Some devices, like the I2C controller on SH7372, are not
necessary for providing power to their children or forwarding
wakeup signals (and generally interrupts) from them.  They are
only needed by their children when there's some data to transfer,
so they may be suspended for the majority of time and resumed
on demand, when the children have data to send or receive.  For this
purpose, however, their power.ignore_children flags have to be set,
or the PM core wouldn't allow them to be suspended while their
children were active.

Unfortunately, in some situations it may take too much time to
resume such devices so that they can assist their children in
transferring data.  For example, if such a device belongs to a PM
domain which goes to the "power off" state when that device is
suspended, it may take too much time to restore power to the
domain in response to the request from one of the device's
children.  In that case, if the parent's resume time is critical,
the domain should stay in the "power on" state, although it still may
be desirable to power manage the parent itself (e.g. by manipulating
its clock).

In general, device PM QoS may be used to address this problem.
Namely, if the device's children added PM QoS latency constraints
for it, they would be able to prevent it from being put into an
overly deep low-power state.  However, in some cases the devices
needing to be serviced are not the immediate children of a
"children-ignoring" device, but its grandchildren or even less
direct descendants.  In those cases, the entity wanting to add a
PM QoS request for a given device's ancestor that ignores its
children will have to find it in the first place, so introduce a new
helper function that may be used to achieve that.  This function,
dev_pm_qos_add_ancestor_request(), will search for the first
ancestor of the given device whose power.ignore_children flag is
set and will add a device PM QoS latency request for that ancestor
on behalf of the caller.  The request added this way may be removed
with the help of dev_pm_qos_remove_request() in the future, like
any other device PM QoS latency request.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

SH7372 uses two independent mechanisms for ensuring that power
domains will never be turned off: the stay_on flag and the "always
on" domain governor.  Moreover, the "always on" governor is only taken
into accout by runtime PM code paths, while the stay_on flag affects
all attempts to turn the given domain off.  Thus setting the stay_on
flag causes the "always on" governor to be unnecessary, which is
quite confusing.

However, the stay_on flag is currently only set for two domains: A3SP
and A4S.  Moreover, it only is set for the A3SP domain if
console_suspend_enabled is set, so stay_on won't be necessary for
that domain any more if console_suspend_enabled is checked directly
in its .suspend() routine.  [This requires domain .suspend() to
return a result, but that is a minor modification.]  Analogously,
stay_on won't be necessary for the A4S domain if it's .suspend()
routine always returns an error code.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Acked-by: NMagnus Damm <damm@opensource.se>

Since the SH7372's INTCS in included into syscore suspend/resume,
which causes the chip to be accessed when PM domains have been
turned off during system suspend, the A4R domain containing the
INTCS has to stay on during system sleep, which is suboptimal
from the power consumption point of view.

For this reason, add a new INTC flag, skip_syscore_suspend, to mark
the INTCS for intc_suspend() and intc_resume(), so that they don't
touch it.  This allows the A4R domain to be turned off during
system suspend and the INTCS state is resrored during system
resume by the A4R's "power on" code.
Suggested-by: NMagnus Damm <damm@opensource.se>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Acked-by: NMagnus Damm <damm@opensource.se>

The sh7372 contains a power domain named A4S which in turn
contains power domains for both I/O Devices and CPU cores.

At this point only System wide Suspend-to-RAM is supported,
but the the hardware can also support CPUIdle. With more
efforts in the future CPUIdle can work with bot A4S and A3SM.

Tested on the sh7372 Mackerel board.

[rjw: Rebased on top of the current linux-pm tree.]
Signed-off-by: NMagnus Damm <damm@opensource.se>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

The S3C64xx SoCs contain a set of gateable power domains which can be
enabled and disabled at runtime in order to save power.  Use the generic
power domain code to implement support for these in software, enabling
runtime control of most domains:

 - ETM (not supported in mainline).
 - Domain G: 3D acceleration (no mainline support).
 - Domain V: MFC (no mainline support).
 - Domain I: JPEG and camera interface (no mainline support).
 - Domain P: 2D acceleration, TV encoder and scaler (no mainline support)
 - Domain S: Security (no mainline support).
 - Domain F: LCD (driver already uses runtime PM), post processing and
   rotation (no mainline support).

The IROM domain is marked as always enabled as we should arrange for it
to be enabled when we suspend which will need a bit more work.

Due to all the conditional device registration that the platform does
wrap s3c_pm_init() with s3c64xx_pm_init() which actually puts the device
into the power domain after the machines have registered, looking for
platform data to tell if the device was registered. Since currently only
Cragganmore actually sets up PM that is the only machine updated.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NKukjin Kim <kgene.kim@samsung.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Saves a tiny amount of code.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: NMagnus Damm <damm@opensource.se>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Since systems are likely to have power domains that can't be turned off
for various reasons at least temporarily while implementing power domain
support provide a default governor which will always refuse to power off
the domain, saving platforms having to implement their own.

Since the code is so tiny don't bother with a Kconfig symbol for it.
Signed-off-by: NMark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Commit d23b9b00 (PM / Domains: Rework
system suspend callback routines (v2)) broke the system suspend and
resume handling by devices belonging to generic PM domains, because
it used freeze/thaw callbacks instead of suspend/resume ones and
didn't initialize device callbacks for system suspend/resume
properly at all.  Fix those problems.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Add a name member pointer to struct generic_pm_domain and use it in
diagnostic messages regarding the domain power-off and power-on
latencies.  Update the ARM shmobile SH7372 code to assign names to
the PM domains used by it.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Acked-by: NMagnus Damm <damm@opensource.se>

Fix the following compalitaion breakage:

In file included from linux/drivers/sh/pm_runtime.c:15:
linux/include/linux/pm_domain.h: In function 'dev_to_genpd':
linux/include/linux/pm_domain.h:142: error: implicit declaration of function 'ERR_PTR'
linux/include/linux/pm_domain.h:142: warning: return makes pointer from integer without a cast
In file included from linux/include/linux/sh_clk.h:10,
                 from linux/drivers/sh/pm_runtime.c:19:
linux/include/linux/err.h: At top level:
linux/include/linux/err.h:22: error: conflicting types for 'ERR_PTR'
linux/include/linux/pm_domain.h:142: note: previous implicit declaration of 'ERR_PTR' was here
make[3]: *** [drivers/sh/pm_runtime.o] Error 1
Reported-by: NNobuhiro Iwamatsu <nobuhiro.iwamatsu.yj@renesas.com>
Signed-off-by: NGuennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Measure the time of execution of the .stop(), .start(), .save_state()
and .restore_state() PM domain device callbacks and if the result
is greater than the corresponding latency value stored in the
device's struct generic_pm_domain_data object, replace the inaccurate
value with the measured time.

Do analogously for the PM domains' .power_off() and .power_off()
callbacks.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Add a function deciding whether or not a given PM domain should
be powered off on the basis of the PM QoS constraints of devices
belonging to it and their PM QoS timing data.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Add a function deciding whether or not devices should be stopped in
pm_genpd_runtime_suspend() depending on their PM QoS constraints
and stop/start timing values.  Make it possible to add information
used by this function to device objects.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Acked-by: NMagnus Damm <damm@opensource.se>

The current generic PM domains code attempts to use the generic
system suspend operations along with the domains' device stop/start
routines, which requires device drivers to assume that their
system suspend/resume (and hibernation/restore) callbacks will always
be used with generic PM domains.  However, in theory, the same
hardware may be used in devices that don't belong to any PM domain,
in which case it would be necessary to add "fake" PM domains to
satisfy the above assumption.  Also, the domain the hardware belongs
to may not be handled with the help of the generic code.

To allow device drivers that may be used along with the generic PM
domains code of more flexibility, add new device callbacks,
.suspend(), .suspend_late(), .resume_early(), .resume(), .freeze(),
.freeze_late(), .thaw_early(), and .thaw(), that can be supplied by
the drivers in addition to their "standard" system suspend and
hibernation callbacks.  These new callbacks, if defined, will be used
by the generic PM domains code for the handling of system suspend and
hibernation instead of the "standard" ones.  This will allow drivers
to be designed to work with generic PM domains as well as without
them.

For backwards compatibility, introduce default implementations of the
new callbacks for PM domains that will execute pm_generic_suspend(),
pm_generic_suspend_noirq(), pm_generic_resume_noirq(),
pm_generic_resume(), pm_generic_freeze(), pm_generic_freeze_noirq(),
pm_generic_thaw_noirq(), and pm_generic_thaw(), respectively, for the
given device if its driver doesn't define those callbacks.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

The current PM domains code uses device drivers' .runtime_suspend()
and .runtime_resume() callbacks as the "save device state" and
"restore device state" operations, which may not be appropriate in
general, because it forces drivers to assume that they always will
be used with generic PM domains.  However, in theory, the same
hardware may be used in devices that don't belong to any PM
domain, in which case it would be necessary to add "fake" PM
domains to satisfy the above assumption.  It also may be located in
a PM domain that's not handled with the help of the generic code.

To allow device drivers that may be used along with the generic PM
domains code of more flexibility, introduce new device callbacks,
.save_state() and .restore_state(), that can be supplied by the
drivers in addition to their "standard" runtime PM callbacks.  This
will allow the drivers to be designed to work with generic PM domains
as well as without them.

For backwards compatibility, introduce default .save_state() and
.restore_state() callback routines for PM domains that will execute
a device driver's .runtime_suspend() and .runtime_resume() callbacks,
respectively, for the given device if the driver doesn't provide its
own implementations of .save_state() and .restore_state().
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

The current generic PM domains code requires that the same .stop(),
.start() and .active_wakeup() device callback routines be used for
all devices in the given domain, which is inflexible and may not
cover some specific use cases.  For this reason, make it possible to
use device specific .start()/.stop() and .active_wakeup() callback
routines by adding corresponding callback pointers to struct
generic_pm_domain_data.  Add a new helper routine,
pm_genpd_register_callbacks(), that can be used to populate
the new per-device callback pointers.

Modify the shmobile's power domains code to allow drivers to add
their own code to be run during the device stop and start operations
with the help of the new callback pointers.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Acked-by: NMagnus Damm <damm@opensource.se>

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: fix meta data raid-repair merge problem
  Btrfs: skip allocation attempt from empty cluster
  Btrfs: skip block groups without enough space for a cluster
  Btrfs: start search for new cluster at the beginning
  Btrfs: reset cluster's max_size when creating bitmap
  Btrfs: initialize new bitmaps' list
  Btrfs: fix oops when calling statfs on readonly device
  Btrfs: Don't error on resizing FS to same size
  Btrfs: fix deadlock on metadata reservation when evicting a inode
  Fix URL of btrfs-progs git repository in docs
  btrfs scrub: handle -ENOMEM from init_ipath()

Commit 4a54c8c1 introduced raid-repair, killing the individual
readpage_io_failed_hook entries from inode.c and disk-io.c. Commit
4bb31e92 introduced new readahead code, adding a readpage_io_failed_hook to
disk-io.c.

The raid-repair commit had logic to disable raid-repair, if
readpage_io_failed_hook is set. Thus, the readahead commit effectively
disabled raid-repair for meta data.

This commit changes the logic to always attempt raid-repair when needed and
call the readpage_io_failed_hook in case raid-repair fails. This is much
more straight forward and should have been like that from the beginning.
Signed-off-by: NJan Schmidt <list.btrfs@jan-o-sch.net>
Reported-by: NStefan Behrens <sbehrens@giantdisaster.de>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  IB: Fix RCU lockdep splats
  IB/ipoib: Prevent hung task or softlockup processing multicast response
  IB/qib: Fix over-scheduling of QSFP work
  RDMA/cxgb4: Fix retry with MPAv1 logic for MPAv2
  RDMA/cxgb4: Fix iw_cxgb4 count_rcqes() logic
  IB/qib: Don't use schedule_work()

* 'dt-for-linus' of git://sources.calxeda.com/kernel/linux:
  of: Add Silicon Image vendor prefix
  of/irq: of_irq_init: add check for parent equal to child node

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: twl: fix twl4030 support for smps regulators
  regulator: fix use after free bug
  regulator: aat2870: Fix the logic of checking if no id is matched in aat2870_get_regulator

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (45 commits)
  ARM: ux500: update defconfig
  ARM: u300: update defconfig
  ARM: at91: enable additional boards in existing soc defconfig files
  ARM: at91: refresh soc defconfig files for 3.2
  ARM: at91: rename defconfig files appropriately
  ARM: OMAP2+: Fix Compilation error when omap_l3_noc built as module
  ARM: OMAP2+: Remove empty io.h
  ARM: OMAP2: select ARM_AMBA if OMAP3_EMU is defined
  ARM: OMAP: smartreflex: fix IRQ handling bug
  ARM: OMAP: PM: only register TWL with voltage layer when device is present
  ARM: OMAP: hwmod: Fix the addr space, irq, dma count APIs
  arm: mx28: fix bit operation in clock setting
  ARM: imx: export imx_ioremap
  ARM: imx/mm-imx3: conditionally compile i.MX31 and i.MX35 code
  ARM: mx5: Fix checkpatch warnings in cpu-imx5.c
  MAINTAINERS: Add missing directory
  ARM: imx: drop 'ARCH_MX31' and 'ARCH_MX35'
  ARM: imx6q: move clock register map to machine_desc.map_io
  ARM: pxa168/gplugd: add the correct SSP device
  ARM: Update mach-types to fix mxs build breakage
  ...

If we don't have a cluster, don't bother trying to allocate from it,
jumping right away to the attempt to allocate a new cluster.
Signed-off-by: NAlexandre Oliva <oliva@lsd.ic.unicamp.br>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

We test whether a block group has enough free space to hold the
requested block, but when we're doing clustered allocation, we can
save some cycles by testing whether it has enough room for the cluster
upfront, otherwise we end up attempting to set up a cluster and
failing.  Only in the NO_EMPTY_SIZE loop do we attempt an unclustered
allocation, and by then we'll have zeroed the cluster size, so this
patch won't stop us from using the block group as a last resort.
Signed-off-by: NAlexandre Oliva <oliva@lsd.ic.unicamp.br>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

Instead of starting at zero (offset is always zero), request a cluster
starting at search_start, that denotes the beginning of the current
block group.
Signed-off-by: NAlexandre Oliva <oliva@lsd.ic.unicamp.br>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

The field that indicates the size of the largest contiguous chunk of
free space in the cluster is not initialized when setting up bitmaps,
it's only increased when we find a larger contiguous chunk.  We end up
retaining a larger value than appropriate for highly-fragmented
clusters, which may cause pointless searches for large contiguous
groups, and even cause clusters that do not meet the density
requirements to be set up.
Signed-off-by: NAlexandre Oliva <oliva@lsd.ic.unicamp.br>
Signed-off-by: NChris Mason <chris.mason@oracle.com>

We're failing to create clusters with bitmaps because
setup_cluster_no_bitmap checks that the list is empty before inserting
the bitmap entry in the list for setup_cluster_bitmap, but the list
field is only initialized when it is restored from the on-disk free
space cache, or when it is written out to disk.

Besides a potential race condition due to the multiple use of the list
field, filesystem performance severely degrades over time: as we use
up all non-bitmap free extents, the try-to-set-up-cluster dance is
done at every metadata block allocation.  For every block group, we
fail to set up a cluster, and after failing on them all up to twice,
we fall back to the much slower unclustered allocation.

To make matters worse, before the unclustered allocation, we try to
create new block groups until we reach the 1% threshold, which
introduces additional bitmaps and thus block groups that we'll iterate
over at each metadata block request.

To reproduce this bug:

  # dd if=/dev/zero of=img bs=1M count=256
  # mkfs.btrfs img
  # losetup -r /dev/loop1 img
  # mount /dev/loop1 /mnt
  OOPS!!

It triggered BUG_ON(!nr_devices) in btrfs_calc_avail_data_space().

To fix this, instead of checking write-only devices, we check all open
deivces:

  # df -h /dev/loop1
  Filesystem            Size  Used Avail Use% Mounted on
  /dev/loop1            250M   28K  238M   1% /mnt
Signed-off-by: NLi Zefan <lizf@cn.fujitsu.com>

It seems overly harsh to fail a resize of a btrfs file system to the
same size when a shrink or grow would succeed.  User app GParted trips
over this error.  Allow it by bypassing the shrink or grow operation.
Signed-off-by: NMike Fleetwood <mike.fleetwood@googlemail.com>

When I ran the xfstests, I found the test tasks was blocked on meta-data
reservation.

By debugging, I found the reason of this bug:
   start transaction
        |
	v
   reserve meta-data space
	|
	v
   flush delay allocation -> iput inode -> evict inode
	^					|
	|					v
   wait for delay allocation flush <- reserve meta-data space

And besides that, the flush on evicting inode will block the thread, which
is reclaiming the memory, and make oom happen easily.

Fix this bug by skipping the flush step when evicting inode.
Signed-off-by: NMiao Xie <miaox@cn.fujitsu.com>

	The location of the btrfs-progs repository has been changed.
	This patch updates the documentation accordingly.
Signed-off-by: NArnd Hannemann <arnd@arndnet.de>

init_ipath() can return an ERR_PTR(-ENOMEM).
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>

* 'pm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  PM: Update comments describing device power management callbacks
  PM / Sleep: Update documentation related to system wakeup
  PM / Runtime: Make documentation follow the new behavior of irq_safe
  PM / Sleep: Correct inaccurate information in devices.txt
  PM / Domains: Document how PM domains are used by the PM core
  PM / Hibernate: Do not leak memory in error/test code paths

Commit f2c31e32 ("net: fix NULL dereferences in check_peer_redir()")
forgot to take care of infiniband uses of dst neighbours.

Many thanks to Marc Aurele who provided a nice bug report and feedback.
Reported-by: NMarc Aurele La France <tsi@ualberta.ca>
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Cc: David Miller <davem@davemloft.net>
Cc: <stable@kernel.org>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

This following can occur with ipoib when processing a multicast reponse:

    BUG: soft lockup - CPU#0 stuck for 67s! [ib_mad1:982]
    Modules linked in: ...
    CPU 0:
    Modules linked in: ...
    Pid: 982, comm: ib_mad1 Not tainted 2.6.32-131.0.15.el6.x86_64 #1 ProLiant DL160 G5
    RIP: 0010:[<ffffffff814ddb27>]  [<ffffffff814ddb27>] _spin_unlock_irqrestore+0x17/0x20
    RSP: 0018:ffff8802119ed860  EFLAGS: 00000246
    0000000000000004 RBX: ffff8802119ed860 RCX: 000000000000a299
    RDX: ffff88021086c700 RSI: 0000000000000246 RDI: 0000000000000246
    RBP: ffffffff8100bc8e R08: ffff880210ac229c R09: 0000000000000000
    R10: ffff88021278aab8 R11: 0000000000000000 R12: ffff8802119ed860
    R13: ffffffff8100be6e R14: 0000000000000001 R15: 0000000000000003
    FS:  0000000000000000(0000) GS:ffff880028200000(0000) knlGS:0000000000000000
    CS:  0010 DS: 0018 ES: 0018 CR0: 000000008005003b
    CR2: 00000000006d4840 CR3: 0000000209aa5000 CR4: 00000000000406f0
    DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
    DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
    Call Trace:
    [<ffffffffa032c247>] ? ipoib_mcast_send+0x157/0x480 [ib_ipoib]
    [<ffffffff8100bc8e>] ? apic_timer_interrupt+0xe/0x20
    [<ffffffff8100bc8e>] ? apic_timer_interrupt+0xe/0x20
    [<ffffffffa03283d4>] ? ipoib_path_lookup+0x124/0x2d0 [ib_ipoib]
    [<ffffffffa03286fc>] ? ipoib_start_xmit+0x17c/0x430 [ib_ipoib]
    [<ffffffff8141e758>] ? dev_hard_start_xmit+0x2c8/0x3f0
    [<ffffffff81439d0a>] ? sch_direct_xmit+0x15a/0x1c0
    [<ffffffff81423098>] ? dev_queue_xmit+0x388/0x4d0
    [<ffffffffa032d6b7>] ? ipoib_mcast_join_finish+0x2c7/0x510 [ib_ipoib]
    [<ffffffffa032dab8>] ? ipoib_mcast_sendonly_join_complete+0x1b8/0x1f0 [ib_ipoib]
    [<ffffffffa02a0946>] ? mcast_work_handler+0x1a6/0x710 [ib_sa]
    [<ffffffffa015f01e>] ? ib_send_mad+0xfe/0x3c0 [ib_mad]
    [<ffffffffa00f6c93>] ? ib_get_cached_lmc+0xa3/0xb0 [ib_core]
    [<ffffffffa02a0f9b>] ? join_handler+0xeb/0x200 [ib_sa]
    [<ffffffffa029e4fc>] ? ib_sa_mcmember_rec_callback+0x5c/0xa0 [ib_sa]
    [<ffffffffa029e79c>] ? recv_handler+0x3c/0x70 [ib_sa]
    [<ffffffffa01603a4>] ? ib_mad_completion_handler+0x844/0x9d0 [ib_mad]
    [<ffffffffa015fb60>] ? ib_mad_completion_handler+0x0/0x9d0 [ib_mad]
    [<ffffffff81088830>] ? worker_thread+0x170/0x2a0
    [<ffffffff8108e160>] ? autoremove_wake_function+0x0/0x40
    [<ffffffff810886c0>] ? worker_thread+0x0/0x2a0
    [<ffffffff8108ddf6>] ? kthread+0x96/0xa0
    [<ffffffff8100c1ca>] ? child_rip+0xa/0x20

Coinciding with stack trace is the following message:

    ib0: ib_address_create failed

The code below in ipoib_mcast_join_finish() will note the above
failure in the address handle but otherwise continue:

                ah = ipoib_create_ah(dev, priv->pd, &av);
                if (!ah) {
                        ipoib_warn(priv, "ib_address_create failed\n");
                } else {

The while loop at the bottom of ipoib_mcast_join_finish() will attempt
to send queued multicast packets in mcast->pkt_queue and eventually
end up in ipoib_mcast_send():

        if (!mcast->ah) {
                if (skb_queue_len(&mcast->pkt_queue) < IPOIB_MAX_MCAST_QUEUE)
                        skb_queue_tail(&mcast->pkt_queue, skb);
                else {
                        ++dev->stats.tx_dropped;
                        dev_kfree_skb_any(skb);
                }

My read is that the code will requeue the packet and return to the
ipoib_mcast_join_finish() while loop and the stage is set for the
"hung" task diagnostic as the while loop never sees a non-NULL ah, and
will do nothing to resolve.

There are GFP_ATOMIC allocates in the provider routines, so this is
possible and should be dealt with.

The test that induced the failure is associated with a host SM on the
same server during a shutdown.

This patch causes ipoib_mcast_join_finish() to exit with an error
which will flush the queued mcast packets.  Nothing is done to unwind
the QP attached state so that subsequent sends from above will retry
the join.
Reviewed-by: NRam Vepa <ram.vepa@qlogic.com>
Reviewed-by: NGary Leshner <gary.leshner@qlogic.com>
Signed-off-by: NMike Marciniszyn <mike.marciniszyn@qlogic.com>
Signed-off-by: NRoland Dreier <roland@purestorage.com>

* 'slab/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/linux:
  slub: avoid potential NULL dereference or corruption
  slub: use irqsafe_cpu_cmpxchg for put_cpu_partial
  slub: move discard_slab out of node lock
  slub: use correct parameter to add a page to partial list tail

* 'dev' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix racy use-after-free in ext4_end_io_dio()

"sil" is the most commonly used abbreviation for Silicon
Image products.
Signed-off-by: NPawel Moll <pawel.moll@arm.com>