Doing a mark_page_accessed at fault-time, then doing SetPageReferenced at
unmap-time if the pte is young has a number of problems.

mark_page_accessed is supposed to be roughly the equivalent of a young pte
for unmapped references. Unfortunately it doesn't come with any context:
after being called, reclaim doesn't know who or why the page was touched.

So calling mark_page_accessed not only adds extra lru or PG_referenced
manipulations for pages that are already going to have pte_young ptes anyway,
but it also adds these references which are difficult to work with from the
context of vma specific references (eg. MADV_SEQUENTIAL pte_young may not
wish to contribute to the page being referenced).

Then, simply doing SetPageReferenced when zapping a pte and finding it is
young, is not a really good solution either. SetPageReferenced does not
correctly promote the page to the active list for example. So after removing
mark_page_accessed from the fault path, several mmap()+touch+munmap() would
have a very different result from several read(2) calls for example, which
is not really desirable.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: Johannes Weiner <hannes@saeurebad.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

The KernelPageSize entry in /proc/pid/smaps is the pagesize used by the
kernel to back a VMA.  This matches the size used by the MMU in the
majority of cases.  However, one counter-example occurs on PPC64 kernels
whereby a kernel using 64K as a base pagesize may still use 4K pages for
the MMU on older processor.  To distinguish, this patch reports
MMUPageSize as the pagesize used by the MMU in /proc/pid/smaps.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: "KOSAKI Motohiro" <kosaki.motohiro@jp.fujitsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

It is useful to verify a hugepage-aware application is using the expected
pagesizes for its memory regions. This patch creates an entry called
KernelPageSize in /proc/pid/smaps that is the size of page used by the
kernel to back a VMA. The entry is not called PageSize as it is possible
the MMU uses a different size. This extension should not break any sensible
parser that skips lines containing unrecognised information.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: "KOSAKI Motohiro" <kosaki.motohiro@jp.fujitsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-2.6-dm:
  dm snapshot: extend exception store functions
  dm snapshot: split out exception store implementations
  dm snapshot: rename struct exception_store
  dm snapshot: separate out exception store interface
  dm mpath: move trigger_event to system workqueue
  dm: add name and uuid to sysfs
  dm table: rework reference counting
  dm: support barriers on simple devices
  dm request: extend target interface
  dm request: add caches
  dm ioctl: allow dm_copy_name_and_uuid to return only one field
  dm log: ensure log bitmap fits on log device
  dm log: move region_size validation
  dm log: avoid reinitialising io_req on every operation
  dm: consolidate target deregistration error handling
  dm raid1: fix error count
  dm log: fix dm_io_client leak on error paths
  dm snapshot: change yield to msleep
  dm table: drop reference at unbind

Supply dm_add_exception as a callback to the read_metadata function.
Add a status function ready for a later patch and name the functions
consistently.
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Move the existing snapshot exception store implementations out into
separate files.  Later patches will place these behind a new
interface in preparation for alternative implementations.
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Rename struct exception_store to dm_exception_store.
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Pull structures that bridge the gap between snapshot and
exception store out of dm-snap.h and put them in a new
.h file - dm-exception-store.h.  This file will define the
API for new exception stores.

Ultimately, dm-snap.h is unnecessary, since only dm-snap.c
should be using it.
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

The same workqueue is used both for sending uevents and processing queued I/O.
Deadlock has been reported in RHEL5 when sending a uevent was blocked waiting
for the queued I/O to be processed.  Use scheduled_work() for the asynchronous
uevents instead.
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Implement simple read-only sysfs entry for device-mapper block device.

This patch adds a simple sysfs directory named "dm" under block device
properties and implements
	- name attribute (string containing mapped device name)
	- uuid attribute (string containing UUID, or empty string if not set)

The kobject is embedded in mapped_device struct, so no additional
memory allocation is needed for initializing sysfs entry.

During the processing of sysfs attribute we need to lock mapped device
which is done by a new function dm_get_from_kobj, which returns the md
associated with kobject and increases the usage count.

Each 'show attribute' function is responsible for its own locking.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Rework table reference counting.

The existing code uses a reference counter. When the last reference is
dropped and the counter reaches zero, the table destructor is called.
Table reference counters are acquired/released from upcalls from other
kernel code (dm_any_congested, dm_merge_bvec, dm_unplug_all).
If the reference counter reaches zero in one of the upcalls, the table
destructor is called from almost random kernel code.

This leads to various problems:
* dm_any_congested being called under a spinlock, which calls the
  destructor, which calls some sleeping function.
* the destructor attempting to take a lock that is already taken by the
  same process.
* stale reference from some other kernel code keeps the table
  constructed, which keeps some devices open, even after successful
  return from "dmsetup remove". This can confuse lvm and prevent closing
  of underlying devices or reusing device minor numbers.

The patch changes reference counting so that the table destructor can be
called only at predetermined places.

The table has always exactly one reference from either mapped_device->map
or hash_cell->new_map. After this patch, this reference is not counted
in table->holders.  A pair of dm_create_table/dm_destroy_table functions
is used for table creation/destruction.

Temporary references from the other code increase table->holders. A pair
of dm_table_get/dm_table_put functions is used to manipulate it.

When the table is about to be destroyed, we wait for table->holders to
reach 0. Then, we call the table destructor.  We use active waiting with
msleep(1), because the situation happens rarely (to one user in 5 years)
and removing the device isn't performance-critical task: the user doesn't
care if it takes one tick more or not.

This way, the destructor is called only at specific points
(dm_table_destroy function) and the above problems associated with lazy
destruction can't happen.

Finally remove the temporary protection added to dm_any_congested().
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Implement barrier support for single device DM devices

This patch implements barrier support in DM for the common case of dm linear
just remapping a single underlying device. In this case we can safely
pass the barrier through because there can be no reordering between
devices.

 NB. Any DM device might cease to support barriers if it gets
     reconfigured so code must continue to allow for a possible
     -EOPNOTSUPP on every barrier bio submitted.  - agk
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

This patch adds the following target interfaces for request-based dm.

  map_rq    : for mapping a request

  rq_end_io : for finishing a request

  busy      : for avoiding performance regression from bio-based dm.
              Target can tell dm core not to map requests now, and
              that may help requests in the block layer queue to be
              bigger by I/O merging.
              In bio-based dm, this behavior is done by device
              drivers managing the block layer queue.
              But in request-based dm, dm core has to do that
              since dm core manages the block layer queue.
Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

This patch prepares some kmem_caches for request-based dm.
Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Allow NULL buffer in dm_copy_name_and_uuid if you only want to return one of
the fields.

(Required by a following patch that adds these fields to sysfs.)
Signed-off-by: Milan Broz <mbroz@redhat.com>
Reviewed-by: Alasdair G Kergon <agk@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Check that the log bitmap will fit within the log device.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Move log size validation from mirror target to log constructor.

Removed PAGE_SIZE restriction we no longer think necessary.
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

rw_header function updates three members of io_req data every time
when I/O is processed. bi_rw and notify.fn are never modified once
they get initialized, and so they can be set in advance.

header_to_disk() can also be pulled out of write_header() since only one
caller needs it and write_header() can be replaced by rw_header()
directly.
Signed-off-by: Takahiro Yasui <tyasui@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Change dm_unregister_target to return void and use BUG() for error
reporting.

dm_unregister_target can only fail because of programming bug in the
target driver. It can't fail because of user's behavior or disk errors.

This patch changes unregister_target to return void and use BUG if
someone tries to unregister non-registered target or unregister target
that is in use.

This patch removes code duplication (testing of error codes in all dm
targets) and reports bugs in just one place, in dm_unregister_target. In
some target drivers, these return codes were ignored, which could lead
to a situation where bugs could be missed.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Always increase the error count when I/O on a leg of a mirror fails.

The error count is used to decide whether to select an alternative
mirror leg.  If the target doesn't use the "handle_errors" feature, the
error count is not updated and the bio can get requeued forever by the
read callback.

Fix it by increasing error_count before the handle_errors feature
checking.

Cc: stable@kernel.org
Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

In create_log_context function, dm_io_client_destroy function needs
to be called, when memory allocation of disk_header, sync_bits and
recovering_bits failed, but dm_io_client_destroy is not called.

Cc: stable@kernel.org
Signed-off-by: Takahiro Yasui <tyasui@redhat.com>
Acked-by: Jonathan Brassow <jbrassow@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Change yield() to msleep(1). If the thread had realtime priority,
yield() doesn't really yield, so the yielding process would loop
indefinitely and cause machine lockup.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Move one dm_table_put() so that the last reference in the thread
gets dropped in __unbind().

This is required for a following patch,
dm-table-rework-reference-counting.patch, which will change the logic in
such a way that table destructor is called only at specific points in
the code.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

* 'for-next' of git://git.o-hand.com/linux-mfd: (30 commits)
  mfd: Fix section mismatch in da903x
  mfd: move drivers/i2c/chips/menelaus.c to drivers/mfd
  mfd: move drivers/i2c/chips/tps65010.c to drivers/mfd
  mfd: dm355evm msp430 driver
  mfd: Add missing break from wm3850-core
  mfd: Add WM8351 support
  mfd: Support configurable numbers of DCDCs and ISINKs on WM8350
  mfd: Handle missing WM8350 platform data
  mfd: Add WM8352 support
  mfd: Use irq_to_desc in twl4030 code
  power_supply: Add Dialog DA9030 battery charger driver
  mfd: Dialog DA9030 battery charger MFD driver
  mfd: Register WM8400 codec device
  mfd: Pass driver_data onto child devices
  mfd: Fix twl4030-core.c build error
  mfd: twl4030 regulator bug fixes
  mfd: twl4030: create some regulator devices
  mfd: twl4030: cleanup symbols and OMAP dependency
  mfd: twl4030: simplified child creation code
  power_supply: Add battery health reporting for WM8350
  ...

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  module: convert to stop_machine_create/destroy.
  stop_machine: introduce stop_machine_create/destroy.
  parisc: fix module loading failure of large kernel modules
  module: fix module loading failure of large kernel modules for parisc
  module: fix warning of unused function when !CONFIG_PROC_FS
  kernel/module.c: compare symbol values when marking symbols as exported in /proc/kallsyms.
  remove CONFIG_KMOD

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

* 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  swiotlb: Don't include linux/swiotlb.h twice in lib/swiotlb.c
  intel-iommu: fix build error with INTR_REMAP=y and DMAR=n
  swiotlb: add missing __init annotations

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/dlm:
  dlm: fs/dlm/ast.c: fix warning
  dlm: add new debugfs entry
  dlm: add time stamp of blocking callback
  dlm: change lock time stamping
  dlm: improve how bast mode handling
  dlm: remove extra blocking callback check
  dlm: replace schedule with cond_resched
  dlm: remove kmap/kunmap
  dlm: trivial annotation of be16 value
  dlm: fix up memory allocation flags

* 'i2c-next' of git://aeryn.fluff.org.uk/bjdooks/linux:
  i2c-omap: fix type of irq handler function
  i2c-s3c2410: Change IRQ to be plain integer.
  i2c-s3c2410: Allow more than one i2c-s3c2410 adapter
  i2c-s3c2410: Remove default platform data.
  i2c-s3c2410: Use platform data for gpio configuration
  i2c-s3c2410: Fixup style problems from checkpatch.pl
  i2c-omap: Enable I2C wakeups for 34xx
  i2c-omap: reprogram OCP_SYSCONFIG register after reset
  i2c-omap: convert 'rev1' flag to generic 'rev' u8
  i2c-omap: fix I2C timeouts due to recursive omap_i2c_{un,}idle()
  i2c-omap: Clean-up i2c-omap
  i2c-omap: Don't compile in OMAP15xx I2C ISR for non-OMAP15xx builds
  i2c-omap: Mark init-only functions as __init
  i2c-omap: Add support for omap34xx
  i2c-omap: FIFO handling support and broken hw workaround for i2c-omap
  i2c-omap: Add high-speed support to omap-i2c
  i2c-omap: Close suspected race between omap_i2c_idle() and omap_i2c_isr()
  i2c-omap: Do not use interruptible wait call in omap_i2c_xfer_msg

Fix up apparently-trivial conflict in drivers/i2c/busses/i2c-s3c2410.c

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid: (22 commits)
  HID: fix error condition propagation in hid-sony driver
  HID: fix reference count leak hidraw
  HID: add proper support for pensketch 12x9 tablet
  HID: don't allow DealExtreme usb-radio be handled by usb hid driver
  HID: fix default Kconfig setting for TopSpeed driver
  HID: driver for TopSeed Cyberlink quirky remote
  HID: make boot protocol drivers depend on EMBEDDED
  HID: avoid sparse warning in HID_COMPAT_LOAD_DRIVER
  HID: hiddev cleanup -- handle all error conditions properly
  HID: force feedback driver for GreenAsia 0x12 PID
  HID: switch specialized drivers from "default y" to !EMBEDDED
  HID: set proper dev.parent in hidraw
  HID: add dynids facility
  HID: use GFP_KERNEL in hid_alloc_buffers
  HID: usbhid, use usb_endpoint_xfer_int
  HID: move usbhid flags to usbhid.h
  HID: add n-trig digitizer support
  HID: add phys and name ioctls to hidraw
  HID: struct device - replace bus_id with dev_name(), dev_set_name()
  HID: automatically call usbhid_set_leds in usbhid driver
  ...

* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-2.6-nmw: (27 commits)
  GFS2: Use DEFINE_SPINLOCK
  GFS2: Fix use-after-free bug on umount (try #2)
  Revert "GFS2: Fix use-after-free bug on umount"
  GFS2: Streamline alloc calculations for writes
  GFS2: Send useful information with uevent messages
  GFS2: Fix use-after-free bug on umount
  GFS2: Remove ancient, unused code
  GFS2: Move four functions from super.c
  GFS2: Fix bug in gfs2_lock_fs_check_clean()
  GFS2: Send some sensible sysfs stuff
  GFS2: Kill two daemons with one patch
  GFS2: Move gfs2_recoverd into recovery.c
  GFS2: Fix "truncate in progress" hang
  GFS2: Clean up & move gfs2_quotad
  GFS2: Add more detail to debugfs glock dumps
  GFS2: Banish struct gfs2_rgrpd_host
  GFS2: Move rg_free from gfs2_rgrpd_host to gfs2_rgrpd
  GFS2: Move rg_igeneration into struct gfs2_rgrpd
  GFS2: Banish struct gfs2_dinode_host
  GFS2: Move i_size from gfs2_dinode_host and rename it to i_disksize
  ...

When using "min()", the types of both sides should match.  With the cpu
mask changes, the type of num_online_cpus() will now depend on config
options. Use "min_t()" with an explicit type instead.

And make the rx/tx case look the same too, just for sanity.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6: (30 commits)
  sparc: Fix minor SPARC32 compile error
  sparc: Remove reg*.h from Kbuild
  sparc: Clean arch-specific code in prom_common.c
  sparc: Kill asm/reg*.h
  sparc: Use 64BIT config entry
  MAINTAINERS: update sparc maintainer
  sparc: unify ipcbuf.h
  sparc: Update 64-bit defconfig.
  sparc: remove NO_PROC_ID - it is no longer used
  sparc: drop get_tbr() in traps.h
  sparc: fix warning in userspace header traps.h
  sparc: fix warnings in userspace header byteorder.h
  sparc: fix warning in userspace header jsflash.h
  sparc: unify openprom.h
  sparc64: delete unused linux_prom64_ranges from openprom_64.h
  sparc: prepare openprom for unification
  sparc: remove linux_prom_pci_assigned_addresses from openprom_32.h
  sparc: remove ebus definitions from openprom*.h
  sparc: unify siginfo.h
  sparc: unify ptrace.h
  ...

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (44 commits)
  qlge: Fix sparse warnings for tx ring indexes.
  qlge: Fix sparse warning regarding rx buffer queues.
  qlge: Fix sparse endian warning in ql_hw_csum_setup().
  qlge: Fix sparse endian warning for inbound packet control block flags.
  qlge: Fix sparse warnings for byte swapping in qlge_ethool.c
  myri10ge: print MAC and serial number on probe failure
  pkt_sched: cls_u32: Fix locking in u32_change()
  iucv: fix cpu hotplug
  af_iucv: Free iucv path/socket in path_pending callback
  af_iucv: avoid left over IUCV connections from failing connects
  af_iucv: New error return codes for connect()
  net/ehea: bitops work on unsigned longs
  Revert "net: Fix for initial link state in 2.6.28"
  tcp: Kill extraneous SPLICE_F_NONBLOCK checks.
  tcp: don't mask EOF and socket errors on nonblocking splice receive
  dccp: Integrate the TFRC library with DCCP
  dccp: Clean up ccid.c after integration of CCID plugins
  dccp: Lockless integration of CCID congestion-control plugins
  qeth: get rid of extra argument after printk to dev_* conversion
  qeth: No large send using EDDP for HiperSockets.
  ...

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ALSA: ice1724 - Fix a typo in IEC958 PCM name
  ASoC: fix davinci-sffsdr buglet
  ALSA: sound/usb: Use negated usb_endpoint_xfer_control, etc
  ALSA: hda - cxt5051 report jack state
  ALSA: hda - add basic jack reporting functions to patch_conexant.c
  ALSA: Use usb_set/get_intfdata
  ASoC: Clean up kerneldoc warnings
  ASoC: Fix pxa2xx-pcm checks for invalid DMA channels
  LSA: hda - Add HP Acacia detection
  ALSA: hda - fix name for ALC1200
  ALSA: sound/usb: use USB API functions rather than constants
  ASoC: TWL4030: DAPM based capture implementation
  ASoC: TWL4030: Make the enum filter generic for twl4030

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq:
  [CPUFREQ] Fix on resume, now preserves user policy min/max.
  [CPUFREQ] Add Celeron Core support to p4-clockmod.
  [CPUFREQ] add to speedstep-lib additional fsb values for core processors
  [CPUFREQ] Disable sysfs ui for p4-clockmod.
  [CPUFREQ] p4-clockmod: reduce noise
  [CPUFREQ] clean up speedstep-centrino and reduce cpumask_t usage

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2: (138 commits)
  ocfs2: Access the right buffer_head in ocfs2_merge_rec_left.
  ocfs2: use min_t in ocfs2_quota_read()
  ocfs2: remove unneeded lvb casts
  ocfs2: Add xattr support checking in init_security
  ocfs2: alloc xattr bucket in ocfs2_xattr_set_handle
  ocfs2: calculate and reserve credits for xattr value in mknod
  ocfs2/xattr: fix credits calculation during index create
  ocfs2/xattr: Always updating ctime during xattr set.
  ocfs2/xattr: Remove extend_trans call and add its credits from the beginning
  ocfs2/dlm: Fix race during lockres mastery
  ocfs2/dlm: Fix race in adding/removing lockres' to/from the tracking list
  ocfs2/dlm: Hold off sending lockres drop ref message while lockres is migrating
  ocfs2/dlm: Clean up errors in dlm_proxy_ast_handler()
  ocfs2/dlm: Fix a race between migrate request and exit domain
  ocfs2: One more hamming code optimization.
  ocfs2: Another hamming code optimization.
  ocfs2: Don't hand-code xor in ocfs2_hamming_encode().
  ocfs2: Enable metadata checksums.
  ocfs2: Validate superblock with checksum and ecc.
  ocfs2: Checksum and ECC for directory blocks.
  ...

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  inotify: fix type errors in interfaces
  fix breakage in reiserfs_new_inode()
  fix the treatment of jfs special inodes
  vfs: remove duplicate code in get_fs_type()
  add a vfs_fsync helper
  sys_execve and sys_uselib do not call into fsnotify
  zero i_uid/i_gid on inode allocation
  inode->i_op is never NULL
  ntfs: don't NULL i_op
  isofs check for NULL ->i_op in root directory is dead code
  affs: do not zero ->i_op
  kill suid bit only for regular files
  vfs: lseek(fd, 0, SEEK_CUR) race condition

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>

Warnings:
drivers/net/qlge/qlge_main.c:1474:34: warning: restricted degrades to integer
drivers/net/qlge/qlge_main.c:1475:36: warning: restricted degrades to integer
drivers/net/qlge/qlge_main.c:1592:51: warning: restricted degrades to integer
drivers/net/qlge/qlge_main.c:1941:20: warning: incorrect type in assignment (different base types)
drivers/net/qlge/qlge_main.c:1941:20:    expected restricted unsigned int [usertype] tid
drivers/net/qlge/qlge_main.c:1941:20:    got int [signed] index
drivers/net/qlge/qlge_main.c:1945:24: warning: incorrect type in assignment (different base types)
drivers/net/qlge/qlge_main.c:1945:24:    expected restricted unsigned int [usertype] txq_idx
drivers/net/qlge/qlge_main.c:1945:24:    got unsigned int [unsigned] [usertype] tx_ring_idx
Signed-off-by: Ron Mercer <ron.mercer@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>