Pull Ceph fix from Sage Weil:
 "It's a simple fix for a hard to hit race, but low-risk and clearly
  correct"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: do a safe list traversal in rbd_img_request_submit()

In the very unlikely event where a guest would be foolish enough to
*read* from a write-only cache maintainance register, we end up
with preemption disabled, due to a misplaced get_cpu().

Just move the "is_write" test outside of the critical section.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

It's possible that the reference to the object request dropped
inside the loop in rbd_img_request_submit() will be the last
one, in which case the content of the object pointer can't be
trusted.

Use a safe form of the object request list traversal to avoid
problems.

This resolves:
    http://tracker.ceph.com/issues/4705Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>

Pull networking fixes from David Miller:

 1) Fix erroneous netfilter drop of SIP packets generated by some Cisco
    phones, from Patrick McHardy.

 2) Fix netfilter IPSET refcounting in list_set_add(), from Jozsef
    Kadlecsik.

 3) Fix TCP syncookies route lookup key, we don't use the same values we
    would use for the usual SYN receive processing, from Dmitry Popov.

 4) Fix NULL deref in bond_slave_netdev_event(), from Nikolay
    Aleksandrov.

 5) When bonding enslave fails, we can forget to clear the IFF_BONDING
    bit, fix also from Nikolay Aleksandrov.

 6) skb->csum_start is 16-bits, which is almost always just fine.  But
    if we reallocate the headroom of an SKB this can push the
    skb->csum_start value outside of it's valid range.  This can easily
    happen when collapsing multiple SKBs from the retransmit queue
    together.

    Fix from Thomas Graf.

 7) Fix NULL deref in be2net driver due to missing check of
    __vlan_put_tag() return value, from Ivan Vecera.

 8) tun_set_iff() returns zero instead of error code on failure, fix
    from Wei Yongjun.

 9) Like GARP, 802 MRP needs to hold the app->lock when adding MAD
    events and queueing PDUs.  Fix from David Ward.

10) Build fix, MVMDIO needs PHYLIB, from Thomas Petazzoni..

11) Fix mac80211 static with ipv6 modular build, from Cong Wang.

12) If userland specifies a path cost explicitly, do not override it
    when the carrier state changes.  From Stephen Hemminger.

13) mvnets calculates the TX queue to use incorrectly resulting in
    garbage pointer derefs and crashes, fix from Willy Tarreau.

14) cdc_mbim does erroneous sizeof(ETH_HLEN).  Fix from Bjorn Mork.

15) IP fragmentation can leak a refcount-less route out from an RCU
    protected section.  This results in crashes and all sorts of hard to
    diagnose behavior.  Fix from Eric Dumazet.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (24 commits)
  qlcnic: fix beaconing test for 82xx adapter
  net: drop dst before queueing fragments
  net: fec: fix regression in link change accounting
  net: cdc_mbim: remove bogus sizeof()
  drivers: net: ethernet: cpsw: get slave VLAN id from slave node instead of cpsw node
  net: mvneta: fix improper tx queue usage in mvneta_tx()
  esp4: fix error return code in esp_output()
  bridge: make user modified path cost sticky
  ipv6: statically link register_inet6addr_notifier()
  net: mvmdio: add select PHYLIB
  net/802/mrp: fix possible race condition when calling mrp_pdu_queue()
  tuntap: fix error return code in tun_set_iff()
  be2net: take care of __vlan_put_tag return value
  can: sja1000: fix handling on dt properties on little endian systems
  can: mcp251x: add missing IRQF_ONESHOT to request_threaded_irq
  netfilter: nf_nat: fix race when unloading protocol modules
  tcp: Reallocate headroom if it would overflow csum_start
  stmmac: prevent interrupt loop with MMC RX IPC Counter
  bonding: IFF_BONDING is not stripped on enslave failure
  bonding: fix netdev event NULL pointer dereference
  ...

Commit b4cbb197 ("vm: add vm_iomap_memory() helper function") added
a helper function wrapper around io_remap_pfn_range(), and every other
architecture defined it in <asm/pgtable.h>.

The s390 choice of <asm/io.h> may make sense, but is not very convenient
for this case, and gratuitous differences like that cause unexpected errors like this:

   mm/memory.c: In function 'vm_iomap_memory':
   mm/memory.c:2439:2: error: implicit declaration of function 'io_remap_pfn_range' [-Werror=implicit-function-declaration]

Glory be the kbuild test robot who noticed this, bisected it, and
reported it to the guilty parties (ie me).

Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

o Commit 319ecf12
  ("qlcnic: 83xx sysfs routines") introduced regression
  for beaconing test while refactoring 82xx code. This patch is to
  revert code to fix beaconing test for 82xx adapter.
Signed-off-by: Himanshu Madhani <himanshu.madhani@qlogic.com>
Signed-off-by: Shahed Shaikh <shahed.shaikh@qlogic.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Commit 4a94445c (net: Use ip_route_input_noref() in input path)
added a bug in IP defragmentation handling, as non refcounted
dst could escape an RCU protected section.

Commit 64f3b9e2 (net: ip_expire() must revalidate route) fixed
the case of timeouts, but not the general problem.

Tom Parkin noticed crashes in UDP stack and provided a patch,
but further analysis permitted us to pinpoint the root cause.

Before queueing a packet into a frag list, we must drop its dst,
as this dst has limited lifetime (RCU protected)

When/if a packet is finally reassembled, we use the dst of the very
last skb, still protected by RCU and valid, as the dst of the
reassembled packet.

Use same logic in IPv6, as there is no need to hold dst references.
Reported-by: Tom Parkin <tparkin@katalix.com>
Tested-by: Tom Parkin <tparkin@katalix.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull ARM fix from Russell King:
 "A build fix for an incomplete change to the ARM cpu suspend code"

* branch 'fixes' of git://git.linaro.org/people/rmk/linux-arm:
  ARM: Do 15e0d9e3 (ARM: pm: let platforms select cpu_suspend support) properly

Pull kvm fixes from Marcelo Tosatti:
 "PPC and ARM KVM fixes"

* git://git.kernel.org/pub/scm/virt/kvm/kvm:
  ARM: KVM: fix L_PTE_S2_RDWR to actually be Read/Write
  ARM: KVM: fix KVM_CAP_ARM_SET_DEVICE_ADDR reporting
  kvm/ppc/e500: eliminate tlb_refs
  kvm/ppc/e500: g2h_tlb1_map: clear old bit before setting new bit
  kvm/ppc/e500: h2g_tlb1_rmap: esel 0 is valid
  kvm/powerpc/e500mc: fix tlb invalidation on cpu migration

Pull powerpc fixes from Stephen Rothwell:
 "Three regresions in the PowerPC code.  One from v3.7 the others from
  this merge window."

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sfr/next-fixes:
  powerpc: add a missing label in resume_kernel
  powerpc: Fix audit crash due to save/restore PPR changes
  powerpc: fix compiling CONFIG_PPC_TRANSACTIONAL_MEM when CONFIG_ALTIVEC=n

Pull kbuild fix from Michal Marek:
 "Fix for a missing dependency when generating scripts/mod/devicetable-offsets.h.
  This dependency got introduced in v3.9-rc1."

* 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:
  kbuild: generate generic headers before recursing into scripts

Pull input fixes from Dmitry Torokhov:
 "Two small fixups to the Wacom driver"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: wacom - correct reported resolution for Intuos4 Wireless
  Input: wacom - fix "can not retrieve extra class descriptor" for 24HDT

* 'kvm-arm-fixes-3.9' of git://github.com/columbia/linux-kvm-arm:
  ARM: KVM: fix L_PTE_S2_RDWR to actually be Read/Write
  ARM: KVM: fix KVM_CAP_ARM_SET_DEVICE_ADDR reporting
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Various drivers end up replicating the code to mmap() their memory
buffers into user space, and our core memory remapping function may be
very flexible but it is unnecessarily complicated for the common cases
to use.

Our internal VM uses pfn's ("page frame numbers") which simplifies
things for the VM, and allows us to pass physical addresses around in a
denser and more efficient format than passing a "phys_addr_t" around,
and having to shift it up and down by the page size.  But it just means
that drivers end up doing that shifting instead at the interface level.

It also means that drivers end up mucking around with internal VM things
like the vma details (vm_pgoff, vm_start/end) way more than they really
need to.

So this just exports a function to map a certain physical memory range
into user space (using a phys_addr_t based interface that is much more
natural for a driver) and hides all the complexity from the driver.
Some drivers will still end up tweaking the vm_page_prot details for
things like prefetching or cacheability etc, but that's actually
relevant to the driver, rather than caring about what the page offset of
the mapping is into the particular IO memory region.
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Looks like our L_PTE_S2_RDWR definition is slightly wrong,
and is actually write only (see ARM ARM Table B3-9, Stage 2 control
of access permissions). Didn't make a difference for normal pages,
as we OR the flags together, but I'm still wondering how it worked
for Stage-2 mapped devices, such as the GIC.

Brown paper bag time, again.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>

Commit 3401d546 (KVM: ARM: Introduce KVM_ARM_SET_DEVICE_ADDR
ioctl) added support for the KVM_CAP_ARM_SET_DEVICE_ADDR capability,
but failed to add a break in the relevant case statement, returning
the number of CPUs instead.

Luckilly enough, the CONFIG_NR_CPUS=0 patch hasn't been merged yet
(https://lkml.org/lkml/diff/2012/3/31/131/1), so the bug wasn't
noticed.

Just give it a break!
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>

A link-down isn't properly saved in the FEC state, so we wouldn't restart the
FEC after a repeated link-up.

Regression was introduced with commit
d97e7497 "net: fec: restart the FEC when PHY speed changes"
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Tested-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The intention was to test against the constant, not the size of
the constant.
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>

Dual EMAC slave VLAN id must be got from slave node instead of cpsw node as
VLAN id for each slave will be different.
Reported-by: Mark Jackson <mpfj-list@mimc.co.uk>
Signed-off-by: Mugunthan V N <mugunthanvnm@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mvneta_tx() was using a static tx queue number causing crashes as
soon as a little bit of traffic was sent via the interface, because
it is normally expected that the same queue should be used as in
dev_queue_xmit().

As suggested by Ben Hutchings, let's use skb_get_queue_mapping() to
get the proper Tx queue number, and use alloc_etherdev_mqs() instead
of alloc_etherdev_mq() to create the queues.

Both my Mirabox and my OpenBlocks AX3 used to crash without this patch
and don't anymore with it. The issue appeared in 3.8 but became more
visible after the fix allowing GSO to be enabled.

Original work was done by Dmitri Epshtein and Thomas Petazzoni. I
just adapted it to take care of Ben's comments.
Signed-off-by: Willy Tarreau <w@1wt.eu>
Cc: Dmitri Epshtein <dima@marvell.com>
Cc: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Cc: Gregory CLEMENT <gregory.clement@free-electrons.com>
Cc: Ben Hutchings <bhutchings@solarflare.com>
Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix to return a negative error code from the error handling
case instead of 0, as returned elsewhere in this function.
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Keep a STP port path cost value if it was set by a user.
Don't replace it with the link-speed based path cost
whenever the link goes down and comes back up.
Reported-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Reported-by: Przemo Firszt <przemo@firszt.eu>
Signed-off-by: Jason Gerecke <killertofu@gmail.com>
Tested-by: Przemo Firszt <przemo@firszt.eu>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>

Merge branches 'timers-urgent-for-linus', 'irq-urgent-for-linus' and 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull {timer,irq,core} fixes from Thomas Gleixner:

 - timer: bug fix for a cpu hotplug race.

 - irq: single bugfix for a wrong return value, which prevents the
   calling function to invoke the software fallback.

 - core: bugfix which plugs two race confitions which can cause hotplug
   per cpu threads to end up on the wrong cpu.

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  hrtimer: Don't reinitialize a cpu_base lock on CPU_UP

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  irqchip: gic: fix irq_trigger return

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  kthread: Prevent unpark race which puts threads on the wrong cpu

A label 0 was missed in the patch a9c4e541 (powerpc/kprobe: Complete
kprobe and migrate exception frame). This will cause the kernel
branch to an undetermined address if there really has a conflict when
updating the thread flags.
Signed-off-by: Kevin Hao <haokexin@gmail.com>
Cc: stable@vger.kernel.org
Acked-By: Tiejun Chen <tiejun.chen@windriver.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>

The current mainline crashes when hitting userspace with the following:

kernel BUG at kernel/auditsc.c:1769!
cpu 0x1: Vector: 700 (Program Check) at [c000000023883a60]
    pc: c0000000001047a8: .__audit_syscall_entry+0x38/0x130
    lr: c00000000000ed64: .do_syscall_trace_enter+0xc4/0x270
    sp: c000000023883ce0
   msr: 8000000000029032
  current = 0xc000000023800000
  paca    = 0xc00000000f080380   softe: 0        irq_happened: 0x01
    pid   = 1629, comm = start_udev
kernel BUG at kernel/auditsc.c:1769!
enter ? for help
[c000000023883d80] c00000000000ed64 .do_syscall_trace_enter+0xc4/0x270
[c000000023883e30] c000000000009b08 syscall_dotrace+0xc/0x38
 --- Exception: c00 (System Call) at 0000008010ec50dc

Bisecting found the following patch caused it:

commit 44e9309f
Author: Haren Myneni <haren@linux.vnet.ibm.com>
powerpc: Implement PPR save/restore

It was found this patch corrupted r9 when calling
SET_DEFAULT_THREAD_PPR()

Using r10 as a scratch register instead of r9 solved the problem.
Signed-off-by: Alistair Popple <alistair@popple.id.au>
Acked-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>

Tomas reported the following build error:

net/built-in.o: In function `ieee80211_unregister_hw':
(.text+0x10f0e1): undefined reference to `unregister_inet6addr_notifier'
net/built-in.o: In function `ieee80211_register_hw':
(.text+0x10f610): undefined reference to `register_inet6addr_notifier'
make: *** [vmlinux] Error 1

when built IPv6 as a module.

So we have to statically link these symbols.
Reported-by: Tomas Melin <tomas.melin@iki.fi>
Cc: Tomas Melin <tomas.melin@iki.fi>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: YOSHIFUJI Hidaki <yoshfuji@linux-ipv6.org>
Signed-off-by: Cong Wang <amwang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull x86 fixes from Ingo Molnar:
 "Misc fixes"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mm: Flush lazy MMU when DEBUG_PAGEALLOC is set
  x86/mm/cpa/selftest: Fix false positive in CPA self test
  x86/mm/cpa: Convert noop to functional fix
  x86, mm: Patch out arch_flush_lazy_mmu_mode() when running on bare metal
  x86, mm, paravirt: Fix vmalloc_fault oops during lazy MMU updates

Pull scheduler fixes from Ingo Molnar:
 "Misc fixlets"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/cputime: Fix accounting on multi-threaded processes
  sched/debug: Fix sd->*_idx limit range avoiding overflow
  sched_clock: Prevent 64bit inatomicity on 32bit systems
  sched: Convert BUG_ON()s in try_to_wake_up_local() to WARN_ON_ONCE()s

Pull perf fixes from Ingo Molnar:
 "Misc fixlets"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf: Fix error return code
  ftrace: Fix strncpy() use, use strlcpy() instead of strncpy()
  perf: Fix strncpy() use, use strlcpy() instead of strncpy()
  perf: Fix strncpy() use, always make sure it's NUL terminated
  perf: Fix ring_buffer perf_output_space() boundary calculation
  perf/x86: Fix uninitialized pt_regs in intel_pmu_drain_bts_buffer()

Pull drm fixes from Dave Airlie:
 "One fix for a hotplug locking regressions, and one fix for an oops if
  you unplug the monitor at an inopportune moment on the udl device."

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/fb-helper: Fix locking in drm_fb_helper_hotplug_event
  udl: handle EDID failure properly.

Pull m68knommu fix from Greg Ungerer:
 "This contains only a single compilation fix for ColdFire m68k targets
  that use local non-GPIOLIB support."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu:
  m68k: define a local gpio_request_one() function

Pull watchdog fix from Wim Van Sebroeck:
 "It will fix compile errors for the at91rm9200_wdt driver"

* git://www.linux-watchdog.org/linux-watchdog:
  watchdog: Revert the AT91RM9200_WATCHDOG dependency

Pull one more btrfs fix from Chris Mason:
 "This has a recent fix from Josef for our tree log replay code.  It
  fixes problems where the inode counter for the number of bytes in the
  file wasn't getting updated properly during fsync replay.

  The commit did get rebased this morning, but it was only to clean up
  the subject line.  The code hasn't changed."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: make sure nbytes are right after log replay

Merge tag 'trace-fixes-v3.9-rc-v3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace

Pull ftrace fixes from Steven Rostedt:
 "Namhyung Kim found and fixed a bug that can crash the kernel by simply
  doing: echo 1234 | tee -a /sys/kernel/debug/tracing/set_ftrace_pid

  Luckily, this can only be done by root, but still is a nasty bug."

* tag 'trace-fixes-v3.9-rc-v3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  ftrace: Move ftrace_filter_lseek out of CONFIG_DYNAMIC_FTRACE section
  tracing: Fix possible NULL pointer dereferences

Nothing is using it yet, but this will allow us to delay the open-time
checks to use time, without breaking the normal UNIX permission
semantics where permissions are determined by the opener (and the file
descriptor can then be passed to a different process, or the process can
drop capabilities).
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Compiling the at91rm9200_wdt.c driver without at91rm9200
support was leading to several errors:

drivers/built-in.o: In function `at91_wdt_close':
at91_adc.c:(.text+0xc9fe4): undefined reference to `at91_st_base'
drivers/built-in.o: In function `at91_wdt_write':
at91_adc.c:(.text+0xca004): undefined reference to `at91_st_base'
drivers/built-in.o: In function `at91wdt_shutdown':
at91_adc.c:(.text+0xca01c): undefined reference to `at91_st_base'
drivers/built-in.o: In function `at91wdt_suspend':
at91_adc.c:(.text+0xca038): undefined reference to `at91_st_base'
drivers/built-in.o: In function `at91_wdt_open':
at91_adc.c:(.text+0xca0cc): undefined reference to `at91_st_base'
drivers/built-in.o:at91_adc.c:(.text+0xca2c8): more undefined references to
`at91_st_base' follow

So, reverting the modification of the "depends" Kconfig line
introduced by patch a6a1bcd3 (watchdog: at91rm9200: add DT support)
seems to be the good solution.
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Acked-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>

Revert commit 62a3ddef ("vfs: fix spinning prevention in prune_icache_sb").

This commit doesn't look right: since we are looking at the tail of the
list (sb->s_inode_lru.prev) if we want to skip an inode, we should put
it back at the head of the list instead of the tail, otherwise we will
keep spinning on it.

Discovered when investigating why prune_icache_sb came top in perf
reports of a swapping load.
Signed-off-by: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: stable@vger.kernel.org # v3.2+
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Anatol Pomozov identified a race condition that hits module unloading
and re-loading.  To quote Anatol:

 "This is a race codition that exists between kset_find_obj() and
  kobject_put().  kset_find_obj() might return kobject that has refcount
  equal to 0 if this kobject is freeing by kobject_put() in other
  thread.

  Here is timeline for the crash in case if kset_find_obj() searches for
  an object tht nobody holds and other thread is doing kobject_put() on
  the same kobject:

    THREAD A (calls kset_find_obj())     THREAD B (calls kobject_put())
    splin_lock()
                                         atomic_dec_return(kobj->kref), counter gets zero here
                                         ... starts kobject cleanup ....
                                         spin_lock() // WAIT thread A in kobj_kset_leave()
    iterate over kset->list
    atomic_inc(kobj->kref) (counter becomes 1)
    spin_unlock()
                                         spin_lock() // taken
                                         // it does not know that thread A increased counter so it
                                         remove obj from list
                                         spin_unlock()
                                         vfree(module) // frees module object with containing kobj

    // kobj points to freed memory area!!
    kobject_put(kobj) // OOPS!!!!

  The race above happens because module.c tries to use kset_find_obj()
  when somebody unloads module.  The module.c code was introduced in
  commit 6494a93d"

Anatol supplied a patch specific for module.c that worked around the
problem by simply not using kset_find_obj() at all, but rather than make
a local band-aid, this just fixes kset_find_obj() to be thread-safe
using the proper model of refusing the get a new reference if the
refcount has already dropped to zero.

See examples of this proper refcount handling not only in the kref
documentation, but in various other equivalent uses of this pattern by
grepping for atomic_inc_not_zero().

[ Side note: the module race does indicate that module loading and
  unloading is not properly serialized wrt sysfs information using the
  module mutex.  That may require further thought, but this is the
  correct fix at the kobject layer regardless. ]
Reported-analyzed-and-tested-by: Anatol Pomozov <anatol.pomozov@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>