sched: fix __set_task_cpu() SMP race

Grant Wilson has reported rare SCHED_FAIR_USER crashes on his quad-core
system, which crashes can only be explained via runqueue corruption.

there is a narrow SMP race in __set_task_cpu(): after ->cpu is set up to
a new value, task_rq_lock(p, ...) can be successfuly executed on another
CPU. We must ensure that updates of per-task data have been completed by
this moment.

this bug has been hiding in the Linux scheduler for an eternity (we never
had any explicit barrier for task->cpu in set_task_cpu() - so the bug was
introduced in 2.5.1), but only became visible via set_task_cfs_rq() being
accidentally put after the task->cpu update. It also probably needs a
sufficiently out-of-order CPU to trigger.

Reported-by: Grant Wilson <grant.wilson@zen.co.uk>
Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

sched: fix SCHED_FIFO tasks & FAIR_GROUP_SCHED

Suppose that the SCHED_FIFO task does

switch_uid(new_user);

Now, p->se.cfs_rq and p->se.parent both point into the old
user_struct->tg because sched_move_task() doesn't call set_task_cfs_rq()
for !fair_sched_class case.

Suppose that old user_struct/task_group is freed/reused, and the task
does

sched_setscheduler(SCHED_NORMAL);

__setscheduler() sets fair_sched_class, but doesn't update
->se.cfs_rq/parent which point to the freed memory.

This means that check_preempt_wakeup() doing

while (!is_same_group(se, pse)) {
se = parent_entity(se);
pse = parent_entity(pse);
}

may OOPS in a similar way if rq->curr or p did something like above.

Perhaps we need something like the patch below, note that
__setscheduler() can't do set_task_cfs_rq().

Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

sched: fix accounting of interrupts during guest execution on s390

Currently the scheduler checks for PF_VCPU to decide if this timeslice
has to be accounted as guest time. On s390 host interrupts are not
disabled during guest execution. This causes theses interrupts to be
accounted as guest time if CONFIG_VIRT_CPU_ACCOUNTING is set. Solution
is to check if an interrupt triggered account_system_time. As the tick
is timer interrupt based, we have to subtract hardirq_offset.

I tested the patch on s390 with CONFIG_VIRT_CPU_ACCOUNTING and on
x86_64. Seems to work.

CC: Avi Kivity <avi@qumranet.com>
CC: Laurent Vivier <Laurent.Vivier@bull.net>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge git://git./linux/kernel/git/herbert/crypto-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  [CRYPTO] geode: Fix not inplace encryption

Merge branch 'for-linus' of git://git./linux/kernel/git/roland/infiniband

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  mlx4_core: Fix thinko in QP destroy (incorrect bitmap_free)
  RDMA/cxgb3: Set the max_qp_init_rd_atom attribute in query_device
  IB/ehca: Fix static rate calculation
  IB/ehca: Return physical link information in query_port()
  IB/ipath: Fix race with ACK retry timeout list management
  IB/ipath: Fix memory leak in ipath_resize_cq() if copy_to_user() fails
  mlx4_core: Fix possible bad free in mlx4_buf_free()

Merge git://git./pub/scm/linux/kernel/git/sam/x86

* git://git.kernel.org/pub/scm/linux/kernel/git/sam/x86:
  x86: enable "make ARCH=x86"
  x86: do not use $(ARCH) when not needed
  kconfig: use $K64BIT to set 64BIT with all*config targets
  kconfig: add helper to set config symbol from environment variable
  kconfig: factor out code in confdata.c
  x86: move the rest of the menu's to Kconfig
  x86: move all simple arch settings to Kconfig
  x86: copy x86_64 specific Kconfig symbols to Kconfig.i386
  x86: add X86_64 dependency to x86_64 specific symbols in Kconfig.x86_64
  x86: add X86_32 dependency to i386 specific symbols in Kconfig.i386
  x86: arch/x86/Kconfig.cpu unification
  x86: start unification of arch/x86/Kconfig.*
  x86: unification of cfufreq/Kconfig

Merge branch 'master' of /linux/kernel/git/davem/net-2.6

* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [NET]: rt_check_expire() can take a long time, add a cond_resched()
  [ISDN] sc: Really, really fix warning
  [ISDN] sc: Fix sndpkt to have the correct number of arguments
  [TCP] FRTO: Clear frto_highmark only after process_frto that uses it
  [NET]: Remove notifier block from chain when register_netdevice_notifier fails
  [FS_ENET]: Fix module build.
  [TCP]: Make sure write_queue_from does not begin with NULL ptr
  [TCP]: Fix size calculation in sk_stream_alloc_pskb
  [S2IO]: Fixed memory leak when MSI-X vector allocation fails
  [BONDING]: Fix resource use after free
  [SYSCTL]: Fix warning for token-ring from sysctl checker
  [NET] random : secure_tcp_sequence_number should not assume CONFIG_KTIME_SCALAR
  [IWLWIFI]: Not correctly dealing with hotunplug.
  [TCP] FRTO: Plug potential LOST-bit leak
  [TCP] FRTO: Limit snd_cwnd if TCP was application limited
  [E1000]: Fix schedule while atomic when called from mii-tool.
  [NETX]: Fix build failure added by 2.6.24 statistics cleanup.
  [EP93xx_ETH]: Build fix after 2.6.24 NAPI changes.
  [PKT_SCHED]: Check subqueue status before calling hard_start_xmit

CRISv10 fasttimer: Scrap INLINE and name timeval_cmp better

Scrap the local __INLINE__ macro, and rename timeval_cmp to fasttime_cmp.

Inline macro was completely unnecessary since the macro was defined
locally to be inline.
timeval_cmp was inaccurately named since it does comparison on
struct fasttimer_t and not on struct timeval.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRISv10 memset library add lineendings to asm

Add \n\ at end of lines inside asm statement to avoid warning.

No change except adding \n\ to end of line and correcting
whitespace has been done.
Removes warning about multi-line string literals when compiling
arch/cris/arch-v10/lib/memset.c

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRISv10 string library add lineendings to asm

Add \n\ at end of lines inside asm statement to avoid warning.

No change except adding \n\ to end of line and correcting
whitespace has been done.
Removes warning about multi-line string literals when compiling
arch/cris/arch-v10/lib/string.c

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRISv10 usercopy library add lineendings to asm

Removes warning when compiling arch/cris/arch-v10/lib/usercopy.c

No change except adding \n\ on the end of the lines has been done.
Removes warning about multi-line string literals.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRISv10 Ethernet declare mac fix

Declare mac using DECLARE_MAC_BUF for use when calling print_mac().

This fixes compile error where mac was undeclared.
Also, remove unused variable i.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRIS tlb.h should include linux/pagemap.h

Include linux/pagemap.h for release_pages and page_cache_release.

Fixes compilation error in arch/cris/mm/init.c when CONFIG_SWAP is unset.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRISv32: add cache flush operations

These are needed due to a cache bug, and can be used to make sure that the
DMA descriptors are flushed to memory and can be safely handled by DMA.

flush_dma_descr - Flush one DMA descriptor.
flush_dma_list - Flush a complete list of DMA descriptors.
cris_flush_cache - Flush the complete cache.
cris_flush_cache_range - Flush only the specified range

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRISv10 improve and bugfix fasttimer

Improve and bugfix CRIS v10 fast timers.

- irq_handler_t now only takes two arguments.
- Keep interrupts disabled as long as we have a reference to the
  fasttimer list and only enable them while doing the callback.
  del_fast_timer may be called from other interrupt context.
- Fix bug where debug code could return without calling local_irq_restore.
- Use jiffies instead of usec (change from struct timeval to fasttime_t).
- Don't initialize static variables to zero.
- Remove obsolete #ifndef DECLARE_WAITQUEUE code.
- fast_timer_init should be __initcall.
- Change status/debug variables to unsigned.
- Remove CVS log and CVS id.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRISv10: fix timer interrupt parameters

Fix timer interrupt parameters for CRIS v10.

- irq_handler_t only takes two arguments, use get_irq_regs to get
  regs for profiling.
- Call update_process_times.
- Remove CVS tag.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRIS: remove MTD_AMSTD and MTD_OBSOLETE_CHIPS take two

Remove MTD_AMDSTD and MTD_OBSOLETE_CHIPS from defconfig, Kconfig and code,
instead we'll use MTD_CFI or MTD_JEDECPROBE.

[akpm@linux-foundation.org: codingl-style cleanups]
Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRISv10 serial driver rewrite

New and improved serial driver for CRISv10, take three, with improvements
suggested by Jiri Slaby.

- Call wait_event_interruptible with a _correct_ and sensible condition.
- Removed superfluous test of info->flags & ASYNC_CLOSING, since that is done
  by wait_event_interruptible.
- Moved common code for deregistering DMA and IRQ to deinit_port function.
- Use setup_timer when initializing flush_timer.
- Convert bit-field for uses_dma_in and uses_dma_out to regular bytes.
- Removed CVS tags.
- Removed defines and comments for CRIS_BUF_SIZE and TTY_THRESHOLD_THROTTLE
  (no longer used).
- Cleaned up code to pass checkpatch.
- Add crisv10.h header file.
- Merge of CRISv10 from Axis internal CVS.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Reviewed-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRIS don't include bitops.h in posix_types.h

In file included from include/asm/byteorder.h:23,
                 from include/asm-generic/bitops/le.h:5,
                 from include/asm-generic/bitops/ext2-non-atomic.h:4,
                 from include/asm/bitops.h:163,
                 from include/linux/bitops.h:17,
                 from include/asm/posix_types.h:55,
                 from include/linux/posix_types.h:47,
                 from include/linux/types.h:11,
                 from include/linux/capability.h:16,
                 from include/linux/sched.h:49,
                 from arch/cris/kernel/asm-offsets.c:1:
include/linux/byteorder/little_endian.h:43: parse error before "__cpu_to_le64p"
include/linux/byteorder/little_endian.h:44: warning: return type defaults to `int'
include/linux/byteorder/little_endian.h: In function `__cpu_to_le64p':
include/linux/byteorder/little_endian.h:45: `__le64' undeclared (first use in this function)

Remove include of asm/bitops.h, not needed here, corrects compilation error
(__le64 undeclared).

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris: ARRAY_SIZE() cleanup

I'm converting most array size calculations under arch/ to use
ARRAY_SIZE().  This is the patch for CRIS.

Signed-off-by: Alejandro Martinez Ruiz <alex@flawedcode.org>
Cc: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: defconfig updates

Remove some unused config lines, and add some others that needs to be off
for a succesful compile. This is the minimal set of changes for 2.6.23.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: fix crisksyms.c

- Add __negdi2 to exported symbols.
- Remove string functions from exported symbols.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: setup.c needs param.h

Include linux/param.h for HZ.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: fixes in arch/cris/kernel/time.c

- Remove debug print.
- Change #if to #ifdef to avoid compile time warning if CONFIG_PROFILING
  isn't set.
- Number of parameters to profile_tick has changed, drop the regs parameter.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: corrected and improved NMI and IRQ handling

Corrects compile errors and the following:

- Remove oldset parameter from do_signal and do_notify_resume.

- Modified to fit new consolidated IRQ handling code.

- Reverse check order between external nmi and watchdog nmi to avoid false
  watchdog oops in case of a glitch on the nmi pin.

- Return from an pin-generated NMI the same way as for other interrupts.

- Moved blocking of ethernet rx/tx irq from ethernet interrupt handler to
  low-level asm interrupt handlers.  Fixed in the multiple interrupt
  handler also.

- Add space for thread local storage in thread_info struct.

- Add NO_DMA to Kconfig, and include arch specific Kconfig using arch
  independent path. Include subsystem Kconfigs for pcmcia, usb, i2c,
  rtc and pci.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris-build-fixes-update-eth_v10c-ethernet-driver-fix

Fix locking bug noted by  Roel Kluin <12o3l@tiscali.nl>.

Cc: Jeff Garzik <jeff@garzik.org>
Cc: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: Roel Kluin <12o3l@tiscali.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: update eth_v10.c ethernet driver

New (updated) version of ethernet driver for cris v10.

- First steps to simplify and make the MII code more similar
  between the etrax100 and etraxfs ports.

- Start the transmit queue before enabling tx interrupts
  to avoid race with the first frame.

- Flip the comparition statement to stick to physical addresses
  to avoid phys_to_virt mapping a potential null pointer.
  This was not an error but the change simplifies debugging
  of address-space mappings.

- Made myPrevRxDesc local to e100_rx since it was only used there.
  Fixed out of memory handling in e100_rx.  If dev_alloc_skb() fails
  persistently the system is hosed anyway but at least it won't
  loop in an interrupt handler.

- Correct some code formatting issues.

- Add defines SET_ETH_ENABLE_LEDS, SET_ETH_DISABLE_LEDS
  and SET_ETH_AUTONEG used in new cris v10 ethernet driver.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: add baud rate defines

Add missing defines for (unsupported) baud rates.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: sys_cris.c needs fs.h

Include linux/fs.h, needed for do_pipe.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: irq fixes

- New file include/asm-cris/irq_regs.h.
- Change handling of registers for do_IRQ.
- Add GENERIC_HARDIRQS to Kconfig.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris-build-fixes-atomich-needs-compilerh-fix

Cc: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: atomic.h needs compiler.h

Include file linux/compiler.h is needed for 'likely'.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: hardirq.h: include asm/irq.h

Include asm/irq.h to avoid undefined value warning.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: Add missing syscalls

Add missing syscalls to cris architecture.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris build fixes: fix csum_tcpudp_magic() declaration

Remove int from prototype, no longer needed and causes compile error.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Acked-by: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

aic94xx_sds: rename FLASH_SIZE

arm:

drivers/scsi/aic94xx/aic94xx_sds.c:381:1: warning: "FLASH_SIZE" redefined
In file included from include/asm/arch/irqs.h:22,
                 from include/asm/irq.h:4,
                 from include/asm/hardirq.h:6,
                 from include/linux/hardirq.h:7,
                 from include/asm-generic/local.h:5,
                 from include/asm/local.h:1,
                 from include/linux/module.h:19,
                 from include/linux/device.h:21,
                 from include/linux/pci.h:52,
                 from drivers/scsi/aic94xx/aic94xx_sds.c:28:
include/asm/arch/platform.h:444:1: warning: this is the location of the previous definition

Cc: Gilbert Wu <gilbert_wu@adaptec.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

x86: don't call mce_create_device on CPU_UP_PREPARE

Fix regression introduced with d435d862baca3e25e5eec236762a43251b1e7ffc
("cpu hotplug: mce: fix cpu hotplug error handling").

A CPU which was not brought up during boot (using maxcpus and
additional_cpus parameters) couldn't be onlined anymore.  For such a CPU it
seemed that MCE was not supported during CPU_UP_PREPARE-time which caused
mce_cpu_callback to return NOTIFY_BAD to notifier_call_chain.  To fix this
we:

 - call mce_create_device for CPU_ONLINE event (instead of CPU_UP_PREPARE),
 - avoid mce_remove_device() for the CPU that is not correctly initialized
   by mce_create_device() failure,
 - make mce_cpu_callback always return NOTIFY_OK for CPU_ONLINE event.
   Because CPU_ONLINE callback return value is always ignored.

[akinobu.mita@gmail.com: avoid mce_remove_device() for not initialized device]
[akinobu.mita@gmail.com: make mce_cpu_callback always return NOTIFY_OK]
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oprofile: fix oops on x86 32-bit

x86 32-bit isn't saving the stack pointer to pt_regs->esp when an
interrupt occurs.

Signed-off-by: Jan Blunck <jblunck@suse.de>
Tested-by: Robert Fitzsimons <robfitz@273k.net>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Philippe Elie <phil.el@wanadoo.fr>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

x86: disable preemption in delay_tsc()

Marin Mitov points out that delay_tsc() can misbehave if it is preempted and
rescheduled on a different CPU which has a skewed TSC.  Fix it by disabling
preemption.

(I assume that the worst-case behaviour here is a stall of 2^32 cycles)

Cc: Andi Kleen <ak@suse.de>
Cc: Marin Mitov <mitov@issp.bas.bg>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

keyspan: init termios properly

Remove redundant code leading to NULL ptr deref and let terminal config
settings take place in the proper initialization path in
usb_console_setup().

Signed-off-by: Borislav Petkov <bbpetkov@yahoo.de>
Cc: <lucy@keyspan.com>
Cc: Greg KH <greg@kroah.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ACPI: AC: Update AC state on resume

Check if AC state has changed across resume and notify userspace if so.

Fixes "[2.6.24-rc1 regression] AC adapter state does not change after resume"

Signed-off-by: Alexey Starikovskiy <astarikovskiy@suse.de>
Tested-by: Andrey Borzenkov <arvidjaar@mail.ru>
Cc: Len Brown <lenb@kernel.org>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

make /proc/acpi/ac_adapter dependent on ACPI_PROCFS

Do not provide /proc/acpi/ac_adapter if ACPI_PROCFS is not defined.  This
eliminates duplicated power adapters in HAL and makes it consistent with
battery module

Signed-off-by: Andrey Borzenkov <arvidjaar@mail.ru>
Acked-by: Alexey Starikovskiy <astarikovskiy@suse.de>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/taskstats.c: fix bogus nlmsg_free()

We'd better not nlmsg_free on a pointer containing an undefined value
(and without having anything allocated).

Spotted by the Coverity checker.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

feature-removal-schedule: remove SA_* flags entry

No reason to keep the feature-removal-schedule.txt entry after the code
was removed.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

file capabilities: allow sigcont within session

Fix http://bugzilla.kernel.org/show_bug.cgi?id=9247

Allow sigcont to be sent to a process with greater capabilities if it is in
the same session.  Otherwise, a shell from which I've started a root shell
and done 'suspend' can't be restarted by the parent shell.

Also don't do file-capabilities signaling checks when uids for the
processes don't match, since the standard check_kill_permission will have
done those checks.

[akpm@linux-foundation.org: coding-style cleanups]
Signed-off-by: Serge E. Hallyn <serue@us.ibm.com>
Acked-by: Andrew Morgan <morgan@kernel.org>
Cc: Chris Wright <chrisw@sous-sol.org>
Tested-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Stephen Smalley <sds@epoch.ncsc.mil>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Chris Wright <chrisw@sous-sol.org>
Cc: James Morris <jmorris@namei.org>
Cc: Stephen Smalley <sds@tycho.nsa.gov>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Swap delay accounting, include lock_page() delays

The delay incurred in lock_page() should also be accounted in swap delay
accounting

Reported-by: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: fix build for !CONFIG_PRINTK

Handle the case of CONFIG_PRINTK being disabled.  This requires a do-nothing
stub to be present in arch/um/include/user.h so that we don't get references
to printk from libc code.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: fix build for !CONFIG_TCP

Make UML build in the absence of CONFIG_INET by making the inetaddr_notifier
registration depend on it.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: remove last include of libc asm/page.h

asm/page.h is disappearing from the libc headers and we don't need it anyway.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: fix spurious IRQ testing

The spurious IRQ testing in request_irq is mishandled in um_request_irq, which
sets the incoming file descriptors non-blocking only after request_irq
succeeds.  This results in the spurious irq calling read on a blocking
descriptor, and a hang.

Fixed by reversing the O_NONBLOCK setting and the request_irq call.

Signed-off-by: Jeff Dike <jdike@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Fix 64KB blocksize in ext3 directories

With 64KB blocksize, a directory entry can have size 64KB which does not
fit into 16 bits we have for entry lenght.  So we store 0xffff instead and
convert value when read from / written to disk.  The patch also converts
some places to use ext3_next_entry() when we are changing them anyway.

[akpm@linux-foundation.org: coding-style cleanups]
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

smbfs: fix debug builds

Fix some warnings with SMBFS_DEBUG_* builds.  This patch makes it so that
builds with -Werror don't fail.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hibernate: fix lockdep report

Lockdep reports a circular locking dependency in the hibernate code
because
 - during system boot hibernate code (from an initcall) locks pm_mutex
   and then a sysfs buffer mutex via name_to_dev_t
 - during regular operation hibernate code locks pm_mutex under a
   sysfs buffer mutex because it's called from sysfs methods.

The deadlock can never happen because during initcall invocation nothing
can write to sysfs yet. This removes the lockdep report by marking the
initcall locking as being in a different class.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

__do_IRQ does not check IRQ_DISABLED when IRQ_PER_CPU is set

In __do_IRQ(), the normal case is that IRQ_DISABLED is checked and if set
the handler (handle_IRQ_event()) is not called.

Earlier in __do_IRQ(), if IRQ_PER_CPU is set the code does not check
IRQ_DISABLED and calls the handler even though IRQ_DISABLED is set.  This
behavior seems unintentional.

One user encountering this behavior is the CPE handler (in
arch/ia64/kernel/mca.c).  When the CPE handler encounters too many CPEs
(such as a solid single bit error), it sets up a polling timer and disables
the CPE interrupt (to avoid excessive overhead logging the stream of single
bit errors).  disable_irq_nosync() is called which sets IRQ_DISABLED.  The
IRQ_PER_CPU flag was previously set (in ia64_mca_late_init()).  The net
result is the CPE handler gets called even though it is marked disabled.

If the behavior of not checking IRQ_DISABLED when IRQ_PER_CPU is set is
intentional, it would be worthy of a comment describing the intended
behavior.  disable_irq_nosync() does call chip->disable() to provide a
chipset specifiec interface for disabling the interrupt, which avoids this
issue when used.

Signed-off-by: Russ Anderson <rja@sgi.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pidns: Place under CONFIG_EXPERIMENTAL

This is my trivial patch to swat innumerable little bugs with a single
blow.

After some intensive review (my apologies for not having gotten to this
sooner) what we have looks like a good base to build on with the current
pid namespace code but it is not complete, and it is still much to simple
to find issues where the kernel does the wrong thing outside of the initial
pid namespace.

Until the dust settles and we are certain we have the ABI and the
implementation is as correct as humanly possible let's keep process ID
namespaces behind CONFIG_EXPERIMENTAL.

Allowing us the option of fixing any ABI or other bugs we find as long as
they are minor.

Allowing users of the kernel to avoid those bugs simply by ensuring their
kernel does not have support for multiple pid namespaces.

[akpm@linux-foundation.org: coding-style cleanups]
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Cedric Le Goater <clg@fr.ibm.com>
Cc: Adrian Bunk <bunk@kernel.org>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Kir Kolyshkin <kir@swsoft.com>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

vmstat: fix section mismatch warning

Mark start_cpu_timer() as __cpuinit instead of __devinit.
Fixes this section warning:

WARNING: vmlinux.o(.text+0x60e53): Section mismatch: reference to .init.text:start_cpu_timer (between 'vmstat_cpuup_callback' and 'vmstat_show')

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gbefb: fix section mismatch warnings

Make 'default_mode' and 'default_var' be __initdata.
Fixes these section warnings:

WARNING: vmlinux.o(.data+0x128e0): Section mismatch: reference to .init.data:default_mode_CRT (between 'default_mode' and 'default_var')
WARNING: vmlinux.o(.data+0x128e4): Section mismatch: reference to .init.data:default_var_CRT (between 'default_var' and 'dev_attr_size')

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mark sys_open/sys_read exports unused

sys_open / sys_read were used in the early 1.2 days to load firmware from
disk inside drivers.  Since 2.0 or so this was deprecated behavior, but
several drivers still were using this.  Since a few years we have a
request_firmware() API that implements this in a nice, consistent way.
Only some old ISA sound drivers (pre-ALSA) still straggled along for some
time....  however with commit c2b1239a9f22f19c53543b460b24507d0e21ea0c the
last user is now gone.

This is a good thing, since using sys_open / sys_read etc for firmware is a
very buggy to dangerous thing to do; these operations put an fd in the
process file descriptor table....  which then can be tampered with from
other threads for example.  For those who don't want the firmware loader,
filp_open()/vfs_read are the better APIs to use, without this security
issue.

The patch below marks sys_open and sys_read unused now that they're
really not used anymore, and for deletion in the 2.6.25 timeframe.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fix param_sysfs_builtin name length check

Commit faf8c714f4508207a9c81cc94dafc76ed6680b44 caused a regression:
parameter names longer than MAX_KBUILD_MODNAME will now be rejected,
although we just need to keep the module name part that short.  This patch
restores the old behaviour while still avoiding that memchr is called with
its length parameter larger than the total string length.

Signed-off-by: Jan Kiszka <jan.kiszka@web.de>
Cc: Dave Young <hidave.darkstar@gmail.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

proc: simplify and correct proc_flush_task

Currently we special case when we have only the initial pid namespace.
Unfortunately in doing so the copied case for the other namespaces was
broken so we don't properly flush the thread directories :(

So this patch removes the unnecessary special case (removing a usage of
proc_mnt) and corrects the flushing of the thread directories.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Al Viro <viro@ftp.linux.org.uk>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: Kirill Korotaev <dev@sw.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mips: undo locking on error path returns

[akpm@linux-foundation.org: coding-style cleanups]
Signed-off-by: Roel Kluin <12o3l@tiscali.nl>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris gpio: undo locks before returning

Signed-off-by: Roel Kluin <12o3l@tiscali.nl>
Cc: Mikael Starvik <starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rd: fix data corruption on memory pressure

We have seen ramdisk based install systems, where some pages of mapped
libraries and programs were suddendly zeroed under memory pressure.  This
should not happen, as the ramdisk avoids freeing its pages by keeping them
dirty all the time.

It turns out that there is a case, where the VM makes a ramdisk page clean,
without telling the ramdisk driver.  On memory pressure shrink_zone runs
and it starts to run shrink_active_list.  There is a check for
buffer_heads_over_limit, and if true, pagevec_strip is called.
pagevec_strip calls try_to_release_page.  If the mapping has no releasepage
callback, try_to_free_buffers is called.  try_to_free_buffers has now a
special logic for some file systems to make a dirty page clean, if all
buffers are clean.  Thats what happened in our test case.

The simplest solution is to provide a noop-releasepage callback for the
ramdisk driver.  This avoids try_to_free_buffers for ramdisk pages.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Nick Piggin <npiggin@suse.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tle62x0 driver stops ignoring read errors

The tle62x0 driver was ignoring all read errors.  This patch makes it
pass such errors up the stack, instead of returning bogus data.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fuse_file_alloc(): fix NULL dereferences

Fix obvious NULL dereferences spotted by the Coverity checker.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fix mm/util.c:krealloc()

Commit ef8b4520bd9f8294ffce9abd6158085bde5dc902 added one NULL check for
"p" in krealloc(), but that doesn't seem to be enough since there
doesn't seem to be any guarantee that memcpy(ret, NULL, 0) works
(spotted by the Coverity checker).

For making it clearer what happens this patch also removes the pointless
min().

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sunrpc/xprtrdma/transport.c: fix use-after-free

Fix an obvious use-after-free spotted by the Coverity checker.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Cc: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Neil Brown <neilb@suse.de>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

serial: only use PNP IRQ if it's valid

"Luming Yu" <luming.yu@gmail.com> says:

  There is a "ttyS1 irq is -1" problem observed on tiger4 which cause the
  serial port broken.

  It is because that there is __no__ ACPI IRQ resource assigned for the
  serial port.  So the value of the IRQ for the port is never changed since it
  got initialized to -1.

If PNP supplies a valid IRQ, use it.  Otherwise, leave port.irq == 0, which
means "no IRQ" to the serial core.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Yu Luming <luming.yu@intel.com>
Acked-by: Matthew Wilcox <matthew@wil.cx>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

i5000_edac: no need to __stringify() KBUILD_BASENAME

The i5000_edac driver's PCI registration structure has the name
""i5000_edac"" (with extra set of double-quotes) which is probably not
intentional.  Get rid of __stringify.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Cc: Doug Thompson <norsk5@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rtc: fall back to requesting only the ports we actually use

Firmware like PNPBIOS or ACPI can report the address space consumed by the
RTC.  The actual space consumed may be less than the size (RTC_IO_EXTENT)
assumed by the RTC driver.

The PNP core doesn't request resources yet, but I'd like to make it do so.
If/when it does, the RTC_IO_EXTENT request may fail, which prevents the RTC
driver from loading.

Since we only use the RTC index and data registers at RTC_PORT(0) and
RTC_PORT(1), we can fall back to requesting just enough space for those.

If the PNP core requests resources, this results in typical I/O port usage
like this:

    0070-0073 : 00:06 <-- PNP device 00:06 responds to 70-73
      0070-0071 : rtc <-- RTC driver uses only 70-71

instead of the current:

    0070-0077 : rtc <-- RTC_IO_EXTENT == 8

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rtc: release correct region in error path

The misc_register() error path always released an I/O port region,
even if the region was memory-mapped (only mips uses memory-mapped RTC,
as far as I can see).

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

reiserfs: don't drop PG_dirty when releasing sub-page-sized dirty file

This is not a new problem in 2.6.23-git17.  2.6.22/2.6.23 is buggy in the
same way.

Reiserfs could accumulate dirty sub-page-size files until umount time.
They cannot be synced to disk by pdflush routines or explicit `sync'
commands.  Only `umount' can do the trick.

The direct cause is: the dirty page's PG_dirty is wrongly _cleared_.
Call trace:
 [<ffffffff8027e920>] cancel_dirty_page+0xd0/0xf0
 [<ffffffff8816d470>] :reiserfs:reiserfs_cut_from_item+0x660/0x710
 [<ffffffff8816d791>] :reiserfs:reiserfs_do_truncate+0x271/0x530
 [<ffffffff8815872d>] :reiserfs:reiserfs_truncate_file+0xfd/0x3b0
 [<ffffffff8815d3d0>] :reiserfs:reiserfs_file_release+0x1e0/0x340
 [<ffffffff802a187c>] __fput+0xcc/0x1b0
 [<ffffffff802a1ba6>] fput+0x16/0x20
 [<ffffffff8029e676>] filp_close+0x56/0x90
 [<ffffffff8029fe0d>] sys_close+0xad/0x110
 [<ffffffff8020c41e>] system_call+0x7e/0x83

Fix the bug by removing the cancel_dirty_page() call. Tests show that
it causes no bad behaviors on various write sizes.

=== for the patient ===
Here are more detailed demonstrations of the problem.

1) the page has both PG_dirty(D)/PAGECACHE_TAG_DIRTY(d) after being written to;
   and then only PAGECACHE_TAG_DIRTY(d) remains after the file is closed.

------------------------------ screen 0 ------------------------------
[T0] root /home/wfg# cat > /test/tiny
[T1] hi
[T2] root /home/wfg#

------------------------------ screen 1 ------------------------------
[T1] root /home/wfg# echo /test/tiny > /proc/filecache
[T1] root /home/wfg# cat /proc/filecache
     # file /test/tiny
     # flags R:referenced A:active M:mmap U:uptodate D:dirty W:writeback O:owner B:buffer d:dirty w:writeback
     # idx   len     state   refcnt
     0       1       ___UD__Bd_      2
[T2] root /home/wfg# cat /proc/filecache
     # file /test/tiny
     # flags R:referenced A:active M:mmap U:uptodate D:dirty W:writeback O:owner B:buffer d:dirty w:writeback
     # idx   len     state   refcnt
     0       1       ___U___Bd_      2

2) note the non-zero 'cancelled_write_bytes' after /tmp/hi is copied.

------------------------------ screen 0 ------------------------------
[T0] root /home/wfg# echo hi > /tmp/hi
[T1] root /home/wfg# cp /tmp/hi /dev/stdin /test
[T2] hi
[T3] root /home/wfg#

------------------------------ screen 1 ------------------------------
[T1] root /proc/4397# cd /proc/`pidof cp`
[T1] root /proc/4713# cat io
     rchar: 8396
     wchar: 3
     syscr: 20
     syscw: 1
     read_bytes: 0
     write_bytes: 20480
     cancelled_write_bytes: 4096
[T2] root /proc/4713# cat io
     rchar: 8399
     wchar: 6
     syscr: 21
     syscw: 2
     read_bytes: 0
     write_bytes: 24576
     cancelled_write_bytes: 4096

//Question: the 'write_bytes' is a bit more than expected ;-)

Tested-by: Maxim Levitsky <maximlevitsky@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn>
Reviewed-by: Chris Mason <chris.mason@oracle.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

I/OAT: Add support for version 2 of ioatdma device

Add support for version 2 of the ioatdma device.  This device handles
the descriptor chain and DCA services slightly differently:
 - Instead of moving the dma descriptors between a busy and an idle chain,
   this new version uses a single circular chain so that we don't have
   rewrite the next_descriptor pointers as we add new requests, and the
   device doesn't need to re-read the last descriptor.
 - The new device has the DCA tags defined internally instead of needing
   them defined statically.

Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Cc: "Williams, Dan J" <dan.j.williams@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Linux Kernel Markers: fix samples to follow format string standard

Add the field names to marker example format string.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Linux Kernel Markers: document format string

Describes the format string standard further: Use of field names before the
type specifiers..

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Linux Kernel Markers: fix marker mutex not taken upon module load

Upon module load, we must take the markers mutex.  It implies that the marker
mutex must be nested inside the module mutex.

It implies changing the nesting order : now the marker mutex nests inside the
module mutex.  Make the necessary changes to reverse the order in which the
mutexes are taken.

Includes some cleanup from Dave Hansen <haveblue@us.ibm.com>.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Fixes to the BFS filesystem driver

I found a few bugs in the BFS driver.  Detailed description of the bugs as
well as the steps to reproduce the errors are given in the kernel bugzilla.
 Please follow these links for more information:

http://bugzilla.kernel.org/show_bug.cgi?id=9363
http://bugzilla.kernel.org/show_bug.cgi?id=9364
http://bugzilla.kernel.org/show_bug.cgi?id=9365
http://bugzilla.kernel.org/show_bug.cgi?id=9366

This patch fixes the bugs described above.  Besides, the patch introduces
coding style changes to make the BFS driver conform to the requirements
specified for Linux kernel code.  Finally, I made a few cosmetic changes
such as removal of trivial debug output.

Also, the patch removes the fields `si_lf_ioff' and `si_lf_sblk' of the
in-core superblock structure.  These fields are initialized but never
actually used.

If you are wondering why I need BFS, here is the answer: I am using this
driver in the context of Linux kernel classes I am teaching in the Moscow
State University and in the International Institute of Information
Technology in Pune, India.

Signed-off-by: Dmitri Vorobiev <dmitri.vorobiev@gmail.com>
Cc: Tigran Aivazian <tigran@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

revert "Task Control Groups: example CPU accounting subsystem"

Revert 62d0df64065e7c135d0002f069444fbdfc64768f.

This was originally intended as a simple initial example of how to create a
control groups subsystem; it wasn't intended for mainline, but I didn't make
this clear enough to Andrew.

The CFS cgroup subsystem now has better functionality for the per-cgroup usage
accounting (based directly on CFS stats) than the "usage" status file in this
patch, and the "load" status file is rather simplistic - although having a
per-cgroup load average report would be a useful feature, I don't believe this
patch actually provides it.  If it gets into the final 2.6.24 we'd probably
have to support this interface for ever.

Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: fix i_blocks accounting

For administrative purpose, we want to query actual block usage for
hugetlbfs file via fstat.  Currently, hugetlbfs always return 0.  Fix that
up since kernel already has all the information to track it properly.

Signed-off-by: Ken Chen <kenchen@google.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/hugetlb.c: make a function static

return_unused_surplus_pages() can become static.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: enforce quotas during reservation for shared mappings

When a MAP_SHARED mmap of a hugetlbfs file succeeds, huge pages are reserved
to guarantee no problems will occur later when instantiating pages.  If quotas
are in force, page instantiation could fail due to a race with another process
or an oversized (but approved) shared mapping.

To prevent these scenarios, debit the quota for the full reservation amount up
front and credit the unused quota when the reservation is released.

Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: allow bulk updating in hugetlb_*_quota()

Add a second parameter 'delta' to hugetlb_get_quota and hugetlb_put_quota to
allow bulk updating of the sbinfo->free_blocks counter.  This will be used by
the next patch in the series.

Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: debit quota in alloc_huge_page

Now that quota is credited by free_huge_page(), calls to hugetlb_get_quota()
seem out of place.  The alloc/free API is unbalanced because we handle the
hugetlb_put_quota() but expect the caller to open-code hugetlb_get_quota().
Move the get inside alloc_huge_page to clean up this disparity.

This patch has been kept apart from the previous patch because of the somewhat
dodgy ERR_PTR() use herein.  Moving the quota logic means that
alloc_huge_page() has two failure modes.  Quota failure must result in a
SIGBUS while a standard allocation failure is OOM.  Unfortunately, ERR_PTR()
doesn't like the small positive errnos we have in VM_FAULT_* so they must be
negated before they are used.

Does anyone take issue with the way I am using PTR_ERR.  If so, what are your
thoughts on how to clean this up (without needing an if,else if,else block at
each alloc_huge_page() callsite)?

Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: fix quota management for private mappings

The hugetlbfs quota management system was never taught to handle MAP_PRIVATE
mappings when that support was added.  Currently, quota is debited at page
instantiation and credited at file truncation.  This approach works correctly
for shared pages but is incomplete for private pages.  In addition to
hugetlb_no_page(), private pages can be instantiated by hugetlb_cow(); but
this function does not respect quotas.

Private huge pages are treated very much like normal, anonymous pages.  They
are not "backed" by the hugetlbfs file and are not stored in the mapping's
radix tree.  This means that private pages are invisible to
truncate_hugepages() so that function will not credit the quota.

This patch (based on a prototype provided by Ken Chen) moves quota crediting
for all pages into free_huge_page().  page->private is used to store a pointer
to the mapping to which this page belongs.  This is used to credit quota on
the appropriate hugetlbfs instance.

Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: split alloc_huge_page into private and shared components

Hugetlbfs implements a quota system which can limit the amount of memory that
can be used by the filesystem.  Before allocating a new huge page for a file,
the quota is checked and debited.  The quota is then credited when truncating
the file.  I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED
mappings.  Before detailing the problems and my proposed solutions, we should
agree on a definition of quotas that properly addresses both private and
shared pages.  Since the purpose of quotas is to limit total memory
consumption on a per-filesystem basis, I argue that all pages allocated by the
fs (private and shared) should be charged against quota.

Private Mappings
================

The current code will debit quota for private pages sometimes, but will never
credit it.  At a minimum, this causes a leak in the quota accounting which
renders the accounting essentially useless as it is.  Shared pages have a one
to one mapping with a hugetlbfs file and are easy to account by debiting on
allocation and crediting on truncate.  Private pages are anonymous in nature
and have a many to one relationship with their hugetlbfs files (due to copy on
write).  Because private pages are not indexed by the mapping's radix tree,
thier quota cannot be credited at file truncation time.  Crediting must be
done when the page is unmapped and freed.

Shared Pages
============

I discovered an issue concerning the interaction between the MAP_SHARED
reservation system and quotas.  Since quota is not checked until page
instantiation, an over-quota mmap/reservation will initially succeed.  When
instantiating the first over-quota page, the program will receive SIGBUS.
This is inconsistent since the reservation is supposed to be a guarantee.  The
solution is to debit the full amount of quota at reservation time and credit
the unused portion when the reservation is released.

This patch series brings quotas back in line by making the following
modifications:
 * Private pages
   - Debit quota in alloc_huge_page()
   - Credit quota in free_huge_page()
 * Shared pages
   - Debit quota for entire reservation at mmap time
   - Credit quota for instantiated pages in free_huge_page()
   - Credit quota for unused reservation at munmap time

This patch:

The shared page reservation and dynamic pool resizing features have made the
allocation of private vs.  shared huge pages quite different.  By splitting
out the private/shared-specific portions of the process into their own
functions, readability is greatly improved.  alloc_huge_page now calls the
proper helper and performs common operations.

[akpm@linux-foundation.org: coding-style cleanups]
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

raid5: fix unending write sequence

<debug output from Joel's system>
handling stripe 7629696, state=0x14 cnt=1, pd_idx=2 ops=0:0:0
check 5: state 0x6 toread 0000000000000000 read 0000000000000000 write fffff800ffcffcc0 written 0000000000000000
check 4: state 0x6 toread 0000000000000000 read 0000000000000000 write fffff800fdd4e360 written 0000000000000000
check 3: state 0x1 toread 0000000000000000 read 0000000000000000 write 0000000000000000 written 0000000000000000
check 2: state 0x1 toread 0000000000000000 read 0000000000000000 write 0000000000000000 written 0000000000000000
check 1: state 0x6 toread 0000000000000000 read 0000000000000000 write fffff800ff517e40 written 0000000000000000
check 0: state 0x6 toread 0000000000000000 read 0000000000000000 write fffff800fd4cae60 written 0000000000000000
locked=4 uptodate=2 to_read=0 to_write=4 failed=0 failed_num=0
for sector 7629696, rmw=0 rcw=0
</debug>

These blocks were prepared to be written out, but were never handled in
ops_run_biodrain(), so they remain locked forever.  The operations flags
are all clear which means handle_stripe() thinks nothing else needs to be
done.

This state suggests that the STRIPE_OP_PREXOR bit was sampled 'set' when it
should not have been.  This patch cleans up cases where the code looks at
sh->ops.pending when it should be looking at the consistent stack-based
snapshot of the operations flags.

Report from Joel:
Resync done. Patch fix this bug.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Tested-by: Joel Bertrand <joel.bertrand@systella.fr>
Cc: <stable@kernel.org>
Cc: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: follow_hugetlb_page() for write access

When calling get_user_pages(), a write flag is passed in by the caller to
indicate if write access is required on the faulted-in pages.  Currently,
follow_hugetlb_page() ignores this flag and always faults pages for
read-only access.  This can cause data corruption because a device driver
that calls get_user_pages() with write set will not expect COW faults to
occur on the returned pages.

This patch passes the write flag down to follow_hugetlb_page() and makes
sure hugetlb_fault() is called with the right write_access parameter.

[ezk@cs.sunysb.edu: build fix]
Signed-off-by: Adam Litke <agl@us.ibm.com>
Reviewed-by: Ken Chen <kenchen@google.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Erez Zadok <ezk@cs.sunysb.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

atmel_serial build warnings begone

Remove annoying build warnings about unused variables in atmel_serial,
which afflict both AT91 and AVR32 builds.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Acked-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Cc: Andrew Victor <andrew@sanpeople.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

dmaengine: fix broken device refcounting

When a DMA device is unregistered, its reference count is decremented twice
for each channel: Once dma_class_dev_release() and once in
dma_chan_cleanup().  This may result in the DMA device driver's remove()
function completing before all channels have been cleaned up, causing lots
of use-after-free fun.

Fix it by incrementing the device's reference count twice for each
channel during registration.

[dan.j.williams@intel.com: kill unnecessary client refcounting]
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Shannon Nelson <shannon.nelson@intel.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/misc: Move misplaced pci_dev_put's

Move pci_dev_put outside the loops in which it occurs.  Within the loop,
pci_dev_put is done implicitly by pci_get_device.

The problem was detected using the following semantic patch, and corrected
by hand.

@@
expression dev;
expression E;
@@

- pci_dev_put(dev)
   ... when != dev = E
- pci_get_device(...,dev)

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

paride: pf driver fixes

The pf driver for parallel port floppy drives seems to be broken.  At least
with Imation SuperDisk with EPAT chip, the driver calls pi_connect() and
pi_disconnect after each transferred sector.  At least with EPAT, this
operation is very expensive - causes drive recalibration.  Thus, transferring
even a single byte (dd if=/dev/pf0 of=/dev/null bs=1 count=1) takes 20
seconds, making the driver useless.

The pf_next_buf() function seems to be broken as it returns 1 always (except
when pf_run is non-zero), causing the loop in do_pf_read_drq (and
do_pf_write_drq) to be executed only once.

The following patch fixes this problem.  It also fixes swapped descriptions in
pf_lock() function and removes DBMSG macro, which seems useless.

Signed-off-by: Ondrej Zary <linux@rainbow-software.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

spi: fix error paths on txx9spi_probe

Some error paths in txx9spi_probe wrongly return 0.  This patch fixes them by
using the devres interfaces.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Acked-by: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

spi: fix double-free on spi_unregister_master

After 49dce689ad4ef0fd1f970ef762168e4bd46f69a3, device_for_each_child
iteration hits the master device itself.  Do not call spi_unregister_device()
for the master device.

Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Acked-by: David Brownell <david-b@pacbell.net>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

acpi: make ACPI_PROCFS default to y

Zillions of people are getting my-battery-monitor-doesnt-work problems
(including me).

Lessen the damage by making ACPI_PROCFS default to on.

Cc: Len Brown <lenb@kernel.org>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add IORESOUCE_BUSY flag for System RAM

i386 and x86-64 registers System RAM as IORESOURCE_MEM | IORESOURCE_BUSY.

But ia64 registers it as IORESOURCE_MEM only.
In addition, memory hotplug code registers new memory as IORESOURCE_MEM too.

This difference causes a failure of memory unplug of x86-64.  This patch
fixes it.

This patch adds IORESOURCE_BUSY to avoid potential overlap mapping by PCI
device.

Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Luck, Tony" <tony.luck@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: speed up writeback ramp-up on clean systems

We allow violation of bdi limits if there is a lot of room on the system.
Once we hit half the total limit we start enforcing bdi limits and bdi
ramp-up should happen.  Doing it this way avoids many small writeouts on an
otherwise idle system and should also speed up the ramp-up.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: Fengguang Wu <wfg@mail.ustc.edu.cn>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

make getdelays cgroupstats aware

Update the getdelays utility to become cgroupstats aware.  A new -C option has
been added.  It takes in a control group path and prints out a summary of the
states of tasks in the control group

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memory hotremove: unset migrate type "ISOLATE" after removal

We should unset migrate type "ISOLATE" when we successfully removed memory.
 But current code has BUG and cannot works well.

This patch also includes bugfix?  to change get_pageblock_flags to
get_pageblock_migratetype().

Thanks to Badari Pulavarty for finding this.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Migration: find correct vma in new_vma_page()

We hit the BUG_ON() in mm/rmap.c:vma_address() when trying to migrate via
mbind(MPOL_MF_MOVE) a non-anon region that spans multiple vmas.  For
anon-regions, we just fail to migrate any pages beyond the 1st vma in the
range.

This occurs because do_mbind() collects a list of pages to migrate by
calling check_range().  check_range() walks the task's mm, spanning vmas as
necessary, to collect the migratable pages into a list.  Then, do_mbind()
calls migrate_pages() passing the list of pages, a function to allocate new
pages based on vma policy [new_vma_page()], and a pointer to the first vma
of the range.

For each page in the list, new_vma_page() calls page_address_in_vma()
passing the page and the vma [first in range] to obtain the address to get
for alloc_page_vma().  The page address is needed to get interleaving
policy correct.  If the pages in the list come from multiple vmas,
eventually, new_page_address() will pass that page to page_address_in_vma()
with the incorrect vma.  For !PageAnon pages, this will result in a bug
check in rmap.c:vma_address().  For anon pages, vma_address() will just
return EFAULT and fail the migration.

This patch modifies new_vma_page() to check the return value from
page_address_in_vma().  If the return value is EFAULT, new_vma_page()
searchs forward via vm_next for the vma that maps the page--i.e., that does
not return EFAULT.  This assumes that the pages in the list handed to
migrate_pages() is in address order.  This is currently case.  The patch
documents this assumption in a new comment block for new_vma_page().

If new_vma_page() cannot locate the vma mapping the page in a forward
search in the mm, it will pass a NULL vma to alloc_page_vma().  This will
result in the allocation using the task policy, if any, else system default
policy.  This situation is unlikely, but the patch documents this behavior
with a comment.

Note, this patch results in restarting from the first vma in a multi-vma
range each time new_vma_page() is called.  If this is not acceptable, we
can make the vma argument a pointer, both in new_vma_page() and it's caller
unmap_and_move() so that the value held by the loop in migrate_pages()
always passes down the last vma in which a page was found.  This will
require changes to all new_page_t functions passed to migrate_pages().  Is
this necessary?

For this patch to work, we can't bug check in vma_address() for pages
outside the argument vma.  This patch removes the BUG_ON().  All other
callers [besides new_vma_page()] already check the return status.

Tested on x86_64, 4 node NUMA platform.

Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

proc: fix proc_kill_inodes to kill dentries on all proc superblocks

It appears we overlooked support for removing generic proc files
when we added support for multiple proc super blocks.  Handle
that now.

[akpm@linux-foundation.org: coding-style cleanups]
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Pavel Emelyanov <xemul@openvz.org>
Cc: Alexey Dobriyan <adobriyan@sw.ru>
Acked-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>