After being put on hold for a while (needed fixes to CCISS driver, etc) I
attached a patch that adds pci-x support to the cpqphp driver in 2.5.65.

Changes:
	Adds 66/100/133MHz PCI-X support.
	Adds support for dynamically chaning frequency and mode.
	Changes code to also blink the power LED when powering down.
	Uses the IRQ from PCI config space if the ROM resource table doesn't
	provide one (would previously use the same interrupt as the hotplug
	controller which would lead to bad things when trying to update routing
	tables). Dan Zink should have the credit for this fix.
	Changes find_slot() to cpqhp_find_slot().
	Uses sysfs to display speed/freq.
	Some documentation updates.

Patch from Bob Miller <rem@osdl.org>

The get_disk() function should check the return value from kobject_get()
before passing it to to_disk().  This patch fixes this error.

(Acked by Pat)

Patch from Andreas Gruenbacher <agruen@suse.de>

In both 2.5 and 2.4, the fs/nfsd/vfs.c:nfsd_symlink() function calls down to
notify_change().  If notify_change fails for some reason, the error code is
not converted to an nfs no-the-wire error code as is should.  The attached
patches fix that (one for 2.4, the other for 2.5).

Patch from Martin J. Bligh and Dave Hansen

People with ordinary PCs are accidentally turning on NUMA support, and people
with NUMA machines are finding the NUMA option mysteriously disappearing.
This patch sets the defaults to sane things for everyone, and only allows you
to turn on NUMA with both SMP and 64Gb support on (it's useful for the
distros on non-Summit boxes, but not on their UP kernels ;-)).

I've also moved it below the highmem options, as it logically depends on
them, so this makes more sense.  For those searching for NUMA support on
*real* NUMA machine, Dave has provided some guiding comments to show them
what they messed up (it's totally non-obvious).  Hopefully this will stop
people's recent unfortunate foot-wounds (I think UP machines were defaulting
to NUMA on ...  oops).

This makes sysrq facilities available to remote users.

Writing a 'C' to /proc/sysrq-trigger receives the same treatment as typing
sysrq-C on the local keyboard.

Patch from Dave Hansen <haveblue@us.ibm.com>

Adds some commentary to these newly-introduced macros.

Patch from John Levon <levon@movementarian.org>

wli got an oops from this. The callbacks call mod_timer so the timer had
better be setup by then

Patch from "Martin J. Bligh" <mbligh@aracnet.com>

This removes the DO_ACTION stuff.  The downside is that we add some boring
and repetive code.  The upside is that it's simple, and mere mortals can read
it without screwing their brains into a small piece of silly putty and
bouncing it off the wall.  I think that's more important than pure source
code size.

Patch from "Randy.Dunlap" <rddunlap@osdl.org>

Adds an oops counter to the oops messages, such as:

	Oops: 0002 [#2]

So we can tell whether oops reports refer to the first oops, or to some
less-interesting followon oops.

Patch from george anzinger <george@mvista.com>

Fix the "large sleep returns EINVAL" problem, and clean a few things up.

Patch from Oleg Drokin <green@namesys.com>

    There is trivial memleak on error exit path in
    fs/ufs/util.c::_ubh_bread_()

Patch from Oleg Drokin <green@namesys.com>

   There is trivial memleak on error exit path in nfs get_sb function.

Patch from Christoph Hellwig <hch@lst.de>

It extends the maximum amount of memory which may be kmalloced on nommu
machines.  This is needed because these machines cannot perform vmalloc().

We couldn't really find a way of doing this which avoided the ifdef tangle.

Patch from Christoph Hellwig <hch@lst.de>

This is from the uClinux patches - there are a few more stubs needed
in nommu.c to get the mmuless plattforms working.

Patch from "Randy.Dunlap" <randy.dunlap@verizon.net>

This patch (to 2.5.64) reduces the stack usage in vt_ioctl()
from 0x334 bytes to 0xec bytes (P4, UP, gcc 2.96).

- optimise file_kill() to not take the global lock if the file is not on a
  list.

- Use optimised file_kill() in a few places rather than open-coding the
  list removal.

- Remove the private freelist.  There's no point in special-casing file
  structure allocations in this way.

- Hence the freeing of files can be moved outside file_list_lock()

- Replace euid test with capable(CAP_SYS_ADMIN).

- Tidy various other things up.

Replace the odd handling of f_list.next = NULL with list_emptiness.

release_mem() is altering the file->f_list lists without taking the
appropriate spinlock.

Patch from Manfred Spraul <manfred@colorfullife.com>

Fixes the lock contention over file_list_lock by simply removing the unneeded
anon_list.  So filp allocation does not need to take a global lock any more.

The use of a spinlock to protect files_stat.nr_files is a bit awkward, but
the alternative is a custom sysctl handler, and isn't much more efficient
anyway.

Patch from Andi Kleen <ak@muc.de>

copy_thread could leak memory if you had a io bitmap and passed wrong
arguments to the new clone flags.

Patch from Robert Love <rml@tech9.net>

There are a couple uses of 'p->state=foo' in mm/ which are open coded.
This patch converts them to the proper [__]set_current_state() function.

Patch from Robert Love <rml@tech9.net>

This patch is by Inaky Perez-Gonzalez.

There are a couple uses of 'p->state=foo' in fs/ which are open coded.
This patch converts them to the proper [__]set_current_state() function.

This is a forward-port of Andrea's fix in 2.4.

If a timer handler re-adds a timer to go off right now, __run_timers() will
never terminate.  (I wrote a test.  It happens.)

Fix that up by teaching internal_add_timer() to detect when it is being
called from within the context of __run_timers() and to park newly-added
timers onto a temp list instead.  These timers are then added for real by
__run_timers(), after it has finished processing all pending timers.

- Use list_head functions rather than open-coding them

- Use time comparison macros rather than open-coding them

- Hide some ifdefs

- uninline internal_add_timer().  Saves half a kilobyte of text.

Patch from: Hugh Dickins <hugh@veritas.com>

Removes a ton of dead code from ll_rw_blk.c.  I don't expect we'll be using
this now.

Patch from Anton Blanchard <anton@samba.org>

With ppc64 64bit dynamic executables we have to relocate the contents of the
function descriptor.  Passing in the load address to ELF_PLAT_INIT allows us
to do this.

The patch allows ppc64 to run 64-bit executables and is a no-op for other
architectures.

Patch from Zwane Mwaikambo <zwane@linuxpower.ca>

This patch fixes what seems to have been a longstanding bug. Ever since we
moved cpu bringup later into the boot process, we end up programming the
ioapics before we have any of our possible cpus in the cpu_online_map.
Therefore leading to the following current situation;

For walmart-smp, bigsmp and summit we set the logical destination for cpu
to TARGET_CPUS which can depend on the cpu_online_map, so what you would
normally see with noirqbalance would be all interrupts handled on cpu0
since at that stage no other cpu apart from the BSP is online.

You can check for this by looking at the ioredtbls at boottime for a two
way system;

.... IRQ redirection table:
 NR Log Phy Mask Trig IRR Pol Stat Dest Deli Vect:
 00 000 00  1    0    0   0   0    0    0    00
 01 001 01  0    0    0   0   0    1    1    39
 02 001 01  0    0    0   0   0    1    1    31
 03 001 01  0    0    0   0   0    1    1    41
 04 001 01  0    0    0   0   0    1    1    49
 05 001 01  0    0    0   0   0    1    1    51
 06 001 01  0    0    0   0   0    1    1    59

Notice that 'Log' is set to 1 instead of 3.

This patch will simply reprogram all the ioredtbls to handle the other
online cpus.

Patch tested on my 2way P2-400 and a 16way NUMAQ both with noirqbalance.
It will not affect the irqbalance case because we are simply setting
TARGET_CPUS which is done anyway.

before:
  CPU0       CPU1
  0:    1495632          0    IO-APIC-edge  timer
  1:       4270          0    IO-APIC-edge  i8042
  2:          0          0          XT-PIC  cascade
  8:          1          0    IO-APIC-edge  rtc
 12:      83592          0    IO-APIC-edge  i8042
 14:      93791          0    IO-APIC-edge  ide0
 15:     103167          0    IO-APIC-edge  ide1
 17:    1396088          0   IO-APIC-level  EMU10K1, eth0
 18:      56125          0   IO-APIC-level  aic7xxx, aic7xxx
 19:       2258          0   IO-APIC-level  uhci-hcd, eth1, serial
NMI:          0          0
LOC:    1495566    1497133

after:
           CPU0       CPU1
  0:    1046157    1015670    IO-APIC-edge  timer
  1:       4923       4173    IO-APIC-edge  i8042
  2:          0          0          XT-PIC  cascade
  8:          1          0    IO-APIC-edge  rtc
 12:      48596      48968    IO-APIC-edge  i8042
 14:       4238       3416    IO-APIC-edge  ide0
 15:      25362      31525    IO-APIC-edge  ide1
 17:       3757       4014   IO-APIC-level  EMU10K1, eth0
 18:        335        366   IO-APIC-level  aic7xxx, aic7xxx
 19:       1052        908   IO-APIC-level  uhci-hcd, eth1
NMI:          0          0
LOC:    2061856    2061893

PA-RISC patches for 2.5.65:

 - Only remove palo.conf on a `make mrproper'
 - Add a \ continuation to kernel/Makefile
 - Reindent cache.c
 - Always call schedule_tail
 - Fix some typos in the hardware database
 - Signal handling changes
 - RAID, Device Mapper & BLKSZGET ioctl32 translations
 - Clean up pci host bridge handling a bit.
 - Make IDE link again
 - Fix up compat changes
 - Improve stack dumping code

into kroah.com:/home/greg/linux/BK/gregkh-2.5

This is a fix made almost a month ago, during the flurry of signal changes.
I didn't realize until today that this hadn't made it into 2.5.  Sorry
about the delay.

This fix is necessary to avoid sometimes wedging in uninterruptible sleep
when doing a multithreaded core dump triggered by a process signal (kill)
rather than a trap.  You can reproduce the problem by running your favorite
multithreaded program (NPTL) and then using "kill -SEGV" on it.  It will
often wedge.  The actual fix could be just a two line diff:

+                       if (current->signal->group_exit)
+                               goto dequeue;

after the group_exit_task check.  That is the fix that has been used in
Ingo's backport for weeks and tested heavily (well, as heavily as core
dumping ever gets tested, but it's been in our production systems).

But I broke the hair out into a separate function.  The patch below has the
same effect as the two-liner, and no other difference.  I have tested
2.5.64 with this patch and it works for me, though I haven't beat on it.

The way the wedge happens is that for a core-dump signal group_send_sig_info
does a group stop of other threads before the one thread handles the fatal
signal.  If the fatal thread gets into do_coredump and coredump_wait first,
then other threads see the group stop and suspend with SIGKILL pending.
All other fatal cases clear group_stop_count, so this is the only way this
ever happens.  Checking group_exit fixes it.  I didn't make do_coredump
clear group_stop_count because doing it with the appropriate ordering and
locking doesn't fit the organization that code.

Mounting a non-affs filesystem as affs crashes the kernel.
The reason is the
	sbi = kmalloc(sizeof(struct affs_sb_info), GFP_KERNEL);
	memset(sbi, 0, sizeof(*AFFS_SB));
where the second sizeof is 1, so that sbi is not zeroed at all.

Also, avoid a warning for printk of sector_t in amigaffs.h.

Previously, we only honored barriers wrt merging. Really honor them now,
move all pending requests to dispatch list and add the hard barrier at
the back.

gen-asm-offsets, the most common user of the new filechk function, needs
to be fed input from $< (the first prerequisite).

into kroah.com:/home/linux/linux/BK/gregkh-2.5

Patch from Adrian Bunk <bunk@fs.tum.de>

It would be nice if everyone would try to compile the patched files
before submitting patches...

raid0 doesn't have a thread, so md_wakeup_thread() derefs NULL.

Neil may end up doing this differently, but meanwhile....