[PATCH] stop_machine() vs. synchronous IPI send deadlock

This fixes deadlock of stop_machine() vs.  synchronous IPI send.  The
problem is that stop_machine() disables interrupts before disabling
preemption on other CPUs.  So if another CPU is preempted and then calls
something like flush_tlb_all() it will deadlock with CPU doing
stop_machine() and which can't process IPI due to disabled IRQs.

I changed stop_machine() to do the same things exactly as it does on other
CPUs, i.e.  it should disable preemption first on _all_ CPUs including
itself and only after that disable IRQs.

Signed-off-by: Kirill Korotaev <dev@sw.ru>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: "Andrey Savochkin" <saw@sawoct.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] arch/mips/au1000/common/usbdev.c: don't concatenate __FUNCTION__ with strings

It's deprecated. Use "%s", __FUNCTION__ instead.

Signed-off-by: Clemens Buchacher <drizzd@aon.at>
Signed-off-by: Maximilian Attems <janitor@sternwelten.at>
Signed-off-by: Domen Puncer <domen@coderock.org>
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] atomic: inc_not_zero

Introduce an atomic_inc_not_zero operation.  Make this a special case of
atomic_add_unless because lockless pagecache actually wants
atomic_inc_not_negativeone due to its offset refcount.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] atomic: cmpxchg

Introduce an atomic_cmpxchg operation.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] i386: generic cmpxchg

- Make cmpxchg generally available on the i386 platform.

- Provide emulation of cmpxchg suitable for uniprocessor if built and run on
  386.

From: Christoph Lameter <clameter@sgi.com>

- Cut down patch and small style changes.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] rcutorture: renice to low priority

Make the box usable for interactive work when running the RCU torture test,
by renicing the RCU torture-test threads to +19 by default.  Kthreads run
at nice -5 by default.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Shut up per_cpu_ptr() on UP

Currently per_cpu_ptr() doesn't really do anything with 'cpu' in the UP
case.  This is problematic in the cases where this is the only place the
variable is referenced:

  CC      kernel/workqueue.o
  kernel/workqueue.c: In function `current_is_keventd':
  kernel/workqueue.c:460: warning: unused variable `cpu'

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Fix sparse warning in proc/task_mmu.c

fs/proc/task_mmu.c:198:33: warning: Using plain integer as NULL pointer

Signed-off-by: Luiz Capitulino <lcapitulino@mandriva.com.br>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] synclink: update to use DMA mapping API

Update synclink to use DMA mapping API.  This removes warning about
isa_virt_to_bus() usage on architectures other than i386

Signed-off-by: Paul Fulghum <paulkf@microgate.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ext3: journal handling on error path in ext3_journalled_writepage()

This patch fixes lost referrence on ext3 current handle in
ext3_journalled_writepage().

Signed-Off-By: Denis Lunev <den@sw.ru>
Cc: Kirill Korotaev <dev@sw.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] signal handling: revert sigkill priority fix

This patch reverts commit c33880aaddbbab1ccf36f4457ed1090621f2e39a since
it's not needed anymore. As pointed out by Roland McGrath the real fix
is to deliver all signals before returning to user space.
See http://www.ussg.iu.edu/hypermail/linux/kernel/0509.2/0683.html
A fix for s390 has been merged.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Linus Torvalds <torvalds@osdl.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml: fix daemon transport exit path bug

Fix some exit path bugs in the daemon driver.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml: fix access_ok

The access_ok_tt() macro is bogus, in that a read access is unconditionally
considered valid.

I couldn't find in SCM logs the introduction of this check, but I went back to
2.4.20-1um and the definition was the same.

Possibly this was done to avoid problems with missing set_fs() calls, but
there can't be any I think because they would fail with SKAS mode.
TT-specific code is still to check.

Also, this patch joins common code together, and makes the "address range
wrapping" check happen for all cases, rather than for only some.

This may, possibly, be reoptimized at some time, but the current code doesn't
seem clever, just confused.

* Important: I've also had to change references to access_ok_{tt,skas} back to
  access_ok - the kernel wasn't that happy otherwise.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml console channels: fix the API of console_write

Since the 4th param is unused, remove it altogether.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml console channels: remove console_write wrappers

We were using a long series of (stupid) wrappers which all call
generic_console_write().  Since the wrappers only change the 4th param, which
is unused by the called proc, remove them and call generic_console_write()
directly.

If needed at any time in the future to reintroduce this stuff, the member
could be moved to a generic struct, to avoid this duplicated handling.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml: fix mcast network driver error handling

printk clears the host errno (I verified this in debugging and it's reasonable
enough, given that it ends via a write call on some fd, especially since
printk() goes on /dev/tty0 which is often the host stdout).  So save errno
earlier.  There's no reason to change the printk calls to use -err rather than
errno - the assignment can't clear errno.

And in the first failure path, we used to return 0 too (and this time more
clearly), which is totally wrong.  0 is a success fd, which is then registered
and gives a "registering fd twice" warning.

Finally, fix up some whitespace.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml: fixups for "reuse i386 cpu-specific tuning"

A few fixups - show the new submenu only for x86 subarchitecture (it does not
make sense to show it for x86_64 users) and remove X86_CMPXCHG, which is now a
duplicate of Kconfig.i386, even though Kconfig doesn't complain (we also miss
the dependency on !M386 CPU).

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml: micro fixups to arch Kconfig

Remove a stone-age comment (UM *does* have a MMU, i.e.  the host), and fix a
dependency (introduced in commit 02edeb586ae4cdd17778923674700edb732a4741) to
do what was intended.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] uml: remove bogus WARN_ON, triggerable harmlessly on a page fault race

The below warning was added in place of pte_mkyoung(); if (is_write)
pte_mkdirty();

In fact, if the PTE is not marked young/dirty, our dirty/accessed bit
emulation would cause the TLB permission not to be changed, and so we'd loop,
and given we don't support preemption yet, we'd busy-hang here.

However, I've seen this warning trigger without crashes during a loop of
concurrent kernel builds, at random times (i.e. like a race condition), and I
realized that two concurrent faults on the same page, one on read and one on
write, can trigger it. The read fault gets serviced and the PTE gets marked
writable but clean (it's possible on a shared-writable mapping), while the
generic code sees the PTE was already installed and returns without action. In
this case, we'll see another fault and service it normally.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Kbuild: index asm-$(SUBARCH) headers for UML

In Uml, many definitions are borrowed from underlying subarch headers (with
#include <asm/arch/stuff.h>).  And it has become annoying to keep switching
tag files all time, so by default index the underlying subarch headers too.
Btw, it adds negligible space to the tags file (less than 1M surely, IIRC it
was around 500k over 40M).

Finally, preserve the ALLSOURCE_ARCHS command line option (I hope) - if it is
set, it is used for headers too as before.  But check my construct please, I
didn't test this.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Acked-by: Jeff Dike <jdike@addtoit.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] m68k: convert thread flags to use bit fields

Remove task_work structure, use the standard thread flags functions and use
shifts in entry.S to test the thread flags.  Add a few local labels to entry.S
to allow gas to generate short jumps.

Finally it changes a number of inline functions in thread_info.h to macros to
delay the current_thread_info() usage, which requires on m68k a structure
(task_struct) not yet defined at this point.

Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Cc: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] m68k: m68k-specific thread_info changes

a) added embedded thread_info [m68k processor.h]

b) added missing symbols in asm-offsets.c

c) task_thread_info() and friends in asm-m68k/thread_info.h

d) made m68k thread_info.h included by m68k processor.h, not the other way
   round.

Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] m68k: thread_info header cleanup

a) in smp_lock.h #include of sched.h and spinlock.h moved under #ifdef
   CONFIG_LOCK_KERNEL.

b) interrupt.h now explicitly pulls sched.h (not via smp_lock.h from
   hardirq.h as it used to)

c) in three more places we need changes to compensate for (a) - one place
   in arch/sparc needs string.h now, hardirq.h needs forward declaration of
   task_struct and preempt.h needs direct include of thread_info.h.

d) thread_info-related helpers in sched.h and thread_info.h put under
   ifndef __HAVE_THREAD_FUNCTIONS.  Obviously safe.

Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] m68k: introduce setup_thread_stack() and end_of_stack()

encapsulates the rest of arch-dependent operations with thread_info access.
Two new helpers - setup_thread_stack() and end_of_stack().  For normal case
the former consists of copying thread_info of parent to new thread_info and
the latter returns pointer immediately past the end of thread_info.

Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] m68k: introduce task_thread_info

new helper - task_thread_info(task).  On platforms that have thread_info
allocated separately (i.e.  in default case) it simply returns
task->thread_info.  m68k wants (and for good reasons) to embed its thread_info
into task_struct.  So it will (in later patch) have task_thread_info() of its
own.  For now we just add a macro for generic case and convert existing
instances of its body in core kernel to uses of new macro.  Obviously safe -
all normal architectures get the same preprocessor output they used to get.

Signed-off-by: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>
Signed-off-by: Roman Zippel <zippel@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86: fix cpu_khz with clock=pit

Fix http://bugzilla.kernel.org/show_bug.cgi?id=5546

The cpu_khz global is not initialized and remains 0 if you boot with
clock=pit, even if the processor does have a TSC.  This may have bad
ramifications since the variable is used in various places scattered around
the kernel, though I didn't check them all to see if they can tolerate cpu_khz
= 0.  You can observe the problem by doing "cat /proc/cpuinfo"; the cpu MHz
line says 0.000.

The fix is trivial; call init_cpu_khz() from init_pit(), just as it's called
from the timers/timer_foo.c:init_foo() for other values of foo.

Cc: john stultz <johnstul@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] i386: NMI pointer comparison fix

Instruction pointer comparisons for the NMI on debug stack check/fixup
were incorrect.

From: Jan Beulich <jbeulich@novell.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Zwane Mwaikambo <zwane@holomorphy.com>
Acked-by: "Seth, Rohit" <rohit.seth@intel.com>
Cc: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] arch/i386/mm/init.c: small cleanups

This patch contains the following cleanups:
- make a needlessly global function static
- every file should include the headers containing the prototypes for
  it's global functions

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] powerpc-xmon-build-fix

arch/powerpc/xmon/xmon.c:525: error: syntax error before "xmon_irq"
arch/powerpc/xmon/xmon.c:526: warning: return type defaults to `int'
arch/powerpc/xmon/xmon.c: In function `xmon_irq':
arch/powerpc/xmon/xmon.c:532: error: `IRQ_HANDLED' undeclared (first use in this function)
arch/powerpc/xmon/xmon.c:532: error: (Each undeclared identifier is reported only once
arch/powerpc/xmon/xmon.c:532: error: for each function it appears in.)

Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc32: Add support for handling PCI interrupts on MPC834x PCI expansion card

The MPC8349 PIBs system has a expansion board with 6 PCI slots.  We needed
to update the IDSEL interrupt mapping for it to work properly.  However,
only PCI1 is supported as the first revision of this expansion board
doesn't function properly for PCI2.  For the time being we have zero'd out
the entries for the IDSELs related to PCI2.  When a functioning expansion
board exists we can fix the table.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ppc: add support for new powerbooks

Enablement patch for the new PowerBooks (late 2005 edition).

This enables the ATA controller, Gigabit ethernet and basic AGP setup.
Bluetooth works out-of-the box after running hid2hci.

Still remaining is to get the touchpad to work, the simple change of just
adding the new USB ids isn't enough.

Signed-off-by: Olof Johansson <olof@lixom.net>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] slab: remove alloc_pages() calls

The slab allocator never uses alloc_pages since kmem_getpages() is always
called with a valid nodeid.  Remove the branch and the code from
kmem_getpages()

Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] slab: convert cache to page mapping macros

This patch converts object cache <-> page mapping macros to static inline
functions to make the more explicit and readable.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Manfred Spraul <manfred@colorfullife.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] mm: highmem watermarks

The pages_high - pages_low and pages_low - pages_min deltas are the asynch
reclaim watermarks.  As such, the should be in the same ratios as any other
zone for highmem zones.  It is the pages_min - 0 delta which is the
PF_MEMALLOC reserve, and this is the region that isn't very useful for
highmem.

This patch ensures highmem systems have similar characteristics as non highmem
ones with the same amount of memory, and also that highmem zones get similar
reclaim pressures to other zones.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] mm: gfp_noreclaim cleanup

Remove last remnant of the defunct early reclaim page logic, the no longer
used __GFP_NORECLAIM flag bit.

Signed-off-by: Paul Jackson <pj@sgi.com>
Acked-by: Martin Hicks <mort@bork.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] mm: __alloc_pages cleanup

Clean up of __alloc_pages.

Restoration of previous behaviour, plus further cleanups by introducing an
'alloc_flags', removing the last of should_reclaim_zone.

Signed-off-by: Rohit Seth <rohit.seth@intel.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] mm: ZAP_BLOCK causes redundant work

The address based work estimate for unmapping (for lockbreak) is and always
was horribly inefficient for sparse mappings.  The problem is most simply
explained with an example:

If we find a pgd is clear, we still have to call into unmap_page_range
PGDIR_SIZE / ZAP_BLOCK_SIZE times, each time checking the clear pgd, in
order to progress the working address to the next pgd.

The fundamental way to solve the problem is to keep track of the end
address we've processed and pass it back to the higher layers.

From: Nick Piggin <npiggin@suse.de>

  Modification to completely get away from address based work estimate
  and instead use an abstract count, with a very small cost for empty
  entries as opposed to present pages.

  On 2.6.14-git2, ppc64, and CONFIG_PREEMPT=y, mapping and unmapping 1TB
  of virtual address space takes 1.69s; with the following patch applied,
  this operation can be done 1000 times in less than 0.01s

From: Andrew Morton <akpm@osdl.org>

With CONFIG_HUTETLB_PAGE=n:

mm/memory.c: In function `unmap_vmas':
mm/memory.c:779: warning: division by zero

Due to

zap_work -= (end - start) /
(HPAGE_SIZE / PAGE_SIZE);

So make the dummy HPAGE_SIZE non-zero

Signed-off-by: Robin Holt <holt@sgi.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] mm: __GFP_NOFAIL fix

In __alloc_pages():

if ((p->flags & (PF_MEMALLOC | PF_MEMDIE)) && !in_interrupt()) {
         /* go through the zonelist yet again, ignoring mins */
         for (i = 0; zones[i] != NULL; i++) {
                 struct zone *z = zones[i];

                 page = buffered_rmqueue(z, order, gfp_mask);
                 if (page) {
                         zone_statistics(zonelist, z);
                         goto got_pg;
                 }
         }
         goto nopage;                <<<< HERE!!! FAIL...
}

kswapd (which has PF_MEMALLOC flag) can fail to allocate memory even when
it allocates it with __GFP_NOFAIL flag.

Signed-Off-By: Pavel Emelianov <xemul@sw.ru>
Signed-Off-By: Denis Lunev <den@sw.ru>
Signed-Off-By: Kirill Korotaev <dev@sw.ru>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] shpchp_hpc build fix

Missing include.

Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] pciehp_hpc build fix

drivers/pci/hotplug/pciehp_hpc.c:221: parse error before "pcie_isr"
drivers/pci/hotplug/pciehp_hpc.c:221: warning: type defaults to `int' in declaration of `pcie_isr'
drivers/pci/hotplug/pciehp_hpc.c:221: warning: data definition has no type or storage class
drivers/pci/hotplug/pciehp_hpc.c: In function `hpc_release_ctlr':
drivers/pci/hotplug/pciehp_hpc.c:715: implicit declaration of function `free_irq'
drivers/pci/hotplug/pciehp_hpc.c: At top level:
drivers/pci/hotplug/pciehp_hpc.c:839: parse error before "pcie_isr"
drivers/pci/hotplug/pciehp_hpc.c:840: warning: return type defaults to `int'
drivers/pci/hotplug/pciehp_hpc.c: In function `pcie_isr':
drivers/pci/hotplug/pciehp_hpc.c:850: `IRQ_NONE' undeclared (first use in this function)
drivers/pci/hotplug/pciehp_hpc.c:850: (Each undeclared identifier is reported only once
drivers/pci/hotplug/pciehp_hpc.c:850: for each function it appears in.)
drivers/pci/hotplug/pciehp_hpc.c:979: `IRQ_HANDLED' undeclared (first use in this function)
drivers/pci/hotplug/pciehp_hpc.c: In function `pcie_init':
drivers/pci/hotplug/pciehp_hpc.c:1362: implicit declaration of function `request_irq'

Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] w100fb: platform device conversion fixup

Fix an error in w100fb after the platform device conversion.

Signed-off-by: Richard Purdie <rpurdie@rpsys.net>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] packet writing oops fix

There is an old bug in the pkt_count_states() function that causes stack
corruption.  When compiling with gcc 3.x or 2.x it is harmless, but gcc 4
allocates local variables differently, which makes the bug visible.

Signed-off-by: Peter Osterlund <petero2@telia.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] cpuset: fix return without releasing semaphore

It is wrong to acquire the semaphore and then return from
cpuset_zone_allowed without releasing it.

Signed-off-by: Bob Picco <bob.picco@hp.com>
Acked-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nvidiafb: Fix bug in nvidiafb_pan_display

nvidiafb_pan_display() is incorrectly using the fields in info->var instead
of var passed to the function.

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] rpaphp_pci build fix

(akpm: _machine is some ppc64 thing - this is a powerpc-only driver)

Signed-off-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fix task_struct leak in ptrace

When ptrace_attach fails we need to drop the task_struct reference.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fbdev: fix module dependency loop

Exporting struct fb_display produces this warning error on depmod:

WARNING: Module
/lib/modules/2.6.14-mm2/kernel/drivers/video/console/fbcon_ud.ko
ignored, due to loop
WARNING: Module
/lib/modules/2.6.14-mm2/kernel/drivers/video/console/fbcon_rotate.ko
ignored, due to loop
WARNING: Module
/lib/modules/2.6.14-mm2/kernel/drivers/video/console/fbcon_cw.ko
ignored, due to loop
WARNING: Module
/lib/modules/2.6.14-mm2/kernel/drivers/video/console/fbcon_ccw.ko
ignored, due to loop
WARNING: Module
/lib/modules/2.6.14-mm2/kernel/drivers/video/console/fbcon.ko ignored,
due to loop
WARNING: Loop detected:
/lib/modules/2.6.14-mm2/kernel/drivers/video/console/bitblit.ko needs

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] nv_of.c build fix

drivers/video/nvidia/nv_of.c:33: error: redefinition of `nvidia_probe_of_connector'
drivers/video/nvidia/nv_proto.h:51: error: `nvidia_probe_of_connector' previously defined here

Because the inline version depends on !CONFIG_FB_OF and the out-of-line
version depends on CONFIG_PPC_OF.

Ben said: "Yes, CONFIG_PPC_OF is the right one, must be a typo."

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
C: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Update email address for Kumar

Changed jobs and the Freescale address is no longer valid.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] New Omnikey Cardman 4000 driver

Add new Omnikey Cardman 4000 smartcard reader driver

Signed-off-by: Harald Welte <laforge@gnumonks.org>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] New Omnikey Cardman 4040 driver

Add new Omnikey Cardman 4040 smartcard reader driver

Signed-off-by: Harald Welte <laforge@gnumonks.org>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] move pm_register/etc. to CONFIG_PM_LEGACY, pm_legacy.h

Since few people need the support anymore, this moves the legacy
pm_xxx functions to CONFIG_PM_LEGACY, and include/linux/pm_legacy.h.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] reorder struct files_struct

The file_lock spinlock sits close to mostly read fields of 'struct
files_struct'

In SMP (and NUMA) environments, each time a thread wants to open or close
a file, it has to acquire the spinlock, thus invalidating the cache line
containing this spinlock on other CPUS.  So other threads doing
read()/write()/...  calls that use RCU to access the file table are going
to ask further memory (possibly NUMA) transactions to read again this
memory line.

Move the spinlock to another cache line, so that concurrent threads can
share the cache line containing 'count' and 'fdt' fields.

It's worth up to 9% on a microbenchmark using a 4-thread 2-package x86
machine.  See
http://marc.theaimsgroup.com/?l=linux-kernel&m=112680448713342&w=2

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86_64 two timer entries in /sys

attached patch renames one instance of
/sys/devices/system/timer
to
/sys/devices/system/timer_pit
to avoid a name clash with another instance created in time.c.

Acked-by: Andi Kleen <ak@muc.de>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86_64: fix tss limit

Fix the x86_64 TSS limit in TSS descriptor.

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge branch 'upstream-fixes' of /linux/kernel/git/jgarzik/libata-dev

[libata passthru] address slave devices correctly

[PATCH] libata: fix comments on ata_tf_from_fis()

Fix description on comments for ata_tf_from_fis().

Signed-off-by: Mark Lord <mlord@pobox.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

Merge /pub/scm/linux/kernel/git/tglx/mtd-2.6

[JFFS2] Remove broken and useless debug code

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

[MTD] maps: Replace dependency on non existing config option

CONFIG_ELAN doesn't exist any more; CONFIG_X86_ELAN is too specific
so make ts-5500 memory map dependant on CONFIG_X86.

Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

[PATCH] VFS: local denial-of-service with file leases

 Remove time_out_leases() printk that's easily triggered by users.

Signed-off-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

[PATCH] VFS: Fix memory leak with file leases

 The patch
 http://linux.bkbits.net:8080/linux-2.6/diffs/fs/locks.c@1.70??nav=index.html
 introduced a pretty nasty memory leak in the lease code. When freeing
 the lease, the code in locks_delete_lock() will correctly clean up
 the fasync queue, but when we return to fcntl_setlease(), the freed
 fasync entry will be reinstated.

 This patch ensures that we skip the call to fasync_helper() when we're
 freeing up the lease.

Signed-off-by: J. Bruce Fields <bfields@fieldses.org>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

[PATCH] sil24: add missing ata_pad_free()

sil24_port_stop() is missing call to ata_pad_free() thus leaking pad
buffer when a port is stopped.  This patch adds it.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

[IPV6]: Fix unnecessary GFP_ATOMIC allocation in fib6 dump

Signed-off-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER] {ip,nf}_conntrack TCP: Accept SYN+PUSH like SYN

Some devices (e.g. Qlogic iSCSI HBA hardware like QLA4010 up to firmware
3.0.0.4) initiates TCP with SYN and PUSH flags set.

The Linux TCP/IP stack deals fine with that, but the connection tracking
code doesn't.

This patch alters TCP connection tracking to accept SYN+PUSH as a valid
flag combination.

Signed-off-by: Vlad Drukker <vlad@storewiz.com>
Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Fix rtnetlink dump infinite loop

The recent change to netlink dump "done" callback handling broke IPv6
which played dirty tricks with the "done" callback.  This causes an
infinite loop during a dump.

The following patch fixes it.

This bug was reported by Jeff Garzik.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SBUSFB]: implement ->compat_ioctl

This patch adds a new function, sbusfb_compat_ioctl() to
drivers/video/sbuslib.c and uses it as compat_ioctl in all sbus fb
drivers

This remove the last per-arch compat ioctl bits in
arch/sparc64/kernel/ioctl32.c so it would be nice if people could test
if this actually copiles and works and if yes apply it :)

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: Fix RTC compat ioctl kernel log spam.

On Fri, Nov 11, 2005 at 12:58:40PM -0800, David S. Miller wrote:
>
> This change:
>
> diff-tree 8ca2bdc7a98b9584ac5f640761501405154171c7 (from feee207e44d3643d19e648aAuthor: Christoph Hellwig <hch@lst.de>
> Date:   Wed Nov 9 12:07:18 2005 -0800
>
>     [SPARC] sbus rtc: implement ->compat_ioctl
>
>     Signed-off-by: Christoph Hellwig <hch@lst.de>
>     Signed-off-by: David S. Miller <davem@davemloft.net>
>
> results in the console now getting spewed on sparc64 systems
> with messages like:
>
> [   11.968298] ioctl32(hwclock:464): Unknown cmd fd(3) cmd(401c7014){00} arg(efc
> What's happening is hwclock tries first the SBUS rtc device ioctls
> then the normal rtc driver ones.
>
> So things actually worked better when we had the SBUS rtc compat ioctl
> directly handled via the generic compat ioctl code.
>
> There are _so_ many rtc drivers in the kernel implementing the
> generic rtc ioctls that I don't think putting a ->compat_ioctl
> into all of them to fix this problem is feasible.  Unless we
> write a single rtc_compat_ioctl(), export it to modules, and hook
> it into all of those somehow.
>
> But even that doesn't appear to have any pretty implementation.
>
> Any better ideas?

We had similar problems with other ioctls where userspace did things
like that.  What we did there was to put the compat handler to generic
code.  The patch below does that, adding a big comment about what's
going on and removing the COMPAT_IOCTL entires for these on powerpc
that not only weren't ever useful but are duplicated now aswell.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] I8K: fix /proc reporting of blank service tags

Make /proc/i8k display '?' when service tag is blank in BIOS.
This fixes segfault in i8k gkrellm plugin.

Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] fix b2c2 dvb undefined symbol

This fixes

  drivers/built-in.o: In function `flexcop_frontend_init':
  : undefined reference to `lgdt330x_attach'

[ Side note: I really dislike that dvb people want to include every
  possible frontend into the kernel - I only need the mt312 one for my
  Skystar2 card.  I'd highly appreciate it this would be made selectable
  again... ]

Signed-off-by: Prakash Punnoor <prakash@punnoor.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge branch 'upstream-fixes' of /linux/kernel/git/jgarzik/libata-dev

Revert "[PATCH] fbcon: Add rl (Roman Large) font"

This reverts 998e6d51162707685336ff99c029c8911b270d32 commit.

[BLOCK] elevator: elv_latter/former_request update

With generic dispatch queue update, implicit former/latter request
handling using rq->queuelist.prev/next doesn't work as expected
anymore.  Also, the only iosched dependent on this feature was
noop-iosched and it has been reimplemented to have its own
latter/former methods.  This patch removes implicit former/latter
handling.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] noop-iosched: reimplementation of request dispatching

The original implementation directly used dispatch queue.  As new
generic dispatch queue imposes stricter rules over ioscheds and
dispatch queue usage, this direct use becomes somewhat problematic.
This patch reimplements noop-iosched such that it complies to generic
iosched model better.  Request merging with q->last_merge and
rq->queuelist.prev/next work again now.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de

[BLOCK] cfq-iosched: fix slice_left calculation

When cfq slice expires, remainder of slice is calculated and stored in
cfqq->slice_left.  Current code calculates the opposite of remainder -
how many jiffies the cfqq has used past slice end.  This patch fixes
the bug.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] fix string handling in elv_iosched_store

elv_iosched_store doesn't terminate string passed from userspace if
it's too long.  Also, if the written length is zero (probably not
possible), it accesses elevator_name[-1].  This patch fixes both bugs.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] Implement elv_drain_elevator for improved switch error detection

This patch adds request_queue->nr_sorted which keeps the number of
requests in the iosched and implement elv_drain_elevator which
performs forced dispatching.  elv_drain_elevator checks whether
iosched actually dispatches all requests it has and prints error
message if it doesn't.  As buggy forced dispatching can result in
wrong barrier operations, I think this extra check is worthwhile.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] cfq-iosched: cfq forced dispatching fix

cfq forced dispatching might not return all requests on the queue.
This bug can hang elevator switchinig and corrupt request ordering
during flush sequence.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] elevator: run queue in elevator_switch

elevator_dispatch needs to run queue after forced dispatching;
otherwise, the queue might stall.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] Document the READ/WRITE splitup of the disk stats

Use the symbolic name where appropriate and add a comment to the
disk_stats structure.

Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] elevator init fixes #2

In addition to the first patch, which is probably goodness, I found the
cause of my panic - applying this patch fixes it and now I am booting.
If the chosen_elevator[] is not found, fall back to noop.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jens Axboe <axboe@suse.de>

[BLOCK] elevator init fixes

I got a panic in the elevator code, backtrace :

Unable to handle kernel NULL pointer dereference at virtual address 00000060
..
EIP is at elevator_put+0x0/0x30 (null elevator_type passed)
..
elevator_init+0x38
blk_init_queu_node+0xc9
floppy_init+0xdb
do_initcalls+0x23
init+0x10a
init+0x0

Clearly if the kmalloc here fails, e->elevator_type is not yet set; this
appears to be the correct fix, but I think I probably hit the second case
due to a race condition.  Someone more familiar with the elevator code
should look at this more closely until I can determine if I can reproduce.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Jens Axboe <axboe@suse.de>

[libata ahci] set port ATAPI bit correctly

Although according to the documentation this largely only affects
desktop LED control, let's make sure we set the ATAPI bit when we
have an ATAPI device attached to the port.

[libata ahci, qstor] fix miscount of scatter/gather entries

Don't directly reference qc->n_elem, as that might cause an off-by-one
error for misaligned (padded) ATAPI transfers.

Linux v2.6.15-rc1

As per the new release rules: two weeks of merging, and then an -rc1 and
calming down for the next release.

Merge /pub/scm/linux/kernel/git/davem/net-2.6

Merge master.kernel.org:/home/rmk/linux-2.6-arm

Merge branch 'release' of git://git./linux/kernel/git/aegl/linux-2.6

Merge git://git./linux/kernel/git/bunk/trivial

[SCTP]: Include ulpevents in socket receive buffer accounting.

Also introduces a sysctl option to configure the receive buffer
accounting policy to be either at socket or association level.
Default is all the associations on the same socket share the
receive buffer.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Fix ia64 NaT consumption fault with sctp_sideffect commands.

On ia64, it is possible to get NaT Consumption Fault and a kernel panic
when initializing sctp sideeffect commands arguments.  The union
sctp_arg_t contains different sized elements and when loading a smaller
sized element (32 or 16 bits), it is possible for a speculative load to
fail and result in a NaT bit set which causes a kernel crash.  The easy
way to get around it is to load the largerst member of the union.

Signed-off-by: Vladislav Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Remove timeouts[] array from sctp_endpoint.

The socket level timeout values are maintained in sctp_sock and
association level timeouts are in sctp_association. So there is
no need for ep->timeouts.

Signed-off-by: Vladislav Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SCTP]: Fix potential NULL pointer dereference in sctp_v4_get_saddr

It is possible to get to sctp_v4_get_saddr() without a valid
association.  This happens when processing OOTB packets and
the cached route entry is no longer valid.
However, when responding to OOTB packets we already properly
set the source address based on the information in the OOTB
packet.  So, if we we get to sctp_v4_get_saddr() without an
association we can simply return.

Signed-off-by: Vladislav Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: Sridhar Samudrala <sri@us.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[IPV6]: Fix inet6_init missing unregister.

Based mostly upon a patch from Olaf Kirch <okir@suse.de>

When initialization fails in inet6_init(), we should
unregister the PF_INET6 socket ops.

Also, check sock_register()'s return value for errors.

Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] kill libata scsi_wait_req usage (make libata compile with scsi-misc changes)

scsi_wait_req does not exist any more in the SCSI layer.  This patch
makes it so libata can compile again.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

ppc64: default build as the merged 'powerpc' architecture

After the last merge of the new unified 'powerpc' architecture, ppc64 no
longer compiles cleanly as a standalone architecture.  Some bits and
pieces still exist as files under the old ppc64 hierarchy, but the old
"ARCH=ppc64" is dead.

So if "uname" says ppc64, that now implies that the default architecture
should be "powerpc".

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge /linux/kernel/git/jejb/scsi-for-linus-2.6

Merge /pub/scm/linux/kernel/git/davem/net-2.6