An XFS workload showed up a bug in the lockless pagecache patch. Basically it
would go into an "infinite" loop, although it would sometimes be able to break
out of the loop! The reason is a missing compiler barrier in the "increment
reference count unless it was zero" case of the lockless pagecache protocol in
the gang lookup functions.

This would cause the compiler to use a cached value of struct page pointer to
retry the operation with, rather than reload it. So the page might have been
removed from pagecache and freed (refcount==0) but the lookup would not correctly
notice the page is no longer in pagecache, and keep attempting to increment the
refcount and failing, until the page gets reallocated for something else. This
isn't a data corruption because the condition will be detected if the page has
been reallocated. However it can result in a lockup.

Linus points out that ACCESS_ONCE is also required in that pointer load, even
if it's absence is not causing a bug on our particular build. The most general
way to solve this is just to put an rcu_dereference in radix_tree_deref_slot.

Assembly of find_get_pages,
before:
.L220:
        movq    (%rbx), %rax    #* ivtmp.1162, tmp82
        movq    (%rax), %rdi    #, prephitmp.1149
.L218:
        testb   $1, %dil        #, prephitmp.1149
        jne     .L217   #,
        testq   %rdi, %rdi      # prephitmp.1149
        je      .L203   #,
        cmpq    $-1, %rdi       #, prephitmp.1149
        je      .L217   #,
        movl    8(%rdi), %esi   # <variable>._count.counter, c
        testl   %esi, %esi      # c
        je      .L218   #,

after:
.L212:
        movq    (%rbx), %rax    #* ivtmp.1109, tmp81
        movq    (%rax), %rdi    #, ret
        testb   $1, %dil        #, ret
        jne     .L211   #,
        testq   %rdi, %rdi      # ret
        je      .L197   #,
        cmpq    $-1, %rdi       #, ret
        je      .L211   #,
        movl    8(%rdi), %esi   # <variable>._count.counter, c
        testl   %esi, %esi      # c
        je      .L212   #,

(notice the obvious infinite loop in the first example, if page->count remains 0)

Signed-off-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

... if you revert a commit, revert the fixups elsewhere that had been
triggered by it.  Such as 8c56250f
(lockdep, UML: fix compilation when CONFIG_TRACE_IRQFLAGS_SUPPORT is not set).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

We need to make asm-offsets.h contents visible for objects built
with userland headers.  Instead of creating a symlink, just have the
file with equivalent include (relative to location of header) created
once.  That kills the last symlink used in arch/um builds.

Additionally, both generated headers can become dependencies of
archprepare now, killing the misuse of prepare.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* 'audit.b61' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit-current:
  audit: validate comparison operations, store them in sane form
  clean up audit_rule_{add,del} a bit
  make sure that filterkey of task,always rules is reported
  audit rules ordering, part 2
  fixing audit rule ordering mess, part 1
  audit_update_lsm_rules() misses the audit_inode_hash[] ones
  sanitize audit_log_capset()
  sanitize audit_fd_pair()
  sanitize audit_mq_open()
  sanitize AUDIT_MQ_SENDRECV
  sanitize audit_mq_notify()
  sanitize audit_mq_getsetattr()
  sanitize audit_ipc_set_perm()
  sanitize audit_ipc_obj()
  sanitize audit_socketcall
  don't reallocate buffer in every audit_sockaddr()

Add standard interfaces for alarm/update irqs enabling.  Drivers are no
more required to implement equivalent ioctl code as rtc-dev will provide
it.

UIE emulation should now be handled correctly and will work even for those
RTC drivers who cannot be configured to do both UIE and AIE.

Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Cc: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
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>

With the write_begin/write_end aops, page_symlink was broken because it
could no longer pass a GFP_NOFS type mask into the point where the
allocations happened.  They are done in write_begin, which would always
assume that the filesystem can be entered from reclaim.  This bug could
cause filesystem deadlocks.

The funny thing with having a gfp_t mask there is that it doesn't really
allow the caller to arbitrarily tinker with the context in which it can be
called.  It couldn't ever be GFP_ATOMIC, for example, because it needs to
take the page lock.  The only thing any callers care about is __GFP_FS
anyway, so turn that into a single flag.

Add a new flag for write_begin, AOP_FLAG_NOFS.  Filesystems can now act on
this flag in their write_begin function.  Change __grab_cache_page to
accept a nofs argument as well, to honour that flag (while we're there,
change the name to grab_cache_page_write_begin which is more instructive
and does away with random leading underscores).

This is really a more flexible way to go in the end anyway -- if a
filesystem happens to want any extra allocations aside from the pagecache
ones in ints write_begin function, it may now use GFP_KERNEL (rather than
GFP_NOFS) for common case allocations (eg.  ocfs2_alloc_write_ctxt, for a
random example).

[kosaki.motohiro@jp.fujitsu.com: fix ubifs]
[kosaki.motohiro@jp.fujitsu.com: fix fuse]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: <stable@kernel.org>		[2.6.28.x]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Cleaned up the calling convention: just pass in the AOP flags
  untouched to the grab_cache_page_write_begin() function.  That
  just simplifies everybody, and may even allow future expansion of the
  logic.   - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The function viafb_cursor() uses 2 stack-variables of CURSOR_SIZE bits;
CURSOR_SIZE is defined as (8 * 1024).  Using up twice 1k on stack is too
much for 4k-stack (though it works with 8k-stacks).  Make those two
variables kzalloc'ed to preserve stack space.

Also merge the whole lot of local struct's in viafb_ioctl into a union so
the stack usage gets minimized here as well.  (struct's are only accessed
in their indicidual IOCTL case) This second part is only compile-tested as
I know of no userspace app using the IOCTLs.

Signed-off-by: Bruno Prémont <bonbons@linux-vserver.org>
Cc: <JosephChan@via.com.tw>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

As suggested by Andreas Dilger, introduce a bgl_lock_ptr() helper in
<linux/blockgroup_lock.h> and add separate sb_bgl_lock() helpers to
filesystem specific header files to break the hidden dependency to
struct ext[234]_sb_info.

Also, while at it, convert the macros to static inlines to try make up
for all the times I broke Andrew Morton's tree.

Acked-by: Andreas Dilger <adilger@sun.com>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Include header files as used/needed:

  In file included from drivers/leds/leds-dac124s085.c:16:
  include/linux/spi/spi.h:66: error: field 'dev' has incomplete type
  include/linux/spi/spi.h: In function 'to_spi_device':
  include/linux/spi/spi.h:100: warning: type defaults to 'int' in declaration of '__mptr'
  ...

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The flush_cache_vmap in vmap_page_range() is called with the end of the
range twice.  The following patch fixes this for me.

Signed-off-by: Adam Lackorzynski <adam@os.inf.tu-dresden.de>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The race is calling cgroup_clone() while umounting the ns cgroup subsys,
and thus cgroup_clone() might access invalid cgroup_fs, or kill_sb() is
called after cgroup_clone() created a new dir in it.

The BUG I triggered is BUG_ON(root->number_of_cgroups != 1);

  ------------[ cut here ]------------
  kernel BUG at kernel/cgroup.c:1093!
  invalid opcode: 0000 [#1] SMP
  ...
  Process umount (pid: 5177, ti=e411e000 task=e40c4670 task.ti=e411e000)
  ...
  Call Trace:
   [<c0493df7>] ? deactivate_super+0x3f/0x51
   [<c04a3600>] ? mntput_no_expire+0xb3/0xdd
   [<c04a3ab2>] ? sys_umount+0x265/0x2ac
   [<c04a3b06>] ? sys_oldumount+0xd/0xf
   [<c0403911>] ? sysenter_do_call+0x12/0x31
  ...
  EIP: [<c0456e76>] cgroup_kill_sb+0x23/0xe0 SS:ESP 0068:e411ef2c
  ---[ end trace c766c1be3bf944ac ]---

Cc: Serge E. Hallyn <serue@us.ibm.com>
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Cc: "Serge E. Hallyn" <serue@us.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Don't store the field->op in the messy (and very inconvenient for e.g.
audit_comparator()) form; translate to dense set of values and do full
validation of userland-submitted value while we are at it.

->audit_init_rule() and ->audit_match_rule() get new values now; in-tree
instances updated.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Fix the actual rule listing; add per-type lists _not_ used for matching,
with all exit,... sitting on one such list.  Simplifies "do something
for all rules" logics, while we are at it...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Problem: ordering between the rules on exit chain is currently lost;
all watch and inode rules are listed after everything else _and_
exit,never on one kind doesn't stop exit,always on another from
being matched.

Solution: assign priorities to rules, keep track of the current
highest-priority matching rule and its result (always/never).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* no allocations
* return void
* don't duplicate checked for dummy context

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* no allocations
* return void

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* don't bother with allocations
* don't do double copy_from_user()
* don't duplicate parts of check for audit_dummy_context()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* logging the original value of *msg_prio in mq_timedreceive(2)
  is insane - the argument is write-only (i.e. syscall always
  ignores the original value and only overwrites it).
* merge __audit_mq_timed{send,receive}
* don't do copy_from_user() twice
* don't mess with allocations in auditsc part
* ... and don't bother checking !audit_enabled and !context in there -
  we'd already checked for audit_dummy_context().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* don't copy_from_user() twice
* don't bother with allocations
* don't duplicate parts of audit_dummy_context()
* make it return void

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* get rid of allocations
* make it return void
* don't duplicate parts of audit_dummy_context()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* get rid of allocations
* make it return void
* simplify callers

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* get rid of allocations
* make it return void
* simplify callers

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

* don't bother with allocations
* now that it can't fail, make it return void

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

No need to do that more than once per process lifetime; allocating/freeing
on each sendto/accept/etc. is bloody pointless.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Merge branch 'cpus4096-for-linus-3' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip

* 'cpus4096-for-linus-3' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (77 commits)
  x86: setup_per_cpu_areas() cleanup
  cpumask: fix compile error when CONFIG_NR_CPUS is not defined
  cpumask: use alloc_cpumask_var_node where appropriate
  cpumask: convert shared_cpu_map in acpi_processor* structs to cpumask_var_t
  x86: use cpumask_var_t in acpi/boot.c
  x86: cleanup some remaining usages of NR_CPUS where s/b nr_cpu_ids
  sched: put back some stack hog changes that were undone in kernel/sched.c
  x86: enable cpus display of kernel_max and offlined cpus
  ia64: cpumask fix for is_affinity_mask_valid()
  cpumask: convert RCU implementations, fix
  xtensa: define __fls
  mn10300: define __fls
  m32r: define __fls
  h8300: define __fls
  frv: define __fls
  cris: define __fls
  cpumask: CONFIG_DISABLE_OBSOLETE_CPUMASK_FUNCTIONS
  cpumask: zero extra bits in alloc_cpumask_var_node
  cpumask: replace for_each_cpu_mask_nr with for_each_cpu in kernel/time/
  cpumask: convert mm/
  ...

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-iommu: (89 commits)
  AMD IOMMU: remove now unnecessary #ifdefs
  AMD IOMMU: prealloc_protection_domains should be static
  kvm/iommu: fix compile warning
  AMD IOMMU: add statistics about total number of map requests
  AMD IOMMU: add statistics about allocated io memory
  AMD IOMMU: add stats counter for domain tlb flushes
  AMD IOMMU: add stats counter for single iommu domain tlb flushes
  AMD IOMMU: add stats counter for cross-page request
  AMD IOMMU: add stats counter for free_coherent requests
  AMD IOMMU: add stats counter for alloc_coherent requests
  AMD IOMMU: add stats counter for unmap_sg requests
  AMD IOMMU: add stats counter for map_sg requests
  AMD IOMMU: add stats counter for unmap_single requests
  AMD IOMMU: add stats counter for map_single requests
  AMD IOMMU: add stats counter for completion wait events
  AMD IOMMU: add init code for statistic collection
  AMD IOMMU: add necessary header defines for stats counting
  AMD IOMMU: add Kconfig entry for statistic collection code
  AMD IOMMU: use dev_name in iommu_enable function
  AMD IOMMU: use calc_devid in prealloc_protection_domains
  ...

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6: (34 commits)
  V4L/DVB (10173): Missing v4l2_prio_close in radio_release
  V4L/DVB (10172): add DVB_DEVICE_TYPE= to uevent
  V4L/DVB (10171): Use usb_set_intfdata
  V4L/DVB (10170): tuner-simple: prevent possible OOPS caused by divide by zero error
  V4L/DVB (10168): sms1xxx: fix inverted gpio for lna control on tiger r2
  V4L/DVB (10167): sms1xxx: add support for inverted gpio
  V4L/DVB (10166): dvb frontend: stop using non-C99 compliant comments
  V4L/DVB (10165): Add FE_CAN_2G_MODULATION flag to frontends that support DVB-S2
  V4L/DVB (10164): Add missing S2 caps flag to S2API
  V4L/DVB (10163): em28xx: allocate adev together with struct em28xx dev
  V4L/DVB (10162): tuner-simple: Fix tuner type set message
  V4L/DVB (10161): saa7134: fix autodetection for AVer TV GO 007 FM Plus
  V4L/DVB (10160): em28xx: update chip id for em2710
  V4L/DVB (10157): Add USB ID for the Sil4701 radio from DealExtreme
  V4L/DVB (10156): saa7134: Add support for Avermedia AVer TV GO 007 FM Plus
  V4L/DVB (10155): Add TEA5764 radio driver
  V4L/DVB (10154): saa7134: fix a merge conflict on Behold H6 board
  V4L/DVB (10153): Add the Beholder H6 card to DVB-T part of sources.
  V4L/DVB (10152): Change configuration of the Beholder H6 card
  V4L/DVB (10151): Fix I2C bridge error in zl10353
  ...

those two functions only used in that C file

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc:
  mmc: warn about voltage mismatches
  mmc_spi: Add support for OpenFirmware bindings
  pxamci: fix dma_unmap_sg length
  mmc_block: ensure all sectors that do not have errors are read
  drivers/mmc: Move a dereference below a NULL test
  sdhci: handle built-in sdhci with modular leds class
  mmc: balanc pci_iomap with pci_iounmap
  mmc_block: print better error messages
  mmc: Add mmc_vddrange_to_ocrmask() helper function
  ricoh_mmc: Handle newer models of Ricoh controllers
  mmc: Add 8-bit bus width support
  sdhci: activate led support also when module
  mmc: trivial annotation of 'blocks'
  pci: use pci_ioremap_bar() in drivers/mmc
  sdricoh_cs: Add support for Bay Controller devices
  mmc: at91_mci: reorder timer setup and mmc_add_host() call

* git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog:
  [WATCHDOG] Add support for the WM8350 watchdog
  [WATCHDOG] Add SMSC SCH311x Watchdog Timer.
  [WATCHDOG] ib700wdt - add timeout parameter

* 'cputime' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [PATCH] fast vdso implementation for CLOCK_THREAD_CPUTIME_ID
  [PATCH] improve idle cputime accounting
  [PATCH] improve precision of idle time detection.
  [PATCH] improve precision of process accounting.
  [PATCH] idle cputime accounting
  [PATCH] fix scaled & unscaled cputime accounting

Before, when we only ever printed out the pointer value itself, a NULL
pointer would never cause issues and might as well be printed out as
just its numeric value.

However, with the extended %p formats, especially %pR, we might validly
want to print out resources for debugging.  And sometimes they don't
even exist, and the resource pointer is just NULL.  Print it out as
such, rather than oopsing.

This is a more generic version of a patch done by Trent Piepho (catching
all %p cases rather than just %pR, and using "(null)" instead of
"[NULL]" to match glibc).

Requested-by: Trent Piepho <xyzzy@speakeasy.org>
Acked-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

... just make it a binfmt handler like #! one.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

They are actually alpha vs.  i386/arm/m68k i.e. ecoff vs. aout.

In the only place where we actually tried to handle arm and i386/m68k in
different ways (START_DATA() in coredump handling), the arm variant
works for all of them (i386 and m68k have u.start_code set to 0).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>