Pull networking fixes from David Miller:

1) TCP can chop up SACK'd SKBs below below the unacked send sequence and
   that breaks lots of stuff.  Fix from Neal Cardwell.

2) There is code in ipv6 to properly join and leave the all-routers
   multicast code when the forwarding setting is changed, but once
   forwarding is turned on, we don't do the join for newly registered
   devices.  Fix from Li Wei.

3) Netfilter's NAT module autoload in ctnetlink drops a spinlock around
   a sleeping call, problem is this code path doesn't actually hold that
   lock.  Fix from Pablo Neira Ayuso.

4) TG3 uses the wrong interfaces to hook into the new byte queue limit
   support.  It uses the device level interfaces, which is fine for
   single queue devices, but on more recent chips this driver supports
   multiqueue so we have to use the multiqueue BQL APIs.  Fix from Tom
   Herbert.

5) r8169 resume fix from Francois Romieu.

6) Add some cxgb4 device IDs, from Vipul Pandya.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  IPv6: Fix not join all-router mcast group when forwarding set.
  caif-hsi: Set default MTU to 4096
  cxgb4vf: Add support for Chelsio's T480-CR and T440-LP-CR adapters
  cxgb4: Add support for Chelsio's T480-CR and T440-LP-CR adapters
  mlx4_core: remove buggy sched_queue masking
  netfilter: nf_conntrack: fix early_drop with reliable event delivery
  bridge: netfilter: don't call iptables on vlan packets if sysctl is off
  netfilter: bridge: fix wrong pointer dereference
  netfilter: ctnetlink: remove incorrect spin_[un]lock_bh on NAT module autoload
  netfilter: ebtables: fix wrong name length while copying to user-space
  r8169: runtime resume before shutdown.
  tcp: fix tcp_shift_skb_data() to not shift SACKed data below snd_una
  tg3: Fix to use multi queue BQL interfaces

It turns out that test-compiling this file on x86-64 doesn't really
help, because much of it is x86-32-specific.  And so I hadn't noticed
the slightly over-eager removal of the 'r' from 'addr' variable despite
thinking I had tested it.
Signed-off-by: Linus "oopsie" Torvalds <torvalds@linux-foundation.org>

Several users of "find_vma_prev()" were not in fact interested in the
previous vma if there was no primary vma to be found either.  And in
those cases, we're much better off just using the regular "find_vma()",
and then "prev" can be looked up by just checking vma->vm_prev.

The find_vma_prev() semantics are fairly subtle (see Mikulas' recent
commit 83cd904d: "mm: fix find_vma_prev"), and the whole "return
prev by reference" means that it generates worse code too.

Thus this "let's avoid using this inconvenient and clearly too subtle
interface when we don't really have to" patch.

Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull CIFS fixes from Steve French

* git://git.samba.org/sfrench/cifs-2.6:
  cifs: fix dentry refcount leak when opening a FIFO on lookup
  CIFS: Fix mkdir/rmdir bug for the non-POSIX case

Commit 6bd4837d ("mm: simplify find_vma_prev()") broke memory
management on PA-RISC.

After application of the patch, programs that allocate big arrays on the
stack crash with segfault, for example, this will crash if compiled
without optimization:

  int main()
  {
	char array[200000];
	array[199999] = 0;
	return 0;
  }

The reason is that PA-RISC has up-growing stack and the stack is usually
the last memory area.  In the above example, a page fault happens above
the stack.

Previously, if we passed too high address to find_vma_prev, it returned
NULL and stored the last VMA in *pprev.  After "simplify find_vma_prev"
change, it stores NULL in *pprev.  Consequently, the stack area is not
found and it is not expanded, as it used to be before the change.

This patch restores the old behavior and makes it return the last VMA in
*pprev if the requested address is higher than address of any other VMA.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Xommit ac563761(genirq: Unmask oneshot irqs when thread was not woken)
fails to unmask when a !IRQ_ONESHOT threaded handler is handled by
handle_level_irq.

This happens because thread_mask is or'ed unconditionally in
irq_wake_thread(), but for !IRQ_ONESHOT interrupts never cleared.  So
the check for !desc->thread_active fails and keeps the interrupt
disabled.

Keep the thread_mask zero for !IRQ_ONESHOT interrupts.

Document the thread_mask magic while at it.
Reported-and-tested-by: Sven Joachim <svenjoac@gmx.de>
Reported-and-tested-by: Stefan Lippers-Hollmann <s.l-h@gmx.de>
Cc: stable@vger.kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

When forwarding was set and a new net device is register,
we need add this device to the all-router mcast group.
Signed-off-by: Li Wei <lw@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Currently error is -ENOMEM when rejecting VM_GROWSDOWN|VM_GROWSUP
from shared anonymous: hoist the file case's -EINVAL up for both.
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Default MTU for CAIF HSI was wrongly set to 15 * 4092 bytes.
The patch sets default MTU size to 4096.
Signed-off-by: Sjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds PCI device ids for Chelsio's T480-CR and T440-LP-CR
adapters.
Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch adds PCI device ids for Chelsio's T480-CR and T440-LP-CR
adapters.
Signed-off-by: Vipul Pandya <vipul@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fixes a bug introduced by commit fe9a2603, where the priority bits
in the schedule queue field were masked out.
Signed-off-by: Amir Vadai <amirv@mellanox.co.il>
Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
Signed-off-by: David S. Miller <davem@davemloft.net>

If reliable event delivery is enabled and ctnetlink fails to deliver
the destroy event in early_drop, the conntrack subsystem cannot
drop any the candidate flow that was planned to be evicted.
Reported-by: Kerin Millar <kerframil@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

When net.bridge.bridge-nf-filter-vlan-tagged is 0 (default), vlan packets
arriving should not be sent to ip(6)tables by bridge netfilter.

However, it turns out that we currently always send VLAN packets to
netfilter, if ..
a), CONFIG_VLAN_8021Q is enabled ; or
b), CONFIG_VLAN_8021Q is not set but rx vlan offload is enabled
   on the bridge port.

This is because bridge netfilter treats skb with
skb->protocol == ETH_P_IP{V6} as "non-vlan packet".

With rx vlan offload on or CONFIG_VLAN_8021Q=y, the vlan header has
already been removed here, and we cannot rely on skb->protocol alone.

Fix this by only using skb->protocol if the skb has no vlan tag,
or if a vlan tag is present and filter-vlan-tagged bridge netfilter
sysctl is enabled.

We cannot remove the skb->protocol == htons(ETH_P_8021Q) test
because the vlan tag is still around in the CONFIG_VLAN_8021Q=n &&
"ethtool -K $itf rxvlan off" case.

reproducer:
iptables -t raw -I PREROUTING -i br0
iptables -t raw -I PREROUTING -i br0.1

Then send packets to an ip address configured on br0.1 interface.
Even with net.bridge.bridge-nf-filter-vlan-tagged=0, the 1st rule
will match instead of the 2nd one.

With this patch applied, the 2nd rule will match instead.
In the non-local address case, netfilter won't be consulted after
this patch unless the sysctl is switched on.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

In adf7ff8, a invalid dereference was added in ebt_make_names.

CC [M]  net/bridge/netfilter/ebtables.o
net/bridge/netfilter/ebtables.c: In function `ebt_make_names':
net/bridge/netfilter/ebtables.c:1371:20: warning: `t' may be used uninitialized in this function [-Wuninitialized]
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since 7d367e06, ctnetlink_new_conntrack is called without holding
the nf_conntrack_lock spinlock. Thus, ctnetlink_parse_nat_setup
does not require to release that spinlock anymore in the NAT module
autoload case.
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

user-space ebtables expects 32 bytes-long names, but xt_match names
use 29 bytes. We have to copy less 29 bytes and then, make sure we
fill the remaining bytes with zeroes.
Signed-off-by: Santosh Nayak <santoshprasadnayak@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

With runtime PM, if the ethernet cable is disconnected, the device is
transitioned to D3 state to conserve energy. If the system is shutdown
in this state, any register accesses in rtl_shutdown are dropped on
the floor. As the device was programmed by .runtime_suspend() to wake
on link changes, it is thus brought back up as soon as the link recovers.

Resuming every suspended device through the driver core would slow things
down and it is not clear how many devices really need it now.

Original report and D0 transition patch by Sameer Nanda. Patch has been
changed to comply with advices by Rafael J. Wysocki and the PM folks.
Reported-by: Sameer Nanda <snanda@chromium.org>
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Hayes Wang <hayeswang@realtek.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: David S. Miller <davem@davemloft.net>

This commit fixes tcp_shift_skb_data() so that it does not shift
SACKed data below snd_una.

This fixes an issue whose symptoms exactly match reports showing
tp->sacked_out going negative since 3.3.0-rc4 (see "WARNING: at
net/ipv4/tcp_input.c:3418" thread on netdev).

Since 2008 (832d11c5)
tcp_shift_skb_data() had been shifting SACKed ranges that were below
snd_una. It checked that the *end* of the skb it was about to shift
from was above snd_una, but did not check that the end of the actual
shifted range was above snd_una; this commit adds that check.

Shifting SACKed ranges below snd_una is problematic because for such
ranges tcp_sacktag_one() short-circuits: it does not declare anything
as SACKed and does not increase sacked_out.

Before the fixes in commits cc9a672e
and daef52ba, shifting SACKed ranges
below snd_una happened to work because tcp_shifted_skb() was always
(incorrectly) passing in to tcp_sacktag_one() an skb whose end_seq
tcp_shift_skb_data() had already guaranteed was beyond snd_una. Hence
tcp_sacktag_one() never short-circuited and always increased
tp->sacked_out in this case.

After those two fixes, my testing has verified that shifting SACKed
ranges below snd_una could cause tp->sacked_out to go negative with
the following sequence of events:

(1) tcp_shift_skb_data() sees an skb whose end_seq is beyond snd_una,
    then shifts a prefix of that skb that is below snd_una

(2) tcp_shifted_skb() increments the packet count of the
    already-SACKed prev sk_buff

(3) tcp_sacktag_one() sees the end of the new SACKed range is below
    snd_una, so it short-circuits and doesn't increase tp->sacked_out

(5) tcp_clean_rtx_queue() sees the SACKed skb has been ACKed,
    decrements tp->sacked_out by this "inflated" pcount that was
    missing a matching increase in tp->sacked_out, and hence
    tp->sacked_out underflows to a u32 like 0xFFFFFFFF, which casted
    to s32 is negative.

(6) this leads to the warnings seen in the recent "WARNING: at
    net/ipv4/tcp_input.c:3418" thread on the netdev list; e.g.:
    tcp_input.c:3418  WARN_ON((int)tp->sacked_out < 0);

More generally, I think this bug can be tickled in some cases where
two or more ACKs from the receiver are lost and then a DSACK arrives
that is immediately above an existing SACKed skb in the write queue.

This fix changes tcp_shift_skb_data() to abort this sequence at step
(1) in the scenario above by noticing that the bytes are below snd_una
and not shifting them.
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull arm-soc bug fixes from Arnd Bergmann:
 "Here are all the fixes I got after sending the last pull request.
  These fix mostly regressions on exynos, at91, pxa and ep93xx.

  Signed-off-by: Arnd Bergmann <arnd@arndb.de>"

* tag 'fixes-3.3-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  ARM: ep93xx: convert vision_ep9307 to MULTI_IRQ_HANDLER
  ARM: EXYNOS: fix touchscreen IRQ setup on Universal C210 board
  ARM: pxa: fix invalid mfp pin issue
  ARM: pxa: remove duplicated registeration on pxa-gpio
  ARM: pxa: add dummy clock for pxa25x and pxa27x
  ARM: S3C24XX: DMA resume regression fix
  ARM: S3C24XX: Fix restart on S3C2442
  ARM: SAMSUNG: Fix memory size for hsotg
  ARM: at91/dma: DMA controller registering with DT support
  ARM: at91/dma: remove platform data from DMA controller

Pull s390 regression fix from Martin Schwidefsky:
 "It is a fix for a regression that has been introduced with git commit
  25f269f1 - "[S390] qdio: EQBS retry after CCQ 96" - and if possible
  we would like to have working code for the fcp data router in 3.3."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  [S390] qdio: fix handler function arguments for zfcp data router

Pull regulator updates from Mark Brown:
 "A simple fix that's obvious from inspection.  There's no mainline
  users of this driver yet (there's some i.MX platforms which will use
  it)."

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: Fix mask parameter in da9052_reg_update calls

kasprintf() (and potentially other functions that I didn't run across so
far) want to evaluate argument lists twice.  Caring to do so for the
primary list is obviously their job, but they can't reasonably be
expected to check the format string for instances of %pV, which however
need special handling too: On architectures like x86-64 (as opposed to
e.g.  ix86), using the same argument list twice doesn't produce the
expected results, as an internally managed cursor gets updated during
the first run.

Fix the problem by always acting on a copy of the original list when
handling %pV.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Why is memcg's swap accounting so broken? Insane counts, wrong
ownership, unfreeable structures, which later get freed and then
accessed after free.

Turns out to be a tiny a little 3.3-rc1 regression in 9fb4b7cc
"page_cgroup: add helper function to get swap_cgroup": the helper
function (actually named lookup_swap_cgroup()) returns an address using
void* arithmetic, but the structure in question is a short.
Signed-off-by: Hugh Dickins <hughd@google.com>
Reviewed-by: Bob Liu <lliubbo@gmail.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <jweiner@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

* 'fixes' of git://github.com/hzhuang1/linux: (3 commits)
  ARM: pxa: fix invalid mfp pin issue
  ARM: pxa: remove duplicated registeration on pxa-gpio
  ARM: pxa: add dummy clock for pxa25x and pxa27x

Includes an update to v3.3-rc6

Merge branch 'v3.3-samsung-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung into fixes

* 'v3.3-samsung-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/kgene/linux-samsung:
  ARM: EXYNOS: fix touchscreen IRQ setup on Universal C210 board
  ARM: S3C24XX: DMA resume regression fix
  ARM: S3C24XX: Fix restart on S3C2442
  ARM: SAMSUNG: Fix memory size for hsotg

As done for the other ep93xx machines in:

commit 9a6879bd
ARM: ep93xx: convert to MULTI_IRQ_HANDLER

Now that there is a generic IRQ handler for multiple VIC devices use it
for vision_ep9307 to help building multi platform kernels.
Signed-off-by: Hartley Sweeten <hsweeten@visionengravers.com>
Acked-by: Ryan Mallon <rmallon@gmail.com>
Reviewed-by: Jamie Iles <jamie@jamieiles.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Fixes atmel_mxt_ts freeze on Universal C210.
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Kukjin Kim <kgene.kim@samsung.com>

Failure is reported on hx4700 with kernel v3.3-rc1.

__mfp_validate: GPIO20 is invalid pin
__mfp_validate: GPIO21 is invalid pin
__mfp_validate: GPIO15 is invalid pin
__mfp_validate: GPIO78 is invalid pin
__mfp_validate: GPIO79 is invalid pin
__mfp_validate: GPIO80 is invalid pin
__mfp_validate: GPIO33 is invalid pin
__mfp_validate: GPIO48 is invalid pin
__mfp_validate: GPIO49 is invalid pin
__mfp_validate: GPIO50 is invalid pin

Since pxa_last_gpio is used in mfp-pxa2xx driver. But it's only
updated in pxa-gpio driver that run after mfp-pxa2xx driver.

So update the pxa_last_gpio first in mfp-pxa2xx driver.
Reported-by: Paul Parsons <lost.distance@yahoo.com>
Signed-off-by: Haojian Zhuang <haojian.zhuang@gmail.com>

Both reboot (via reboot(RB_AUTOBOOT)) and suspend freeze on hx4700.

Registration of pxa_gpio_syscore_ops is moved into pxa-gpio driver,
but it still exists in arch-pxa directory. It resulsts failure on
reboot and suspend.

Now remove the registration code in arch-pxa.
Reported-by: Paul Parsons <lost.distance@yahoo.com>
Signed-off-by: Haojian Zhuang <haojian.zhuang@gmail.com>

gpio-pxa driver is shared among arch-pxa and arch-mmp. Clock is the
essential component on pxa3xx/pxa95x and arch-mmp. So we need to
define dummy clock in pxa25x/pxa27x instead.

This regression was introduced by the commit "ARM: pxa: add dummy
clock for sa1100-rtc", id a55b5ada.
Reported-by: Jonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: Paul Parsons <lost.distance@yahoo.com>
Tested-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Haojian Zhuang <haojian.zhuang@marvell.com>

Fix tg3 to use BQL multi queue related netdev interfaces since the
device supports multi queue.
Signed-off-by: Tom Herbert <therbert@google.com>
Reported-by: Christoph Lameter <cl@gentwo.org>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Pull perf fixes from Ingo Molnar:
 "It contains three cherry-picked fixes from perf/core, which turned out
  to be more urgent than we originally thought."

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf tools: Handle kernels that don't support attr.exclude_{guest,host}
  perf tools: Change perf_guest default back to false
  perf record: No build id option fails

USB: revert a powerpc EHCI patch

There is just one patch in here, a revert of a powerpc EHCI driver
patch that was reported to cause problems.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

* tag 'usb-3.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  Revert "powerpc/usb: fix issue of CPU halt when missing USB PHY clock"

tty: build fix for 3.3-rc6

This contains one build fix for the powerpc udbg driver that was reported.
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

* tag 'tty-3.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  tty/powerpc: early udbg consoles can't be modules

Pull md fixes from Neil Brown:
 "Three fixes for md in 3.3-rc: Two relate to the recently added drive
  replacement.  One fixes the problem where a read error in RAID10 would
  sometimes be retried indefinitely."

* tag 'md-3.3-fixes' of git://neil.brown.name/md:
  md/raid10: fix assembling of arrays with replacement devices.
  md/raid10: fix handling of error on last working device in array.
  md/raid1: fix buglet in md_raid1_contested.

Merge the emailed seties of 19 patches from Andrew Morton

* akpm:
  rapidio/tsi721: fix queue wrapping bug in inbound doorbell handler
  memcg: fix mapcount check in move charge code for anonymous page
  mm: thp: fix BUG on mm->nr_ptes
  alpha: fix 32/64-bit bug in futex support
  memcg: fix GPF when cgroup removal races with last exit
  debugobjects: Fix selftest for static warnings
  floppy/scsi: fix setting of BIO flags
  memcg: fix deadlock by inverting lrucare nesting
  drivers/rtc/rtc-r9701.c: fix crash in r9701_remove()
  c2port: class_create() returns an ERR_PTR
  pps: class_create() returns an ERR_PTR, not NULL
  hung_task: fix the broken rcu_lock_break() logic
  vfork: kill PF_STARTING
  coredump_wait: don't call complete_vfork_done()
  vfork: make it killable
  vfork: introduce complete_vfork_done()
  aio: wake up waiters when freeing unused kiocbs
  kprobes: return proper error code from register_kprobe()
  kmsg_dump: don't run on non-error paths by default

Fix a bug that causes a kernel panic when the number of received doorbells
is larger than number of entries in the inbound doorbell queue (current
default value = 512).

Another possible indication for this bug is large number of spurious
doorbells reported by tsi721 driver after reaching the queue size maximum.
Signed-off-by: Alexandre Bounine <alexandre.bounine@idt.com>
Cc: Chul Kim <chul.kim@idt.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: <stable@vger.kernel.org>		[3.2.x+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Currently the charge on shared anonyous pages is supposed not to moved in
task migration.  To implement this, we need to check that mapcount > 1,
instread of > 2.  So this patch fixes it.
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Dave Jones reports a few Fedora users hitting the BUG_ON(mm->nr_ptes...)
in exit_mmap() recently.

Quoting Hugh's discovery and explanation of the SMP race condition:

  "mm->nr_ptes had unusual locking: down_read mmap_sem plus
   page_table_lock when incrementing, down_write mmap_sem (or mm_users
   0) when decrementing; whereas THP is careful to increment and
   decrement it under page_table_lock.

   Now most of those paths in THP also hold mmap_sem for read or write
   (with appropriate checks on mm_users), but two do not: when
   split_huge_page() is called by hwpoison_user_mappings(), and when
   called by add_to_swap().

   It's conceivable that the latter case is responsible for the
   exit_mmap() BUG_ON mm->nr_ptes that has been reported on Fedora."

The simplest way to fix it without having to alter the locking is to make
split_huge_page() a noop in nr_ptes terms, so by counting the preallocated
pagetables that exists for every mapped hugepage.  It was an arbitrary
choice not to count them and either way is not wrong or right, because
they are not used but they're still allocated.
Reported-by: Dave Jones <davej@redhat.com>
Reported-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Josh Boyer <jwboyer@redhat.com>
Cc: <stable@vger.kernel.org>	[3.0.x, 3.1.x, 3.2.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>