On Cortex-A15 (r0p0..r3p2) the TLBI/DSB are not adequately shooting down
all use of the old entries. This patch implements the erratum workaround
which consists of:

1. Dummy TLBIMVAIS and DSB on the CPU doing the TLBI operation.
2. Send IPI to the CPUs that are running the same mm (and ASID) as the
   one being invalidated (or all the online CPUs for global pages).
3. CPU receiving the IPI executes a DMB and CLREX (part of the exception
   return code already).
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

Commit b8db6b88 (ARM: 7547/4: cache-l2x0: add support for Aurora L2 cache
ctrl) moved the masking of the part ID which caused the RTL version to be
lost. Commit 6248d060 (ARM: 7545/1: cache-l2x0: make outer_cache_fns a
field of l2x0_of_data) changed how .set_debug is initialized. Both commits
break commit 74ddcdb8 (ARM: 7608/1: l2x0: Only set .set_debug
on PL310 r3p0 and earlier) which uses the RTL version to conditionally set
.set_debug function pointer. Commit b8db6b88 also caused the printed cache
ID to be missing the version information.

Fix this by reverting how the part number is masked so the RTL version
info is maintained. The cache-id-part DT property does not set the RTL
bits so masking them should have no effect. Also, re-arrange the order
of the function pointer init so the .set_debug function can be overridden.
Reported-by: Paolo Pisati <paolo.pisati@canonical.com>
Signed-off-by: Rob Herring <rob.herring@calxeda.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Cc: Yehuda Yitschak <yehuday@marvell.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

Jason Cooper reports these build errors:
arch/arm/kernel/built-in.o: In function `iwmmxt_do':
/.../arch/arm/kernel/pj4-cp0.c:36: undefined reference to `iwmmxt_task_release'
/.../arch/arm/kernel/pj4-cp0.c:40: undefined reference to `iwmmxt_task_switch'
make: *** [vmlinux] Error 1

This is caused because the PJ4 code explicitly references the iWMMXt
code, but doesn't require it to be built.  Fix this by ensuring that
iWMMXt is always enabled with PJ4.
Reported-by: Jason Cooper <jason@lakedaemon.net>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

CPU debug features like hardware break, watchpoints can be used only
when the debug mode is enabled and available. Unfortunately on OMAP4
based devices, after a CPU power cycle, the debug feature gets disabled
which leads to a flood of messages coming from reset_ctrl_regs() which
gets called on every CPU_PM_EXIT with CPUidle enabled.

So make use of warn_once() so that system is usable.

Thanks to Will for pointers and Lokesh for the analysis of the issue.
Tested-by: Lokesh Vutla <lokeshvutla@ti.com>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

Some early versions of the Krait CPU design incorrectly indicate
that they only support the UDIV and SDIV instructions in Thumb
mode when they actually support them in ARM and Thumb mode. It
seems that these CPUs follow the DDI0406B ARM ARM which has two
possible values for the divide instructions field, instead of the
DDI0406C document which has three possible values.

Work around this problem by checking the MIDR against Krait CPUs
with this faulty ISAR0 register and force the hwcaps to indicate
support in both modes.

[sboyd: Rewrote commit text to reflect real reasoning now that
	we autodetect udiv/sdiv]
Signed-off-by: Stepan Moskovchenko <stepanm@codeaurora.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

The ISAR0 register indicates support for the SDIV and UDIV
instructions in both the Thumb and ARM instruction set. Read the
register to detect the supported instructions and update the
elf_hwcap mask as appropriate. This is better than adding more
and more cpuid checks in proc-v7.S for each new cpu variant that
supports these instructions.
Acked-by: Will Deacon <will.deacon@arm.com>
Cc: Stepan Moskovchenko <stepanm@codeaurora.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

Don't advertise support for the SDIV/UDIV thumb instructions if
the kernel is not compiled with support for thumb userspace. This
is in line with how we remove the THUMB hwcap in these
configurations.
Acked-by: Will Deacon <will.deacon@arm.com>
Cc: Stepan Moskovchenko <stepanm@codeaurora.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

With LPAE enabled, alloc_init_section() does not map the entire
address space for unaligned addresses.

The issue also reproduced with CMA + LPAE. CMA tries to map 16MB
with page granularity mappings during boot. alloc_init_pte()
is called and out of 16MB, only 2MB gets mapped and rest remains
unaccessible.

Because of this OMAP5 boot is broken with CMA + LPAE enabled.
Fix the issue by ensuring that the entire addresses are
mapped.
Signed-off-by: R Sricharan <r.sricharan@ti.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoffer Dall <chris@cloudcar.com>
Cc: Santosh Shilimkar <santosh.shilimkar@ti.com>
Tested-by: Laura Abbott <lauraa@codeaurora.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Christoffer Dall <chris@cloudcar.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

Merge branch 'kvm-arm/vgic-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms into fixes

Pull CIFS fixes from Steve French:
 "Three small CIFS Fixes (the most important of the three fixes a recent
  problem authenticating to Windows 8 using cifs rather than SMB2)"

* 'for-next' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: ignore everything in SPNEGO blob after mechTypes
  cifs: delay super block destruction until all cifsFileInfo objects are gone
  cifs: map NT_STATUS_SHARING_VIOLATION to EBUSY instead of ETXTBSY

Pull ext4 fixes from Ted Ts'o:
 "Fix a number of regression and other bugs in ext4, most of which were
  relatively obscure cornercases or races that were found using
  regression tests."

* tag 'ext4_for_linue' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (21 commits)
  ext4: fix data=journal fast mount/umount hang
  ext4: fix ext4_evict_inode() racing against workqueue processing code
  ext4: fix memory leakage in mext_check_coverage
  ext4: use s_extent_max_zeroout_kb value as number of kb
  ext4: use atomic64_t for the per-flexbg free_clusters count
  jbd2: fix use after free in jbd2_journal_dirty_metadata()
  ext4: reserve metadata block for every delayed write
  ext4: update reserved space after the 'correction'
  ext4: do not use yield()
  ext4: remove unused variable in ext4_free_blocks()
  ext4: fix WARN_ON from ext4_releasepage()
  ext4: fix the wrong number of the allocated blocks in ext4_split_extent()
  ext4: update extent status tree after an extent is zeroed out
  ext4: fix wrong m_len value after unwritten extent conversion
  ext4: add self-testing infrastructure to do a sanity check
  ext4: avoid a potential overflow in ext4_es_can_be_merged()
  ext4: invalidate extent status tree during extent migration
  ext4: remove unnecessary wait for extent conversion in ext4_fallocate()
  ext4: add warning to ext4_convert_unwritten_extents_endio
  ext4: disable merging of uninitialized extents
  ...

We've had several reports of people attempting to mount Windows 8 shares
and getting failures with a return code of -EINVAL. The default sec=
mode changed recently to sec=ntlmssp. With that, we expect and parse a
SPNEGO blob from the server in the NEGOTIATE reply.

The current decode_negTokenInit function first parses all of the
mechTypes and then tries to parse the rest of the negTokenInit reply.
The parser however currently expects a mechListMIC or nothing to follow the
mechTypes, but Windows 8 puts a mechToken field there instead to carry
some info for the new NegoEx stuff.

In practice, we don't do anything with the fields after the mechTypes
anyway so I don't see any real benefit in continuing to parse them.
This patch just has the kernel ignore the fields after the mechTypes.
We'll probably need to reinstate some of this if we ever want to support
NegoEx.
Reported-by: Jason Burgess <jason@jacknife2.dns2go.com>
Reported-by: Yan Li <elliot.li.tech@gmail.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Steve French <sfrench@us.ibm.com>

Pull thermal management fixes from Zhang Rui.

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux:
  thermal: exynos_thermal: return a proper error code while thermal_zone_device_register fail.
  thermal: rcar_thermal: propagate return value of thermal_zone_device_register
  Thermal: kirkwood: Convert to devm_ioremap_resource()
  Thermal: rcar: Convert to devm_ioremap_resource()
  Thermal: dove: Convert to devm_ioremap_resource()
  thermal: rcar: fix missing unlock on error in rcar_thermal_update_temp()

Pull perf fixes from Ingo Molnar:
 "A fair chunk of the linecount comes from a fix for a tracing bug that
  corrupts latency tracing buffers when the overwrite mode is changed on
  the fly - the rest is mostly assorted fewliner fixlets."

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86: Add SNB/SNB-EP scheduling constraints for cycle_activity event
  kprobes/x86: Check Interrupt Flag modifier when registering probe
  kprobes: Make hash_64() as always inlined
  perf: Generate EXIT event only once per task context
  perf: Reset hwc->last_period on sw clock events
  tracing: Prevent buffer overwrite disabled for latency tracers
  tracing: Keep overwrite in sync between regular and snapshot buffers
  tracing: Protect tracer flags with trace_types_lock
  perf tools: Fix LIBNUMA build with glibc 2.12 and older.
  tracing: Fix free of probe entry by calling call_rcu_sched()
  perf/POWER7: Create a sysfs format entry for Power7 events
  perf probe: Fix segfault
  libtraceevent: Remove hard coded include to /usr/local/include in Makefile
  perf record: Fix -C option
  perf tools: check if -DFORTIFY_SOURCE=2 is allowed
  perf report: Fix build with NO_NEWT=1
  perf annotate: Fix build with NO_NEWT=1
  tracing: Fix race in snapshot swapping

Pull drm fixes from Dave Airlie:
 "Radeon, intel and nouveau, along with one mgag200 fix

   - intel fix for an ioctl overflow, along with a regression fix for
     some phantom irqs on Ironlake.
   - nouveau has a lockdep warning and a bunch of thermal fixes
   - radeon has new pci ids and some minor fixes."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux: (26 commits)
  drm/mgag200: Bug fix: Modified pll algorithm for EH project
  drm/i915: stop using GMBUS IRQs on Gen4 chips
  drm/nv50/kms: prevent lockdep false-positive in page flipping path
  drm/nouveau/core: fix return value of nouveau_object_del()
  MAINTAINERS: intel-gfx is no longer subscribers-only
  drm/i915: Use the fixed pixel clock for eDP in intel_dp_set_m_n()
  drm/nouveau/hwmon: do not expose a buggy temperature if it is unavailable
  drm/nouveau/therm: display the availability of the internal sensor
  drm/nouveau/therm: disable temperature management if the sensor isn't readable
  drm/nouveau/therm: disable auto fan management if temperature is not available
  drm/nv40/therm: reserve negative temperatures for errors
  drm/nv40/therm: disable temperature reading if the bios misses some parameters
  drm/nouveau/therm-ic: the temperature is off by sensor_constant, warn the user
  drm/nouveau/therm: remove some confusion introduced by therm_mode
  drm/nouveau/therm: do not make assumptions on temperature
  drm/nv40/therm: increase the sensor's settling delay to 20ms
  drm/nv40/therm: improve selection between the old and the new style
  Revert "drm/i915: try to train DP even harder"
  drm/radeon: add Richland pci ids
  drm/radeon: add support for Richland APUs
  ...

Pull device-mapper fixes from Alasdair G Kergon:
 "Fix reported data loss with discards and thin snapshots; avoid a
  deadlock observed in dm verity; fix a race in the new dm cache code
  along with some other minor bugs; store the cache policy version on
  disk to make the stored hints format future-proof."

* tag 'dm-3.9-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm:
  dm cache: policy ignore hints if generated by different version
  dm cache: policy change version from string to integer set
  dm cache: fix race in writethrough implementation
  dm cache: metadata clear dirty bits on clean shutdown
  dm cache: avoid calling policy destructor twice on error
  dm cache: detect cache_create failure
  dm cache: avoid 64 bit division on 32 bit
  dm verity: avoid deadlock
  dm thin: fix non power of two discard granularity calc
  dm thin: fix discard corruption

Daniel writes:
Bunch of fixes, all pretty high-priority
- Fix execbuf argument checking (Kees Cook)
- Optionally obfuscate kernel addresses in dumps (Kees Cook)
- Two patches from Takashi Iwai to fix DP link training regressions he's
  seen.
- intel-gfx is no longer subscribers-only (well, just no longer moderated
  in an annoying way for non-subscribers), update MAINTAINERS
- gm45 gmbus irq fallout fix (Jiri Kosina)

* 'drm-intel-fixes' of git://people.freedesktop.org/~danvet/drm-intel:
  drm/i915: stop using GMBUS IRQs on Gen4 chips
  MAINTAINERS: intel-gfx is no longer subscribers-only
  drm/i915: Use the fixed pixel clock for eDP in intel_dp_set_m_n()
  Revert "drm/i915: try to train DP even harder"
  drm/i915: bounds check execbuffer relocation count
  drm/i915: restrict kernel address leak in debugfs

While testing the mgag200 kms driver on the HP ProLiant Gen8, a
bug was seen.  Once the bootloader would load the selected kernel,
the screen would go black.  At first it was assumed that the
mgag200 kms driver was hanging.  But after setting up the grub
serial output, it was seen that the driver was being loaded
properly.  After trying serval monitors, one finaly displayed
the message "Frequency Out of Range".  By comparing the kms pll
algorithm with the previous mgag200 xorg driver pll algorithm,
discrepencies were found.  Once the kms pll algorithm was
modified, the expected pll values were produced.  This fix was
tested on several monitors of varying native resolutions.
Signed-off-by: Julia Lemire <jlemire@matrox.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>

When reading the dm cache metadata from disk, ignore the policy hints
unless they were generated by the same major version number of the same
policy module.

The hints are considered to be private data belonging to the specific
module that generated them and there is no requirement for them to make
sense to different versions of the policy that generated them.
Policy modules are all required to work fine if no previous hints are
supplied (or if existing hints are lost).
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Separate dm cache policy version string into 3 unsigned numbers
corresponding to major, minor and patchlevel and store them at the end
of the on-disk metadata so we know which version of the policy generated
the hints in case a future version wants to use them differently.
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

We have found a race in the optimisation used in the dm cache
writethrough implementation.  Currently, dm core sends the cache target
two bios, one for the origin device and one for the cache device and
these are processed in parallel.  This patch avoids the race by
changing the code back to a simpler (slower) implementation which
processes the two writes in series, one after the other, until we can
develop a complete fix for the problem.

When the cache is in writethrough mode it needs to send WRITE bios to
both the origin and cache devices.

Previously we've been implementing this by having dm core query the
cache target on every write to find out how many copies of the bio it
wants.  The cache will ask for two bios if the block is in the cache,
and one otherwise.

Then main problem with this is it's racey.  At the time this check is
made the bio hasn't yet been submitted and so isn't being taken into
account when quiescing a block for migration (promotion or demotion).
This means a single bio may be submitted when two were needed because
the block has since been promoted to the cache (catastrophic), or two
bios where only one is needed (harmless).

I really don't want to start entering bios into the quiescing system
(deferred_set) in the get_num_write_bios callback.  Instead this patch
simplifies things; only one bio is submitted by the core, this is
first written to the origin and then the cache device in series.
Obviously this will have a latency impact.

deferred_writethrough_bios is introduced to record bios that must be
later issued to the cache device from the worker thread.  This deferred
submission, after the origin bio completes, is required given that we're
in interrupt context (writethrough_endio).
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

When writing the dirty bitset to the metadata device on a clean
shutdown, clear the dirty bits.  Previously they were left indicating
the cache was dirty. This led to confusion about whether there really
was dirty data in the cache or not.  (This was a harmless bug.)
Reported-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

If the cache policy's config values are not able to be set we must
set the policy to NULL after destroying it in create_cache_policy()
so we don't attempt to destroy it a second time later.
Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Return error if cache_create() fails.

A missing return check made cache_ctr continue even after an error in
cache_create() resulting in the cache object being destroyed.  So a
simple failure like an odd number of cache policy config value arguments
would result in an oops.
Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Squash various 32bit link errors.

  >> on i386:
  >> drivers/built-in.o: In function `is_discarded_oblock':
  >> dm-cache-target.c:(.text+0x1ea28e): undefined reference to `__udivdi3'
  ...
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

A deadlock was found in the prefetch code in the dm verity map
function.  This patch fixes this by transferring the prefetch
to a worker thread and skipping it completely if kmalloc fails.

If generic_make_request is called recursively, it queues the I/O
request on the current->bio_list without making the I/O request
and returns. The routine making the recursive call cannot wait
for the I/O to complete.

The deadlock occurs when one thread grabs the bufio_client
mutex and waits for an I/O to complete but the I/O is queued
on another thread's current->bio_list and is waiting to get
the mutex held by the first thread.

The fix recognises that prefetching is not essential.  If memory
can be allocated, it queues the prefetch request to the worker thread,
but if not, it does nothing.
Signed-off-by: Paul Taysom <taysom@chromium.org>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
Cc: stable@kernel.org

Fix a discard granularity calculation to work for non power of 2 block sizes.

In order for thinp to passdown discard bios to the underlying data
device, the data device must have a discard granularity that is a
factor of the thinp block size.  Originally this check was done by
using bitops since the block_size was known to be a power of two.

Introduced by commit f13945d7
("dm thin: support a non power of 2 discard_granularity").
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

Fix a bug in dm_btree_remove that could leave leaf values with incorrect
reference counts.  The effect of this was that removal of a shared block
could result in the space maps thinking the block was no longer used.
More concretely, if you have a thin device and a snapshot of it, sending
a discard to a shared region of the thin could corrupt the snapshot.

Thinp uses a 2-level nested btree to store it's mappings.  This first
level is indexed by thin device, and the second level by logical
block.

Often when we're removing an entry in this mapping tree we need to
rebalance nodes, which can involve shadowing them, possibly creating a
copy if the block is shared.  If we do create a copy then children of
that node need to have their reference counts incremented.  In this
way reference counts percolate down the tree as shared trees diverge.

The rebalance functions were incrementing the children at the
appropriate time, but they were always assuming the children were
internal nodes.  This meant the leaf values (in our case packed
block/flags entries) were not being incremented.

Cc: stable@vger.kernel.org
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>

In data=journal mode, if we unmount the file system before a
transaction has a chance to complete, when the journal inode is being
evicted, we can end up calling into jbd2_log_wait_commit() for the
last transaction, after the journalling machinery has been shut down.

Arguably we should adjust ext4_should_journal_data() to return FALSE
for the journal inode, but the only place it matters is
ext4_evict_inode(), and so to save a bit of CPU time, and to make the
patch much more obviously correct by inspection(tm), we'll fix it by
explicitly not trying to waiting for a journal commit when we are
evicting the journal inode, since it's guaranteed to never succeed in
this case.

This can be easily replicated via: 

     mount -t ext4 -o data=journal /dev/vdb /vdb ; umount /vdb

------------[ cut here ]------------
WARNING: at /usr/projects/linux/ext4/fs/jbd2/journal.c:542 __jbd2_log_start_commit+0xba/0xcd()
Hardware name: Bochs
JBD2: bad log_start_commit: 3005630206 3005630206 0 0
Modules linked in:
Pid: 2909, comm: umount Not tainted 3.8.0-rc3 #1020
Call Trace:
 [<c015c0ef>] warn_slowpath_common+0x68/0x7d
 [<c02b7e7d>] ? __jbd2_log_start_commit+0xba/0xcd
 [<c015c177>] warn_slowpath_fmt+0x2b/0x2f
 [<c02b7e7d>] __jbd2_log_start_commit+0xba/0xcd
 [<c02b8075>] jbd2_log_start_commit+0x24/0x34
 [<c0279ed5>] ext4_evict_inode+0x71/0x2e3
 [<c021f0ec>] evict+0x94/0x135
 [<c021f9aa>] iput+0x10a/0x110
 [<c02b7836>] jbd2_journal_destroy+0x190/0x1ce
 [<c0175284>] ? bit_waitqueue+0x50/0x50
 [<c028d23f>] ext4_put_super+0x52/0x294
 [<c020efe3>] generic_shutdown_super+0x48/0xb4
 [<c020f071>] kill_block_super+0x22/0x60
 [<c020f3e0>] deactivate_locked_super+0x22/0x49
 [<c020f5d6>] deactivate_super+0x30/0x33
 [<c0222795>] mntput_no_expire+0x107/0x10c
 [<c02233a7>] sys_umount+0x2cf/0x2e0
 [<c02233ca>] sys_oldumount+0x12/0x14
 [<c08096b8>] syscall_call+0x7/0xb
---[ end trace 6a954cc790501c1f ]---
jbd2_log_wait_commit: error: j_commit_request=-1289337090, tid=0
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: stable@vger.kernel.org

Commit 84c17543 (ext4: move work from io_end to inode) triggered a
regression when running xfstest #270 when the file system is mounted
with dioread_nolock.

The problem is that after ext4_evict_inode() calls ext4_ioend_wait(),
this guarantees that last io_end structure has been freed, but it does
not guarantee that the workqueue structure, which was moved into the
inode by commit 84c17543, is actually finished.  Once
ext4_flush_completed_IO() calls ext4_free_io_end() on CPU #1, this
will allow ext4_ioend_wait() to return on CPU #2, at which point the
evict_inode() codepath can race against the workqueue code on CPU #1
accessing EXT4_I(inode)->i_unwritten_work to find the next item of
work to do.

Fix this by calling cancel_work_sync() in ext4_ioend_wait(), which
will be renamed ext4_ioend_shutdown(), since it is only used by
ext4_evict_inode().  Also, move the call to ext4_ioend_shutdown()
until after truncate_inode_pages() and filemap_write_and_wait() are
called, to make sure all dirty pages have been written back and
flushed from the page cache first.

BUG: unable to handle kernel NULL pointer dereference at   (null)
IP: [<c01dda6a>] cwq_activate_delayed_work+0x3b/0x7e
*pdpt = 0000000030bc3001 *pde = 0000000000000000 
Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
Modules linked in:
Pid: 6, comm: kworker/u:0 Not tainted 3.8.0-rc3-00013-g84c17543-dirty #91 Bochs Bochs
EIP: 0060:[<c01dda6a>] EFLAGS: 00010046 CPU: 0
EIP is at cwq_activate_delayed_work+0x3b/0x7e
EAX: 00000000 EBX: 00000000 ECX: f505fe54 EDX: 00000000
ESI: ed5b697c EDI: 00000006 EBP: f64b7e8c ESP: f64b7e84
 DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
CR0: 8005003b CR2: 00000000 CR3: 30bc2000 CR4: 000006f0
DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000
DR6: ffff0ff0 DR7: 00000400
Process kworker/u:0 (pid: 6, ti=f64b6000 task=f64b4160 task.ti=f64b6000)
Stack:
 f505fe00 00000006 f64b7e9c c01de3d7 f6435540 00000003 f64b7efc c01def1d
 f6435540 00000002 00000000 0000008a c16d0808 c040a10b c16d07d8 c16d08b0
 f505fe00 c16d0780 00000000 00000000 ee153df4 c1ce4a30 c17d0e30 00000000
Call Trace:
 [<c01de3d7>] cwq_dec_nr_in_flight+0x71/0xfb
 [<c01def1d>] process_one_work+0x5d8/0x637
 [<c040a10b>] ? ext4_end_bio+0x300/0x300
 [<c01e3105>] worker_thread+0x249/0x3ef
 [<c01ea317>] kthread+0xd8/0xeb
 [<c01e2ebc>] ? manage_workers+0x4bb/0x4bb
 [<c023a370>] ? trace_hardirqs_on+0x27/0x37
 [<c0f1b4b7>] ret_from_kernel_thread+0x1b/0x28
 [<c01ea23f>] ? __init_kthread_worker+0x71/0x71
Code: 01 83 15 ac ff 6c c1 00 31 db 89 c6 8b 00 a8 04 74 12 89 c3 30 db 83 05 b0 ff 6c c1 01 83 15 b4 ff 6c c1 00 89 f0 e8 42 ff ff ff <8b> 13 89 f0 83 05 b8 ff 6c c1
 6c c1 00 31 c9 83
EIP: [<c01dda6a>] cwq_activate_delayed_work+0x3b/0x7e SS:ESP 0068:f64b7e84
CR2: 0000000000000000
---[ end trace a1923229da53d8a4 ]---
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jan Kara <jack@suse.cz>

Alex writes:
"Mostly just small bug fixes.  Big change is new pci ids
for Richland APUs."

* 'drm-fixes-3.9' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon: add Richland pci ids
  drm/radeon: add support for Richland APUs
  drm/radeon/benchmark: allow same domains for dma copy
  drm/radeon/benchmark: make sure bo blit copy exists before using it
  drm/radeon: fix backend map setup on 1 RB trinity boards
  drm/radeon: fix S/R on VM systems (cayman/TN/SI)

Merge branch 'drm-nouveau-fixes-3.9' of git://anongit.freedesktop.org/git/nouveau/linux-2.6 into drm-next

Lots of thermal fixes and fix a lockdep warning we've been seeing.

* 'drm-nouveau-fixes-3.9' of git://anongit.freedesktop.org/git/nouveau/linux-2.6:
  drm/nv50/kms: prevent lockdep false-positive in page flipping path
  drm/nouveau/core: fix return value of nouveau_object_del()
  drm/nouveau/hwmon: do not expose a buggy temperature if it is unavailable
  drm/nouveau/therm: display the availability of the internal sensor
  drm/nouveau/therm: disable temperature management if the sensor isn't readable
  drm/nouveau/therm: disable auto fan management if temperature is not available
  drm/nv40/therm: reserve negative temperatures for errors
  drm/nv40/therm: disable temperature reading if the bios misses some parameters
  drm/nouveau/therm-ic: the temperature is off by sensor_constant, warn the user
  drm/nouveau/therm: remove some confusion introduced by therm_mode
  drm/nouveau/therm: do not make assumptions on temperature
  drm/nv40/therm: increase the sensor's settling delay to 20ms
  drm/nv40/therm: improve selection between the old and the new style

Pull vfio fix from Alex Williamson.

* tag 'vfio-v3.9-rc4' of git://github.com/awilliam/linux-vfio:
  vfio: include <linux/slab.h> for kmalloc

Pull kvm fixes from Marcelo Tosatti.

* git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: Fix bounds checking in ioapic indirect register reads (CVE-2013-1798)
  KVM: x86: Convert MSR_KVM_SYSTEM_TIME to use gfn_to_hva_cache functions (CVE-2013-1797)
  KVM: x86: fix for buffer overflow in handling of MSR_KVM_SYSTEM_TIME (CVE-2013-1796)
  KVM: x86: fix deadlock in clock-in-progress request handling
  KVM: allow host header to be included even for !CONFIG_KVM

Commit 28c70f16 ("drm/i915: use the gmbus irq for waits") switched to
using GMBUS irqs instead of GPIO bit-banging for chipset generations 4
and above.

It turns out though that on many systems this leads to spurious interrupts
being generated, long after the register write to disable the IRQs has been
issued.

Typically this results in the spurious interrupt source getting
disabled:

[    9.636345] irq 16: nobody cared (try booting with the "irqpoll" option)
[    9.637915] Pid: 4157, comm: ifup Tainted: GF            3.9.0-rc2-00341-g08637024 #422
[    9.639484] Call Trace:
[    9.640731]  <IRQ>  [<ffffffff8109b40d>] __report_bad_irq+0x1d/0xc7
[    9.640731]  [<ffffffff8109b7db>] note_interrupt+0x15b/0x1e8
[    9.640731]  [<ffffffff810999f7>] handle_irq_event_percpu+0x1bf/0x214
[    9.640731]  [<ffffffff81099a88>] handle_irq_event+0x3c/0x5c
[    9.640731]  [<ffffffff8109c139>] handle_fasteoi_irq+0x7a/0xb0
[    9.640731]  [<ffffffff8100400e>] handle_irq+0x1a/0x24
[    9.640731]  [<ffffffff81003d17>] do_IRQ+0x48/0xaf
[    9.640731]  [<ffffffff8142f1ea>] common_interrupt+0x6a/0x6a
[    9.640731]  <EOI>  [<ffffffff8142f952>] ? system_call_fastpath+0x16/0x1b
[    9.640731] handlers:
[    9.640731] [<ffffffffa000d771>] usb_hcd_irq [usbcore]
[    9.640731] [<ffffffffa0306189>] yenta_interrupt [yenta_socket]
[    9.640731] Disabling IRQ #16

The really curious thing is now that irq 16 is _not_ the interrupt for
the i915 driver when using MSI, but it _is_ the interrupt when not
using MSI. So by all indications it seems like gmbus is able to
generate a legacy (shared) interrupt in MSI mode on some
configurations. I've tried to reproduce this and the differentiating
thing seems to be that on unaffected systems no other device uses irq
16 (which seems to be the non-MSI intel gfx interrupt on all gm45).

I have no idea how that even can happen.

To avoid tempting this elephant into a rage, just disable gmbus
interrupt support on gen 4.

v2: Improve the commit message with exact details of what's going on.
Also add a comment in the code to warn against this particular
elephant in the room.

v3: Move the comment explaing how gen4 blows up next to the definition
of HAS_GMBUS_IRQ to keep the code-flow straight. Suggested by Chris
Wilson.

Signed-off-by: Jiri Kosina <jkosina@suse.cz> (v1)
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
References: https://lkml.org/lkml/2013/3/8/325Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

Pull XFS fixes from Ben Myers:

 - Fix for a potential infinite loop which was introduced in commit
   4d559a3b ("xfs: limit speculative prealloc near ENOSPC
   thresholds")

 - Fix for the return type of xfs_iomap_eof_prealloc_initial_size from
   commit a1e16c26 ("xfs: limit speculative prealloc size on sparse
   files")

 - Fix for a failed buffer readahead causing subsequent callers to fail
   incorrectly

* tag 'for-linus-v3.9-rc4' of git://oss.sgi.com/xfs/xfs:
  xfs: ensure we capture IO errors correctly
  xfs: fix xfs_iomap_eof_prealloc_initial_size type
  xfs: fix potential infinite loop in xfs_iomap_prealloc_size()

Mantas Mikulėnas reported that his graphics hardware failed to
initialise after commit f9a37be0 ("x86: Use PCI setup data").

The aim of this commit was to ensure that ROM images were available on
some Apple systems that don't expose the GPU ROM via any other source.
In this case, UEFI appears to have provided a broken ROM image that we
were using even though there was a perfectly valid ROM available via
other sources.  The simplest way to handle this seems to be to just
re-order pci_map_rom() and leave any firmare-supplied ROM to last.
Signed-off-by: Matthew Garrett <matthew.garrett@nebula.com>
Tested-by: Mantas Mikulėnas <grawity@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull sparc fixes from David Miller:
 "Just some minor fixups, a sunsu console setup panic cure, and
  recognition of a Fujitsu sun4v cpu."

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  sparc: remove unused "config BITS"
  sparc: delete "if !ULTRA_HAS_POPULATION_COUNT"
  sparc64: correctly recognize SPARC64-X chips
  sparc,leon: fix GRPCI2 device0 PCI config space access
  sunsu: Fix panic in case of nonexistent port at "console=ttySY" cmdline option

Pull arm64 fixes from Catalin Marinas:

 - Fix !SMP build error.

 - Fix padding computation in struct ucontext (no ABI change).

 - Minor clean-up after the signal patches (unused var).

 - Two old Kconfig options clean-up.

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/cmarinas/linux-aarch64:
  arm64: Kconfig.debug: Remove unused CONFIG_DEBUG_ERRORS
  arm64: Do not select GENERIC_HARDIRQS_NO_DEPRECATED
  arm64: fix padding computation in struct ucontext
  arm64: Fix build error with !SMP
  arm64: Removed unused variable in compat_setup_rt_frame()

sparc's asm/module.h got removed in commit
786d35d4 ("Make most arch asm/module.h
files use asm-generic/module.h"). That removed the only two uses of this
Kconfig symbol. So we can remove its entry too.

> >From arch/sparc/Makefile:
>     ifeq ($(CONFIG_SPARC32),y)
>     [...]
>
>     [...]
>     export BITS    := 32
>     [...]
>
>     else
>     [...]
>
>     [...]
>     export BITS   := 64
>     [...]
>
> So $(BITS) is set depending on whether CONFIG_SPARC32 is set or not.
> Using $(BITS) in sparc's Makefiles is not using CONFIG_BITS. That
> doesn't count as usage of "config BITS".
Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>