Pull arm64 fixes from Will Deacon:
 "Apologies for this being so late, but we've uncovered a few nasty
  issues on arm64 which didn't settle down until yesterday and the fixes
  all look suitable for 4.3.  Of the four patches, three of them are
  Cc'd to stable, with the remaining patch fixing an issue that only
  took effect during the merge window.

  Summary:

   - Fix corruption in SWP emulation when STXR fails due to contention
   - Fix MMU re-initialisation when resuming from a low-power state
   - Fix stack unwinding code to match what ftrace expects
   - Fix relocation code in the EFI stub when DRAM base is not 2MB aligned"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64/efi: do not assume DRAM base is aligned to 2 MB
  Revert "ARM64: unwind: Fix PC calculation"
  arm64: kernel: fix tcr_el1.t0sz restore on systems with extended idmap
  arm64: compat: fix stxr failure case in SWP emulation

Pull ia64 kcmp syscall from Tony Luck:
 "Missed adding the kcmp() syscall a long time ago.  Now it seems that
  it is essential to build systemd"

* tag 'please-pull-syscalls' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux:
  [IA64] Wire up kcmp syscall

The current arm64 Image relocation code in the UEFI stub assumes that
the dram_base argument it receives is always a multiple of 2 MB. In
reality, it is simply the lowest start address of all RAM entries in
the UEFI memory map, which means it could be any multiple of 4 KB.

Since the arm64 kernel Image needs to reside TEXT_OFFSET bytes beyond
a 2 MB aligned base, or it will fail to boot, make sure we round dram_base
to 2 MB before using it to calculate the relocation address.

Fixes: e38457c3 ("arm64: efi: prefer AllocatePages() over efi_low_alloc() for vmlinux")
Reported-by: Timur Tabi <timur@codeaurora.org>
Tested-by: Timur Tabi <timur@codeaurora.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>

systemd > 218 fails to compile on ia64 with:

     error: ‘__NR_kcmp’ undeclared [1].

I've been told that this is because the kcmp syscall hasn't been wired up
for the ia64 arch [2].

The proposed patch thus wire up the kcmp syscall for the ia64 arch.

[1] https://bugs.gentoo.org/show_bug.cgi?id=560492
[2] https://bugs.gentoo.org/show_bug.cgi?id=560492#c17Signed-off-by: Émeric MASCHINO <emeric.maschino@gmail.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>

This reverts commit e306dfd0.

With this patch applied, we were the only architecture making this sort
of adjustment to the PC calculation in the unwinder. This causes
problems for ftrace, where the PC values are matched against the
contents of the stack frames in the callchain and fail to match any
records after the address adjustment.

Whilst there has been some effort to change ftrace to workaround this,
those patches are not yet ready for mainline and, since we're the odd
architecture in this regard, let's just step in line with other
architectures (like arch/arm/) for now.

Cc: <stable@vger.kernel.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>

Commit dd006da2 ("arm64: mm: increase VA range of identity map")
introduced a mechanism to extend the virtual memory map range
to support arm64 systems with system RAM located at very high offset,
where the identity mapping used to enable/disable the MMU requires
additional translation levels to map the physical memory at an equal
virtual offset.

The kernel detects at boot time the tcr_el1.t0sz value required by the
identity mapping and sets-up the tcr_el1.t0sz register field accordingly,
any time the identity map is required in the kernel (ie when enabling the
MMU).

After enabling the MMU, in the cold boot path the kernel resets the
tcr_el1.t0sz to its default value (ie the actual configuration value for
the system virtual address space) so that after enabling the MMU the
memory space translated by ttbr0_el1 is restored as expected.

Commit dd006da2 ("arm64: mm: increase VA range of identity map")
also added code to set-up the tcr_el1.t0sz value when the kernel resumes
from low-power states with the MMU off through cpu_resume() in order to
effectively use the identity mapping to enable the MMU but failed to add
the code required to restore the tcr_el1.t0sz to its default value, when
the core returns to the kernel with the MMU enabled, so that the kernel
might end up running with tcr_el1.t0sz value set-up for the identity
mapping which can be lower than the value required by the actual virtual
address space, resulting in an erroneous set-up.

This patchs adds code in the resume path that restores the tcr_el1.t0sz
default value upon core resume, mirroring this way the cold boot path
behaviour therefore fixing the issue.

Cc: <stable@vger.kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Fixes: dd006da2 ("arm64: mm: increase VA range of identity map")
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>

If the STXR instruction fails in the SWP emulation code, we leave *data
overwritten with the loaded value, therefore corrupting the data written
by a subsequent, successful attempt.

This patch re-jigs the code so that we only write back to *data once we
know that the update has happened.

Cc: <stable@vger.kernel.org>
Fixes: bd35a4ad ("arm64: Port SWP/SWPB emulation support from arm")
Reported-by: Shengjiu Wang <shengjiu.wang@freescale.com>
Reported-by: Vladimir Murzin <vladimir.murzin@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>

Pull powerpc fix from Michael Ellerman:
 - powerpc/dma: dma_set_coherent_mask() should not be GPL only from Ben

* tag 'powerpc-4.3-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/dma: dma_set_coherent_mask() should not be GPL only

When turning this from inline to an exported function I was a bit
over-eager and made it GPL only. This prevents the use of pretty much
all non-GPL PCI driver which is a bit over the top. Let's bring it
back in line with other architecture.

Fixes: 817820b0 ("powerpc/iommu: Support "hybrid" iommu/direct DMA ops for coherent_mask < dma_mask")
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

In commit b49a0871("block: remove split code in
blkdev_issue_{discard,write_same}"), discard_granularity and alignment
checks were removed. Ideally, with bio late splitting, the upper layers
shouldn't need to depend on device's limits.

Christoph reported a discard regression on the HGST Ultrastar SN100 NVMe
device when mkfs.xfs. We have not found the root cause yet.

This patch re-adds discard_granularity and alignment checks by reverting
the related changes in commit b49a0871. The good thing is now we can
remove the 2G discard size cap and just use UINT_MAX to avoid bi_size
overflow.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Tested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ming Lin <ming.l@ssi.samsung.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

Pull ARM fixes from Russell King:
 "Two fixes for ARM and one for clkdev:

   - Fix another build issue with vdsomunge on non-glibc systems
   - Fix a randconfig build error caused by an invalid configuration
   - Fix a clkdev problem causing the Nokia n700 to no longer boot"

* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
  clkdev: fix clk_add_alias() with a NULL alias device name
  ARM: 8445/1: fix vdsomunge not to depend on glibc specific byteswap.h
  ARM: make RiscPC depend on MMU

Pull blkcg fix from Jens Axboe:
 "One final fix that should go into 4.3.  It's a simple 2x1 liner,
  fixing a blkcg accounting issue.  It was using the wrong bio member to
  look at the sync and write bits..."

* 'for-linus' of git://git.kernel.dk/linux-block:
  blkcg: fix incorrect read/write sync/async stat accounting

Pull crypto fix from Herbert Xu:
 "This fixes a problem in the Crypto API that may cause spurious errors
  when signals are received by the process that made the orignal system
  call into the kernel"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: api - Only abort operations on fatal signal

Pull module preemption fix from Rusty Russell:
 "Turns out we should have always been disabling preemption here;
  someone finally caught it thanks to Peter Z's additional checks"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux:
  module: Fix locking in symbol_put_addr()

While unifying how blkcg stats are collected, 77ea7338 ("blkcg:
move io_service_bytes and io_serviced stats into blkcg_gq")
incorrectly used bio->flags instead of bio->rw to tell the IO type.
This made IOs to be accounted as the wrong type.  Fix it.
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: 77ea7338 ("blkcg: move io_service_bytes and io_serviced stats into blkcg_gq")
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

Pull iommu fixes from Joerg Roedel:
 "Two late fixes for the AMD IOMMU driver:

   - add an additional check to the io page-fault handler to avoid a
     BUG_ON being hit in handle_mm_fault()

   - fix a problem with devices writing to the system management area
     and were blocked by the IOMMU because the driver wrongly cleared
     out the DTE flags allowing that access"

* tag 'iommu-fixes-v4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/amd: Don't clear DTE flags when modifying it
  iommu/amd: Fix BUG when faulting a PROT_NONE VMA

Pull md fixes from Neil Brown:
 "Some raid1/raid10 fixes.

  I meant to get this to you before -rc7, but what with all the travel
  plans..

  Two fixes for bugs that are in both raid1 and raid10.  Both related to
  bad-block-lists and at least one needs to be back ported to 3.1.

  Also a revision for the "new" layout in raid10.  This "new" code
  (which aims to improve robustness) actually reduces robustness in some
  cases.  It probably isn't in use at all as not public user-space code
  makes use of these new layouts.  However just in case someone has
  their own code, it would be good to get the WARNing out for them
  sooner"

* tag 'md/4.3-rc6-fixes' of git://neil.brown.name/md:
  md/raid10: fix the 'new' raid10 layout to work correctly.
  md/raid10: don't clear bitmap bit when bad-block-list write fails.
  md/raid1: don't clear bitmap bit when bad-block-list write fails.
  md/raid10: submit_bio_wait() returns 0 on success
  md/raid1: submit_bio_wait() returns 0 on success

Pull drm fixes from Dave Airlie:
 "Last fixes from me: one amdgpu/radeon suspend resume and one leak fix,
  along with one vmware fix for some issues when command submission
  fails"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/amdgpu: don't try to recreate sysfs entries on resume
  drm/radeon: don't try to recreate sysfs entries on resume
  drm/amdgpu: stop leaking page flip fence
  drm/vmwgfx: Stabilize the command buffer submission code

Pull i2c fixes from Wolfram Sang:
 "Two driver bugfixes for the I2C subsystem"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: pnx: fix runtime warnings caused by enabling unprepared clock
  i2c: mv64xxx: really allow I2C offloading

I'm not sure whether this patch comes in too late, but it would be good to
have it in. It stabilizes command submission in case of command buffer errors.

* 'vmwgfx-fixes-4.3' of git://people.freedesktop.org/~thomash/linux:
  drm/vmwgfx: Stabilize the command buffer submission code

Two regression fixes and a memory leak fix for amdgpu and radeon.

* 'drm-fixes-4.3' of git://people.freedesktop.org/~agd5f/linux:
  drm/amdgpu: don't try to recreate sysfs entries on resume
  drm/radeon: don't try to recreate sysfs entries on resume
  drm/amdgpu: stop leaking page flip fence

In Linux 3.9 we introduce a new 'far' layout for RAID10 which was
supposed to rotate the replicas differently and so provide better
resilience.  In particular it could survive more combinations of 2
drive failures.

Unfortunately. due to a coding error, this some did what was wanted,
sometimes improved less than we hoped, and sometimes - in very
unlikely circumstances - put multiple replicas on the same device so
the redundancy was harmed.

No public user-space tool has created arrays using this layout so it
is very unlikely that zero-redundancy arrays actually exist.  Probably
no arrays using any form of the new layout exist.  But we cannot be
certain.

So use another bit in the 'layout' number and introduce a bug-fixed
version of the layout.
Also when assembling an array, if it has a zero-redundancy layout,
give a warning.
Reported-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>

When a write fails and a bad-block-list is present, we can
update the bad-block-list instead of writing the data.  If
this succeeds then it is OK clear the relevant bitmap-bit as
no further 'sync' of the block is needed.

However if writing the bad-block-list fails then we need to
treat the write as failed and particularly must not clear
the bitmap bit.  Otherwise the device can be re-added (after
any hardware connection issues are resolved) and because the
relevant bit in the bitmap is clear, that block will not be
resynced.  This leads to data corruption.

We already delay the final bio_endio() on the write until
the bad-block-list is written so that when the write
returns: either that data is safe, the bad-block record is
safe, or the fact that the device is faulty is safe.
However we *don't* delay the clearing of the bitmap, so the
bitmap bit can be recorded as cleared before we know if the
bad-block-list was written safely.

So: delay that until the write really is safe.
i.e. move the call to close_write() until just before
calling bio_endio(), and recheck the 'is array degraded'
status before making that call.

This bug goes back to v3.1 when bad-block-lists were
introduced, though it only affects arrays created with
mdadm-3.3 or later as only those have bad-block lists.

Backports will require at least
Commit: 95af587e ("md/raid10: ensure device failure recorded before write request returns.")
as well.  I'll send that to 'stable' separately.

Note that of the two tests of R10BIO_WriteError that this
patch adds, the first is certain to fail and the second is
certain to succeed.  However doing it this way makes the
patch more obviously correct.  I will tidy the code up in a
future merge window.
Reported-by: Nate Dailey <nate.dailey@stratus.com>
Fixes: bd870a16 ("md/raid10:  Handle write errors by updating badblock log.")
Signed-off-by: NeilBrown <neilb@suse.com>

When a write fails and a bad-block-list is present, we can
update the bad-block-list instead of writing the data.  If
this succeeds then it is OK clear the relevant bitmap-bit as
no further 'sync' of the block is needed.

However if writing the bad-block-list fails then we need to
treat the write as failed and particularly must not clear
the bitmap bit.  Otherwise the device can be re-added (after
any hardware connection issues are resolved) and because the
relevant bit in the bitmap is clear, that block will not be
resynced.  This leads to data corruption.

We already delay the final bio_endio() on the write until
the bad-block-list is written so that when the write
returns: either that data is safe, the bad-block record is
safe, or the fact that the device is faulty is safe.
However we *don't* delay the clearing of the bitmap, so the
bitmap bit can be recorded as cleared before we know if the
bad-block-list was written safely.

So: delay that until the write really is safe.
i.e. move the call to close_write() until just before
calling bio_endio(), and recheck the 'is array degraded'
status before making that call.

This bug goes back to v3.1 when bad-block-lists were
introduced, though it only affects arrays created with
mdadm-3.3 or later as only those have bad-block lists.

Backports will require at least
Commit: 55ce74d4 ("md/raid1: ensure device failure recorded before write request returns.")
as well.  I'll send that to 'stable' separately.

Note that of the two tests of R1BIO_WriteError that this
patch adds, the first is certain to fail and the second is
certain to succeed.  However doing it this way makes the
patch more obviously correct.  I will tidy the code up in a
future merge window.
Reported-and-tested-by: Nate Dailey <nate.dailey@stratus.com>
Cc: Jes Sorensen <Jes.Sorensen@redhat.com>
Fixes: cd5ff9a1 ("md/raid1:  Handle write errors by updating badblock log.")
Signed-off-by: NeilBrown <neilb@suse.com>

Pull USB fixes from Greg KH:
 "Here are three xhci driver fixes for reported issues for 4.3-rc7

  All have been in linux-next for a while with no problems"

* tag 'usb-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  xhci: Add spurious wakeup quirk for LynxPoint-LP controllers
  xhci: handle no ping response error properly
  xhci: don't finish a TD if we get a short transfer event mid TD

Pull tty/serial fixes from Greg KH:
 "Here are two fixes that resolve reported issues, one with the 8250
  driver, and the other with the generic fbcon driver.

  Both have been in linux-next for a while"

* tag 'tty-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  fbcon: initialize blink interval before calling fb_set_par
  Revert "serial: 8250_dma: don't bother DMA with small transfers"

Pull staging driver fixes from Greg KH:
 "Here are four iio driver fixes for 4.3-rc7, fixing some reported
  issues.  All of these have been in linux-next for a while"

* tag 'staging-4.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  iio: mxs-lradc: Fix temperature offset
  iio: accel: sca3000: memory corruption in sca3000_read_first_n_hw_rb()
  iio: st_accel: fix interrupt handling on LIS3LV02
  iio: adc: twl4030: Fix ADC[3:6] readings

Pull infiniband fixes from Doug Ledford:
 "It's late in the game, I know, but these fixes seemed important enough
  to warrant a late pull request.  They all involve oopses or use after
  frees or corruptions.

  Six serious fixes:

   - Hold the mutex around the find and corresponding update of our gid

   - The ifa list is rcu protected, copy its contents under rcu to avoid
     using a freed structure

   - On error, netdev might be null, so check it before trying to
     release it

   - On init, if workqueue alloc fails, fail init

   - The new demux patches exposed a bug in mlx5 and ipath drivers, we
     need to use the payload P_Key to determine the P_Key the packet
     arrived on because the hardware doesn't tell us the truth

   - Due to a couple convoluted error flows, it is possible for the CM
     to trigger a use_after_free and a double_free of rb nodes.  Add two
     checks to prevent that.  This code has worked for 10+ years.  It is
     likely that some of the recent changes have caused this issue to
     surface.  The current patch will protect us from nasty events for
     now while we track down why this is just now showing up"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma:
  IB/cm: Fix rb-tree duplicate free and use-after-free
  IB/cma: Use inner P_Key to determine netdev
  IB/ucma: check workqueue allocation before usage
  IB/cma: Potential NULL dereference in cma_id_from_event
  IB/core: Fix use after free of ifa
  IB/core: Fix memory corruption in ib_cache_gid_set_default_gid

Pull device mapper fixes from Mike Snitzer:
 "Three stable fixes (two in btree code used by DM thinp and one to
  properly store flags in DM cache metadata's superblock)"

* tag 'dm-4.3-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm cache: the CLEAN_SHUTDOWN flag was not being set
  dm btree: fix leak of bufio-backed block in btree_split_beneath error path
  dm btree remove: fix a bug when rebalancing nodes after removal

Pull block layer fixes from Jens Axboe:
 "A final set of fixes for 4.3.

  It is (again) bigger than I would have liked, but it's all been
  through the testing mill and has been carefully reviewed by multiple
  parties.  Each fix is either a regression fix for this cycle, or is
  marked stable.  You can scold me at KS.  The pull request contains:

   - Three simple fixes for NVMe, fixing regressions since 4.3.  From
     Arnd, Christoph, and Keith.

   - A single xen-blkfront fix from Cathy, fixing a NULL dereference if
     an error is returned through the staste change callback.

   - Fixup for some bad/sloppy code in nbd that got introduced earlier
     in this cycle.  From Markus Pargmann.

   - A blk-mq tagset use-after-free fix from Junichi.

   - A backing device lifetime fix from Tejun, fixing a crash.

   - And finally, a set of regression/stable fixes for cgroup writeback
     from Tejun"

* 'for-linus' of git://git.kernel.dk/linux-block:
  writeback: remove broken rbtree_postorder_for_each_entry_safe() usage in cgwb_bdi_destroy()
  NVMe: Fix memory leak on retried commands
  block: don't release bdi while request_queue has live references
  nvme: use an integer value to Linux errno values
  blk-mq: fix use-after-free in blk_mq_free_tag_set()
  nvme: fix 32-bit build warning
  writeback: fix incorrect calculation of available memory for memcg domains
  writeback: memcg dirty_throttle_control should be initialized with wb->memcg_completions
  writeback: bdi_writeback iteration must not skip dying ones
  writeback: fix bdi_writeback iteration in wakeup_dirtytime_writeback()
  writeback: laptop_mode_timer_fn() needs rcu_read_lock() around bdi_writeback iteration
  nbd: Add locking for tasks
  xen-blkfront: check for null drvdata in blkback_changed (XenbusStateClosing)

Pull Ceph fixes from Sage Weil:
 "Two fixes.

  One is a stopgap to prevent a stack blowout when users have a deep
  chain of image clones.  (We'll rewrite this code to be non-recursive
  for the next window, but in the meantime this is a simple fix that
  avoids a crash.)

  The second fixes a refcount underflow"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: prevent kernel stack blow up on rbd map
  rbd: don't leak parent_spec in rbd_dev_probe_parent()

Pull btrfs fixes from Chris Mason:
 "I have two more small fixes this week:

  Qu's fix avoids unneeded COW during fallocate, and Christian found a
  memory leak in the error handling of an earlier fix"

* 'for-linus-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  btrfs: fix possible leak in btrfs_ioctl_balance()
  btrfs: Avoid truncate tailing page if fallocate range doesn't exceed inode size

The driver can not be used on a platform with common clock framework
until clk_prepare/clk_unprepare calls are added, otherwise clk_enable
calls will fail and a WARN is generated.
Signed-off-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>

If the CLEAN_SHUTDOWN flag is not set when a cache is loaded then all cache
blocks are marked as dirty and a full writeback occurs.

__commit_transaction() is responsible for setting/clearing
CLEAN_SHUTDOWN (based the flags_mutator that is passed in).

Fix this issue, of the cache's on-disk flags being wrong, by making sure
__commit_transaction() does not reset the flags after the mutator has
altered the flags in preparation for them being serialized to disk.

before:

sb_flags = mutator(le32_to_cpu(disk_super->flags));
disk_super->flags = cpu_to_le32(sb_flags);
disk_super->flags = cpu_to_le32(cmd->flags);

after:

disk_super->flags = cpu_to_le32(cmd->flags);
sb_flags = mutator(le32_to_cpu(disk_super->flags));
disk_super->flags = cpu_to_le32(sb_flags);
Reported-by: Bogdan Vasiliev <bogdan.vasiliev@gmail.com>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org

btree_split_beneath()'s error path had an outstanding FIXME that speaks
directly to the potential for _not_ cleaning up a previously allocated
bufio-backed block.

Fix this by releasing the previously allocated bufio block using
unlock_block().
Reported-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <thornber@redhat.com>
Cc: stable@vger.kernel.org

Commit 4c7e3093 ("dm btree remove: fix bug in redistribute3") wasn't
a complete fix for redistribute3().

The redistribute3 function takes 3 btree nodes and shares out the entries
evenly between them.  If the three nodes in total contained
(MAX_ENTRIES * 3) - 1 entries between them then this was erroneously getting
rebalanced as (MAX_ENTRIES - 1) on the left and right, and (MAX_ENTRIES + 1) in
the center.

Fix this issue by being more careful about calculating the target number
of entries for the left and right nodes.

Unit tested in userspace using this program:
https://github.com/jthornber/redistribute3-test/blob/master/redistribute3_t.cSigned-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org

Mapping an image with a long parent chain (e.g. image foo, whose parent
is bar, whose parent is baz, etc) currently leads to a kernel stack
overflow, due to the following recursion in the reply path:

  rbd_osd_req_callback()
    rbd_obj_request_complete()
      rbd_img_obj_callback()
        rbd_img_parent_read_callback()
          rbd_obj_request_complete()
            ...

Limit the parent chain to 16 images, which is ~5K worth of stack.  When
the above recursion is eliminated, this limit can be lifted.

Fixes: http://tracker.ceph.com/issues/12538

Cc: stable@vger.kernel.org # 3.10+, needs backporting for < 4.2
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>

Currently we leak parent_spec and trigger a "parent reference
underflow" warning if rbd_dev_create() in rbd_dev_probe_parent() fails.
The problem is we take the !parent out_err branch and that only drops
refcounts; parent_spec that would've been freed had we called
rbd_dev_unparent() remains and triggers rbd_warn() in
rbd_dev_parent_put() - at that point we have parent_spec != NULL and
parent_ref == 0, so counter ends up being -1 after the decrement.

Redo rbd_dev_probe_parent() to fix this.

Cc: stable@vger.kernel.org # 3.10+, needs backporting for < 4.2
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>

Fixes an error on resume caused by:
fa022a9bSigned-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org