton [Mon, 10 Feb 2003 15:36:32 +0000 (07:36 -0800)]
[PATCH] hugetlbpage MAP_FIXED fix

We need to validate that the address and length of a MAP_FIXED request are
suitable for hugetlb pages.

23 years ago[PATCH] nforce2 IDE support for the amd74xx driver
Andrew Morton [Mon, 10 Feb 2003 15:36:23 +0000 (07:36 -0800)]
[PATCH] nforce2 IDE support for the amd74xx driver

Patch from James Curbo <phoenix@sandwich.net>

The amd74xx IDE driver in 2.5.59 has support for the nforce IDE controller,
but not explicitly for the nforce2 IDE controller (which has a different PCI
ID, which is in the kernel already).  I'm not sure if the nforce and nforce2
controllers are identical, but I made a small patch that made the amd74xx
driver recognize the nforce2 IDE, and it boots for me, seems to work fine, as
my drives were tuned to their highest transfer rate automatically (udma5).

I don't know if this patch is proper or correct, but it Works for Me [tm].
Patch is attached.

Vojtech has acked this change.

23 years ago[PATCH] Add David Olien MAINTAINERs for DAC960
Andrew Morton [Mon, 10 Feb 2003 15:36:16 +0000 (07:36 -0800)]
[PATCH] Add David Olien MAINTAINERs for DAC960

(I asked Dave for this...)

From: Dave Olien <dmo@osdl.org>

Add me as maintainer for DAC960 driver.

23 years ago[PATCH] DAC960 Stanford Checker fix
Andrew Morton [Mon, 10 Feb 2003 15:36:07 +0000 (07:36 -0800)]
[PATCH] DAC960 Stanford Checker fix

Patch from Dave Olien <dmo@osdl.org>

This was found by the Standford Checker.
The LogicalDeviceNumber bad range test was changed from > to >=
I also replaced a couple of panic() calls with error messages,
since panic-ing seemed a little extreme.

23 years ago[PATCH] ext3: Remove journal_try_start()
Andrew Morton [Mon, 10 Feb 2003 15:36:00 +0000 (07:36 -0800)]
[PATCH] ext3: Remove journal_try_start()

journal_try_start() is a function which nonblockingly attempts to open a JBD
transaction handle.  It was added a long time ago when there were concerns
that ext3_writepage() could block kswapd for too long.

It was never clearly necessary.

So the patch throws it all away and just calls the blocking journal_start()
from ext3_writepage().

23 years ago[PATCH] NUMAQ io_apic programming fix
Andrew Morton [Mon, 10 Feb 2003 15:35:52 +0000 (07:35 -0800)]
[PATCH] NUMAQ io_apic programming fix

Patch from William Lee Irwin III <wli@holomorphy.com>

Don't touch IO-APIC physid's on NUMA-Q. The BIOS pre-programs them.

23 years ago[PATCH] use per-cpu data for ia32 profiler
Andrew Morton [Mon, 10 Feb 2003 15:35:43 +0000 (07:35 -0800)]
[PATCH] use per-cpu data for ia32 profiler

Patch from Ravikiran G Thirumalai <kiran@in.ibm.com>

Converts the ia32 kernel profiler to use use the percpu area infrastructure.

23 years ago[PATCH] uninline get_jiffies_64() for 32-bit architectures
Andrew Morton [Mon, 10 Feb 2003 15:35:35 +0000 (07:35 -0800)]
[PATCH] uninline get_jiffies_64() for 32-bit architectures

uninline get_jiffies_64() for 32-bit architectures

23 years agokbuild: Allow for ',$ in commands
Kai Germaschewski [Mon, 10 Feb 2003 15:27:47 +0000 (09:27 -0600)]
kbuild: Allow for ',$ in commands

Quote ',$ before passing them to the shell - ' must be escaped for
echo, $$ will be converted to $ by make.

This allows to use e.g. perl commands in the build with support
from the generic $(call if_changed,...).

by Konrad Eisele

23 years agokbuild: Customflags for cmd_objcopy
Kai Germaschewski [Mon, 10 Feb 2003 15:20:31 +0000 (09:20 -0600)]
kbuild: Customflags for cmd_objcopy

From Konrad Eisele:

> like with cmd_ld in scripts/Makefile.lib having possibility to add
> customflags with cmd_objcopy would be nice. When building a
> ROMKernel I'd like to use:
> OBJCOPYFLAGS_rompiggydata := --remove-section=.text
> OBJCOPYFLAGS_$(MODEL)piggytext := --only-section=.text

23 years agokbuild: cosmetics
Kai Germaschewski [Mon, 10 Feb 2003 15:17:10 +0000 (09:17 -0600)]
kbuild: cosmetics

From ram <ram@curvesoft.com>:

> Appended below is a small patch to the top-level makefile; it
> -- replaces a call to $(shell/echo/sed) with $(subst) and adds a
>    comment
> -- fixes some typos.

23 years agokbuild: Handle the "no modules" case
Kai Germaschewski [Mon, 10 Feb 2003 15:07:25 +0000 (09:07 -0600)]
kbuild: Handle the "no modules" case

When the user selects CONFIG_MODVERSIONS but doesn't build anything
modular, the post-processing step does nothing (right, as there is
nothing to be done), but it also gave an error, which it shouldn't.

23 years agokbuild: sed compatibility fixes
Kai Germaschewski [Mon, 10 Feb 2003 15:02:39 +0000 (09:02 -0600)]
kbuild: sed compatibility fixes

Some versions of sed seem to think \w, as in word, doesn't include
digits, which breaks the build with CONFIG_MODVERSIONS. So we
just use the more compatible [<space><tab>]*.

23 years ago[PATCH] checker bounds/limits fixes
Randy Dunlap [Mon, 10 Feb 2003 08:39:58 +0000 (00:39 -0800)]
[PATCH] checker bounds/limits fixes

Part of the bounds checking bugs found by the Stanford checker.

The 4 fixes below have been acked by their maintainers.

23 years ago[PATCH] de4x5 compile fix
Andrew Morton [Mon, 10 Feb 2003 08:29:58 +0000 (00:29 -0800)]
[PATCH] de4x5 compile fix

Move the definition of version[] down to where __initdata has been defined.

23 years ago[PATCH] ncpfs compile fix
Andrew Morton [Mon, 10 Feb 2003 08:29:49 +0000 (00:29 -0800)]
[PATCH] ncpfs compile fix

Patch from Joel Becker <Joel.Becker@oracle.com>

The task_struct->sig -> task_struct->signal bits.

23 years ago[PATCH] Fix synchronous writers to wait properly for the result
Andrew Morton [Mon, 10 Feb 2003 08:29:41 +0000 (00:29 -0800)]
[PATCH] Fix synchronous writers to wait properly for the result

Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz> points out a bug in
ll_rw_block() usage.

Typical usage is:

mark_buffer_dirty(bh);
ll_rw_block(WRITE, 1, &bh);
wait_on_buffer(bh);

the problem is that if the buffer was locked on entry to this code sequence
(due to in-progress I/O), ll_rw_block() will not wait, and start new I/O.  So
this code will wait on the _old_ I/O, and will then continue execution,
leaving the buffer dirty.

It turns out that all callers were only writing one buffer, and they were all
waiting on that writeout.  So I added a new sync_dirty_buffer() function:

void sync_dirty_buffer(struct buffer_head *bh)
{
lock_buffer(bh);
if (test_clear_buffer_dirty(bh)) {
get_bh(bh);
bh->b_end_io = end_buffer_io_sync;
submit_bh(WRITE, bh);
} else {
unlock_buffer(bh);
}
}

which allowed a fair amount of code to be removed, while adding the desired
data-integrity guarantees.

UFS has its own wrappers around ll_rw_block() which got in the way, so this
operation was open-coded in that case.

23 years agoMerge tetrachloride.(none):/mnt/stuff/kernel/2.5/bk-linus
Dave Jones [Mon, 10 Feb 2003 18:43:40 +0000 (17:43 -0100)]
Merge tetrachloride.(none):/mnt/stuff/kernel/2.5/bk-linus
into tetrachloride.(none):/mnt/stuff/kernel/2.5/agpgart-respin

23 years ago[AGPGART] Enable support for VIA PLE133 chipset
Dave Jones [Mon, 10 Feb 2003 17:49:26 +0000 (16:49 -0100)]
[AGPGART] Enable support for VIA PLE133 chipset