drm/nouveau: rework init ordering so nv50_instmem.c can be less bad

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50: calculate vram reordering block size

Will be used at a later point when we plug in an alternative VRAM memory
manager for GeForce 8+ boards.

Based on pscnv code to do the same.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Marcin Kościelnicki <koriakin@0x04.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50: allow gpuobjs that aren't mapped into aperture

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: simplify fake gpu objects

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: remove nouveau_gpuobj_ref completely, replace with sanity

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: rebase per-channel pramin heap offsets to 0

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: modify object accessors, offset in bytes rather than dwords

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: move ramht code out of nouveau_object.c, nothing to see here

Reviewed-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: have nv_mask return original register value

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv40: Try to set up CRE_LCD even if it has unknown bits set.

They don't seem to do anything useful, and we really want to program
CRE_LCD if we aren't lucky enough to find the right CRTC binding
already set.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv17-nv4x: Fix analog load detection false positive on rare occasions.

On some boards the residual current DAC outputs can draw when they're
disconnected can be high enough to give a false load detection
positive (I've only seen it in the S-video luma output of some cards,
but just to be sure). The output line capacitance is limited and
sampling twice should fix it reliably.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: require explicit unmap of kmapped bos

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: remove second map of notifier bo

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: move check for no-op bo move before memcpy fallback

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv50: add new accelerated bo move funtion

Hopefully this one will be better able to cope with moving tiled buffers
around without getting them all scrambled as a result.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nv20: Use the nv30 CRTC bandwidth calculation code.

nv2x CRTC FIFOs are as large as in nv3x (4kB it seems), and the FIFO
control registers have the same layout: we can make them share the
same implementation.

Previously we were using the nv1x code, but the calculated FIFO
watermarks are usually too low for nv2x and they cause horrible
scanout artifacts. They've gone unnoticed until now because we've been
leaving one of the bandwidth regs uninitialized (CRE 47, which
contains the most significant bits of FFLWM), so everything seemed to
work fine except in some cases after a cold boot, depending on the
memory bandwidth and pixel clocks used.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drm/nouveau: Fix suspend on some nv4x AGP cards.

On some nv4x cards (specifically, the ones that use an internal
PCIE->AGP bridge) the AGP controller state isn't preserved after a
suspend/resume cycle, and the AGP control registers have moved from
0x18xx to 0x100xx, so the FW check in nouveau_mem_reset_agp() doesn't
quite work. Check "dev->agp->mode" instead.

Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

Merge remote branch 'origin/master' of /home/airlied/kernel//linux-2.6 into drm-core-next

hugetlb, rmap: add BUG_ON(!PageLocked) in hugetlb_add_anon_rmap()

Confirming page lock is held in hugetlb_add_anon_rmap() may be useful
to detect possible future problems.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb, rmap: fix confusing page locking in hugetlb_cow()

The "if (!trylock_page)" block in the avoidcopy path of hugetlb_cow()
looks confusing and is buggy.  Originally this trylock_page() was
intended to make sure that old_page is locked even when old_page !=
pagecache_page, because then only pagecache_page is locked.

This patch fixes it by moving page locking into hugetlb_fault().

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb, rmap: use hugepage_add_new_anon_rmap() in hugetlb_cow()

Obviously, setting anon_vma for COWed hugepage should be done
by hugepage_add_new_anon_rmap() to scan vmas faster.
This patch fixes it.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb, rmap: always use anon_vma root pointer

This patch applies Andrea's fix given by the following patch into hugepage
rmapping code:

  commit 288468c334e98aacbb7e2fb8bde6bc1adcd55e05
  Author: Andrea Arcangeli <aarcange@redhat.com>
  Date:   Mon Aug 9 17:19:09 2010 -0700

This patch uses anon_vma->root and avoids unnecessary overwriting when
anon_vma is already set up.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge branch 'kvm-updates/2.6.36' of git://git./virt/kvm/kvm

* 'kvm-updates/2.6.36' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: Fix reboot on Intel hosts
  KVM: fix irqfd assign/deassign race

Merge branch 'for_linus' of git://git./linux/kernel/git/mjg59/platform-drivers-x86

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mjg59/platform-drivers-x86:
  thinkpad-acpi: avoid keymap pitfall
  Change MSI LAPTOP driver maintainer to Lee, Chun-Yi

thinkpad-acpi: avoid keymap pitfall

Change the code so that it will use the correct size for keymap entries.
Do it in a way that makes it harder to screw it up in the future.

Reported-by: Jaime Velasco Juan <jsagarribay@gmail.com>
Signed-off-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

Change MSI LAPTOP driver maintainer to Lee, Chun-Yi

Change MSI LAPTOP driver maintainer from Lennart Poettering to Lee, Chun-Yi.

MSI is a Taiwan OEM company, Lee, Chun-Yi can more easy to contact with MSI
and maintain msi-laptop driver.
Thank's for Lennart Poettering's contribute, Lee, Chun-Yi will base on his
article to continue maintain the msi-laptop driver.

Signed-off-by: Lee, Chun-Yi <jlee@novell.com>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

MN10300: Arch doesn't support HAVE_ARCH_TRACEHOOK

Remove specification of HAVE_ARCH_TRACEHOOK for MN10300 as the arch does not
support it at this time.

Signed-off-by: Mark Salter <msalter@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MN10300: Fix SIGRTMAX

SIGRTMAX should be _NSIG not _NSIG-1.

Signed-off-by: Mark Salter <msalter@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MN10300: Fix endianess of ext2 bitops

The MN10300 arch ext2 bitops assume a big-endian kernel, but the MN10300
arch only runs in little-endian mode.

Reported-by: Akira Takeuchi <takeuchi.akr@jp.panasonic.com>
Signed-off-by: Mark Salter <msalter@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge branch 'for-linus' of git://git./linux/kernel/git/tj/percpu

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu:
  percpu: fix pcpu_last_unit_cpu

KVM: Fix reboot on Intel hosts

When we reboot, we disable vmx extensions or otherwise INIT gets blocked.
If a task on another cpu hits a vmx instruction, it will fault if vmx is
disabled.  We trap that to avoid a nasty oops and spin until the reboot
completes.

Problem is, we sleep with interrupts disabled.  This blocks smp_send_stop()
from running, and the reboot process halts.

Fix by enabling interrupts before spinning.

KVM-Stable-Tag.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

KVM: fix irqfd assign/deassign race

I think I see the following (theoretical) race:

During irqfd assign, we drop irqfds lock before we
schedule inject work. Therefore, deassign running
on another CPU could cause shutdown and flush to run
before inject, causing user after free in inject.

A simple fix it to schedule inject under the lock.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Gregory Haskins <ghaskins@novell.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

ipmi: fix hardcoded ipmi device exit path warning

When modprobe.conf has
options ipmi_si type="kcs" ports=0xCA2 regspacings="4"

ipmi_si can be loaded properly, but when try to unload it get:

Sep 20 15:00:27 xx abrt: Kerneloops: Reported 1 kernel oopses to Abrt
Sep 20 15:00:27 xx abrtd: Directory 'kerneloops-1285020027-1' creation detected
Sep 20 15:00:27 xx abrtd: New crash /var/spool/abrt/kerneloops-1285020027-1, processing
Sep 20 15:01:09 xx kernel: ------------[ cut here ]------------
Sep 20 15:01:09 xx kernel: WARNING: at drivers/base/driver.c:262 driver_unregister+0x8a/0xa0()
Sep 20 15:01:09 xx kernel: Hardware name: Sun Fire x4800
Sep 20 15:01:09 xx kernel: Unexpected driver unregister!
Sep 20 15:01:09 xx kernel: Modules linked in: ipmi_si(-) ipmi_msghandler ip6table_filter ip6_tables ebtable_nat ebtables ipt_MASQUERADE iptable_nat nf_nat bridge stp llc autofs4 sunrpc cpufreq_ondemand acpi_cpufreq freq_table mperf xt_physdev be2iscsi iscsi_boot_sysfs bnx2i cnic uio cxgb3i iw_cxgb3 cxgb3 mdio ib_iser rdma_cm ib_cm iw_cm ib_sa ib_mad ib_core ib_addr ipv6 iscsi_tcp libiscsi_tcp libiscsi scsi_transport_iscsi dm_mirror dm_region_hash dm_log dm_mod vhost_net macvtap macvlan tun kvm_intel kvm uinput sg ses enclosure ahci libahci pcspkr i2c_i801 i2c_core iTCO_wdt iTCO_vendor_support igb dca i7core_edac edac_core ext3 jbd mbcache sd_mod crc_t10dif megaraid_sas [last unloaded: ipmi_devintf]
Sep 20 15:01:09 xx kernel: Pid: 10625, comm: modprobe Tainted: G        W   2.6.36-rc5-tip+ #6
Sep 20 15:01:09 xx kernel: Call Trace:
Sep 20 15:01:09 xx kernel: [<ffffffff810600df>] warn_slowpath_common+0x7f/0xc0
Sep 20 15:01:09 xx kernel: [<ffffffff810601d6>] warn_slowpath_fmt+0x46/0x50
Sep 20 15:01:09 xx kernel: [<ffffffff812ff60a>] driver_unregister+0x8a/0xa0
Sep 20 15:01:09 xx kernel: [<ffffffff812ae112>] pnp_unregister_driver+0x12/0x20
Sep 20 15:01:09 xx kernel: [<ffffffffa01d0327>] cleanup_ipmi_si+0x3c/0xa7 [ipmi_si]
Sep 20 15:01:09 xx kernel: [<ffffffff81099a60>] sys_delete_module+0x1a0/0x270
Sep 20 15:01:09 xx kernel: [<ffffffff814b7070>] ? do_page_fault+0x150/0x320
Sep 20 15:01:09 xx kernel: [<ffffffff8100b072>] system_call_fastpath+0x16/0x1b
Sep 20 15:01:09 xx kernel: ---[ end trace 0d1967161adcee0d ]---

We need to check if ipmi_pnp_driver is loaded before we try to unload it.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rtc: s3c: balance state changes of wakeup flag

This change resolves a problem about unbalanced calls of
enable_irq_wakeup() and disable_irq_wakeup() for alarm interrupt.

Bug reproduction:

root@eb600:~# echo 0 > /sys/class/rtc/rtc0/wakealarm

WARNING: at kernel/irq/manage.c:361 set_irq_wake+0x7c/0xe4()
 Unbalanced IRQ 46 wake disable
Modules linked in:
[<c0025708>] (unwind_backtrace+0x0/0xd8) from [<c003358c>] (warn_slowpath_common+0x44/0x5c)
[<c003358c>] (warn_slowpath_common+0x44/0x5c) from [<c00335dc>] (warn_slowpath_fmt+0x24/0x30)
[<c00335dc>] (warn_slowpath_fmt+0x24/0x30) from [<c0058c20>] (set_irq_wake+0x7c/0xe4)
[<c0058c20>] (set_irq_wake+0x7c/0xe4) from [<c01b5e80>] (s3c_rtc_setalarm+0xa8/0xb8)
[<c01b5e80>] (s3c_rtc_setalarm+0xa8/0xb8) from [<c01b47a0>] (rtc_set_alarm+0x60/0x74)
[<c01b47a0>] (rtc_set_alarm+0x60/0x74) from [<c01b5a98>] (rtc_sysfs_set_wakealarm+0xc8/0xd8)
[<c01b5a98>] (rtc_sysfs_set_wakealarm+0xc8/0xd8) from [<c01891ec>] (dev_attr_store+0x20/0x24)
[<c01891ec>] (dev_attr_store+0x20/0x24) from [<c00be934>] (sysfs_write_file+0x104/0x13c)
[<c00be934>] (sysfs_write_file+0x104/0x13c) from [<c0080e7c>] (vfs_write+0xb0/0x158)
[<c0080e7c>] (vfs_write+0xb0/0x158) from [<c0080fcc>] (sys_write+0x3c/0x68)
[<c0080fcc>] (sys_write+0x3c/0x68) from [<c0020ec0>] (ret_fast_syscall+0x0/0x28)

Signed-off-by: Vladimir Zapolskiy <vzapolskiy@gmail.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Ben Dooks <ben@fluff.org.uk>
Cc: Atul Dahiya <atul.dahiya@samsung.com>
Cc: Taekgyun Ko <taeggyun.ko@samsung.com>
Cc: Kukjin Kim <kgene.kim@samsung.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mmap: call unlink_anon_vmas() in __split_vma() in case of error

If __split_vma fails because of an out of memory condition the
anon_vma_chain isn't teardown and freed potentially leading to rmap walks
accessing freed vma information plus there's a memleak.

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Johannes Weiner <jweiner@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rmap: fix walk during fork

The below bug in fork led to the rmap walk finding the parent huge-pmd
twice instead of just once, because the anon_vma_chain objects of the
child vma still point to the vma->vm_mm of the parent.

The patch fixes it by making the rmap walk accurate during fork.  It's not
a big deal normally but it worth being accurate considering the cost is
the same.

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Johannes Weiner <jweiner@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/pci/intel-iommu.c: fix build with older gcc's

drivers/pci/intel-iommu.c: In function `__iommu_calculate_agaw':
drivers/pci/intel-iommu.c:437: sorry, unimplemented: inlining failed in call to 'width_to_agaw': function body not available
drivers/pci/intel-iommu.c:445: sorry, unimplemented: called from here

Move the offending function (and its siblings) to top-of-file, remove the
forward declaration.

Addresses https://bugzilla.kernel.org/show_bug.cgi?id=17441

Reported-by: Martin Mokrejs <mmokrejs@ribosome.natur.cuni.cz>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oom: filter unkillable tasks from tasklist dump

/proc/sys/vm/oom_dump_tasks is enabled by default, so it's necessary to
limit as much information as possible that it should emit.

The tasklist dump should be filtered to only those tasks that are eligible
for oom kill.  This is already done for memcg ooms, but this patch extends
it to both cpuset and mempolicy ooms as well as init.

In addition to suppressing irrelevant information, this also reduces
confusion since users currently don't know which tasks in the tasklist
aren't eligible for kill (such as those attached to cpusets or bound to
mempolicies with a disjoint set of mems or nodes, respectively) since that
information is not shown.

Signed-off-by: David Rientjes <rientjes@google.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/video/sis/sis_main.c: prevent reading uninitialized stack memory

The FBIOGET_VBLANK device ioctl allows unprivileged users to read 16 bytes
of uninitialized stack memory, because the "reserved" member of the
fb_vblank struct declared on the stack is not altered or zeroed before
being copied back to the user.  This patch takes care of it.

Signed-off-by: Dan Rosenberg <dan.j.rosenberg@gmail.com>
Cc: Thomas Winischhofer <thomas@winischhofer.net>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uml: fix compile warning

This fixes:
incompatible pointer type:  => 89
       arch/um/kernel/exec.c: warning: passing argument 2 of 'execve1' from
incompatible pointer type:  => 69, 85
       arch/um/kernel/exec.c: warning: passing argument 3 of 'execve1' from
incompatible pointer type:  => 69, 85

which was introduced by d7627467b7a8d ("Make do_execve() take a const
filename pointer")

Signed-off-by: Richard Weinberger <richard@nod.at>
Cc: David Howells <dhowells@redhat.com>
Cc: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

/proc/pid/smaps: fix dirty pages accounting

Currently, /proc/<pid>/smaps has wrong dirty pages accounting.
Shared_Dirty and Private_Dirty output only pte dirty pages and ignore
PG_dirty page flag.  It is difference against documentation, but also
inconsistent against Referenced field.  (Referenced checks both pte and
page flags)

This patch fixes it.

Test program:

 large-array.c
 ---------------------------------------------------
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 #include <unistd.h>

 char array[1*1024*1024*1024L];

 int main(void)
 {
         memset(array, 1, sizeof(array));
         pause();

         return 0;
 }
 ---------------------------------------------------

Test case:
 1. run ./large-array
 2. cat /proc/`pidof large-array`/smaps
 3. swapoff -a
 4. cat /proc/`pidof large-array`/smaps again

Test result:
 <before patch>

00601000-40601000 rw-p 00000000 00:00 0
Size:            1048576 kB
Rss:             1048576 kB
Pss:             1048576 kB
Shared_Clean:          0 kB
Shared_Dirty:          0 kB
Private_Clean:    218992 kB   <-- showed pages as clean incorrectly
Private_Dirty:    829584 kB
Referenced:       388364 kB
Swap:                  0 kB
KernelPageSize:        4 kB
MMUPageSize:           4 kB

 <after patch>

00601000-40601000 rw-p 00000000 00:00 0
Size:            1048576 kB
Rss:             1048576 kB
Pss:             1048576 kB
Shared_Clean:          0 kB
Shared_Dirty:          0 kB
Private_Clean:         0 kB
Private_Dirty:   1048576 kB  <-- fixed
Referenced:       388480 kB
Swap:                  0 kB
KernelPageSize:        4 kB
MMUPageSize:           4 kB

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fbcon: fix lockdep warning from fbcon_deinit()

Fix the lockdep warning:

[   13.657164] INFO: trying to register non-static key.
[   13.657169] the code is fine but needs lockdep annotation.
[   13.657171] turning off the locking correctness validator.
[   13.657177] Pid: 622, comm: modprobe Not tainted 2.6.36-rc3c #8
[   13.657180] Call Trace:
[   13.657194]  [<c13002c8>] ? printk+0x18/0x20
[   13.657202]  [<c1056cf6>] register_lock_class+0x336/0x350
[   13.657208]  [<c1058bf9>] __lock_acquire+0x449/0x1180
[   13.657215]  [<c1059997>] lock_acquire+0x67/0x80
[   13.657222]  [<c1042bf1>] ? __cancel_work_timer+0x51/0x230
[   13.657227]  [<c1042c23>] __cancel_work_timer+0x83/0x230
[   13.657231]  [<c1042bf1>] ? __cancel_work_timer+0x51/0x230
[   13.657236]  [<c10582b2>] ? mark_held_locks+0x62/0x80
[   13.657243]  [<c10b3a2f>] ? kfree+0x7f/0xe0
[   13.657248]  [<c105853c>] ? trace_hardirqs_on_caller+0x11c/0x160
[   13.657253]  [<c105858b>] ? trace_hardirqs_on+0xb/0x10
[   13.657259]  [<c117f4cd>] ? fbcon_deinit+0x16d/0x1e0
[   13.657263]  [<c117f4cd>] ? fbcon_deinit+0x16d/0x1e0
[   13.657268]  [<c1042dea>] cancel_work_sync+0xa/0x10
[   13.657272]  [<c117f444>] fbcon_deinit+0xe4/0x1e0
...

The warning is caused by trying to cancel an uninitialized work from
fbcon_exit().  Fix it by adding a check for queue.func, similarly to other
places in this code.

Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

efifb: support the EFI framebuffer on more Apple hardware

Enable the EFI framebuffer on 14 more Macs, including the iMac11,1
iMac10,1 iMac8,1 Macmini3,1 Macmini4,1 MacBook5,1 MacBook6,1 MacBook7,1
MacBookPro2,2 MacBookPro5,2 MacBookPro5,3 MacBookPro6,1 MacBookPro6,2 and
MacBookPro7,1

Information gathered from various user submissions.

    https://bugzilla.redhat.com/show_bug.cgi?id=528232
    http://ubuntuforums.org/showthread.php?t=1557326

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Luke Macken <lmacken@redhat.com>
Signed-off-by: Peter Jones <pjones@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

efifb: check that the base address is plausible on pci systems

Some Apple machines have identical DMI data but different memory
configurations for the video.  Given that, check that the address in our
table is actually within the range of a PCI BAR on a VGA device in the
machine.

This also fixes up the return value from set_system(), which has always
been wrong, but never resulted in bad behavior since there's only ever
been one matching entry in the dmi table.

The patch

1) stops people's machines from crashing when we get their display wrong,
   which seems to be unfortunately inevitable,

2) allows us to support identical dmi data with differing video memory
   configurations

This also adds me as the efifb maintainer, since I've effectively been
acting as such for quite some time.

Signed-off-by: Peter Jones <pjones@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

aio: do not return ERESTARTSYS as a result of AIO

OCFS2 can return ERESTARTSYS from its write function when the process is
signalled while waiting for a cluster lock (and the filesystem is mounted
with intr mount option).  Generally, it seems reasonable to allow
filesystems to return this error code from its IO functions.  As we must
not leak ERESTARTSYS (and similar error codes) to userspace as a result of
an AIO operation, we have to properly convert it to EINTR inside AIO code
(restarting the syscall isn't really an option because other AIO could
have been already submitted by the same io_submit syscall).

Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

vmscan: check all_unreclaimable in direct reclaim path

M.  Vefa Bicakci reported 2.6.35 kernel hang up when hibernation on his
32bit 3GB mem machine.
(https://bugzilla.kernel.org/show_bug.cgi?id=16771). Also he bisected
the regression to

  commit bb21c7ce18eff8e6e7877ca1d06c6db719376e3c
  Author: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
  Date:   Fri Jun 4 14:15:05 2010 -0700

     vmscan: fix do_try_to_free_pages() return value when priority==0 reclaim failure

At first impression, this seemed very strange because the above commit
only chenged function return value and hibernate_preallocate_memory()
ignore return value of shrink_all_memory().  But it's related.

Now, page allocation from hibernation code may enter infinite loop if the
system has highmem.  The reasons are that vmscan don't care enough OOM
case when oom_killer_disabled.

The problem sequence is following as.

1. hibernation
2. oom_disable
3. alloc_pages
4. do_try_to_free_pages
       if (scanning_global_lru(sc) && !all_unreclaimable)
               return 1;

If kswapd is not freozen, it would set zone->all_unreclaimable to 1 and
then shrink_zones maybe return true(ie, all_unreclaimable is true).  So at
last, alloc_pages could go to _nopage_.  If it is, it should have no
problem.

This patch adds all_unreclaimable check to protect in direct reclaim path,
too.  It can care of hibernation OOM case and help bailout
all_unreclaimable case slightly.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Reported-by: M. Vefa Bicakci <bicave@superonline.com>
Reported-by: <caiqian@redhat.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Tested-by: <caiqian@redhat.com>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/rtc/rtc-ab3100.c: add missing platform_set_drvdata() in ab3100_rtc_probe()

Otherwise, calling platform_get_drvdata() in ab3100_rtc_remove() returns
NULL.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked-by:Wan ZongShun <mcuos.com@gmail.com>
Acked-by: Linus Walleij <linus.walleij@stericsson.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MAINTAINERS: change AVR32 and AT32AP maintainer

Alter the maintainer of the AVR32 architecture and the AVR32/AT32AP
machine support to me.  Haavard is moving on to new challenges, and we've
found it better to transfer the maintainer part to me.  I will have good
contact with Haavard anyway.

Signed-off-by: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>
Acked-by: Haavard Skinnemoen <haavard.skinnemoen@atmel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

vmware balloon: rename module

In an effort to minimize customer confusion we want to unify naming
convention for VMware-provided kernel modules.  This change renames the
balloon driver from vmware_ballon to vmw_balloon.

We expect to follow this naming convention (vmw_<module_name>) for all
modules that are part of mainline kernel and/or being distributed by
VMware, with the sole exception of vmxnet3 driver (since the name of
mainline driver happens to match with the name used in VMware Tools).

Signed-off-by: Dmitry Torokhov <dtor@vmware.com>
Acked-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm: fix "arm: fix pci_set_consistent_dma_mask for dmabounce devices"

This fixes the regression caused by the commit 6fee48cd330c68
("dma-mapping: arm: use generic pci_set_dma_mask and
pci_set_consistent_dma_mask").

ARM needs to clip the dma coherent mask for dmabounce devices. This
restores the old trick.

Note that strictly speaking, the DMA API doesn't allow architectures to do
such but I'm not sure it's worth adding the new API to set the dma mask
that allows architectures to clip it.

Reported-by: Krzysztof Halasa <khc@pm.waw.pl>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

/proc/vmcore: fix seeking

Commit 73296bc611 ("procfs: Use generic_file_llseek in /proc/vmcore")
broke seeking on /proc/vmcore.  This changes it back to use default_llseek
in order to restore the original behaviour.

The problem with generic_file_llseek is that it only allows seeks up to
inode->i_sb->s_maxbytes, which is zero on procfs and some other virtual
file systems.  We should merge generic_file_llseek and default_llseek some
day and clean this up in a proper way, but for 2.6.35/36, reverting vmcore
is the safer solution.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Reported-by: CAI Qian <caiqian@redhat.com>
Tested-by: CAI Qian <caiqian@redhat.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ipmi: fix acpi probe print

After d9e1b6c45059ccf ("ipmi: fix ACPI detection with regspacing") we get

[   11.026326] ipmi_si: probing via ACPI
[   11.030019] ipmi_si 00:09: (null) regsize 1 spacing 1 irq 0
[   11.035594] ipmi_si: Adding ACPI-specified kcs state machine

on an old system with only one range for ipmi kcs range.

Try to fix it by adding another res pointer.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Corey Minyard <cminyard@mvista.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oom: always return a badness score of non-zero for eligible tasks

A task's badness score is roughly a proportion of its rss and swap
compared to the system's capacity.  The scale ranges from 0 to 1000 with
the highest score chosen for kill.  Thus, this scale operates on a
resolution of 0.1% of RAM + swap.  Admin tasks are also given a 3% bonus,
so the badness score of an admin task using 3% of memory, for example,
would still be 0.

It's possible that an exceptionally large number of tasks will combine to
exhaust all resources but never have a single task that uses more than
0.1% of RAM and swap (or 3.0% for admin tasks).

This patch ensures that the badness score of any eligible task is never 0
so the machine doesn't unnecessarily panic because it cannot find a task
to kill.

Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Nitin Gupta <ngupta@vflare.org>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Prevent freeing uninitialized pointer in compat_do_readv_writev

In 32-bit compatibility mode, the error handling for
compat_do_readv_writev() may free an uninitialized pointer, potentially
leading to all sorts of ugly memory corruption.  This is reliably
triggerable by unprivileged users by invoking the readv()/writev()
syscalls with an invalid iovec pointer.  The below patch fixes this to
emulate the non-compat version.

Introduced by commit b83733639a49 ("compat: factor out
compat_rw_copy_check_uvector from compat_do_readv_writev")

Signed-off-by: Dan Rosenberg <dan.j.rosenberg@gmail.com>
Cc: stable@kernel.org (2.6.35)
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge git://git./linux/kernel/git/davem/sparc-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  sparc: Prevent no-handler signal syscall restart recursion.
  sparc: Don't mask signal when we can't setup signal frame.
  sparc64: Fix race in signal instruction flushing.
  sparc64: Support RAW perf events.

powerpc: fix double syscall restarts

Make sigreturn zero regs->trap, make do_signal() do the same on all
paths.  As it is, signal interrupting e.g. read() from fd 512 (==
ERESTARTSYS) with another signal getting unblocked when the first
handler finishes will lead to restart one insn earlier than it ought
to.  Same for multiple signals with in-kernel handlers interrupting
that sucker at the same time.  Same for multiple signals of any kind
interrupting that sucker on 64bit...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  bdi: Fix warnings in __mark_inode_dirty for /dev/zero and friends
  char: Mark /dev/zero and /dev/kmem as not capable of writeback
  bdi: Initialize noop_backing_dev_info properly
  cfq-iosched: fix a kernel OOPs when usb key is inserted
  block: fix blk_rq_map_kern bio direction flag
  cciss: freeing uninitialized data on error path

bdi: Fix warnings in __mark_inode_dirty for /dev/zero and friends

Inodes of devices such as /dev/zero can get dirty for example via
utime(2) syscall or due to atime update. Backing device of such inodes
(zero_bdi, etc.) is however unable to handle dirty inodes and thus
__mark_inode_dirty complains.  In fact, inode should be rather dirtied
against backing device of the filesystem holding it. This is generally a
good rule except for filesystems such as 'bdev' or 'mtd_inodefs'. Inodes
in these pseudofilesystems are referenced from ordinary filesystem
inodes and carry mapping with real data of the device. Thus for these
inodes we have to use inode->i_mapping->backing_dev_info as we did so
far. We distinguish these filesystems by checking whether sb->s_bdi
points to a non-trivial backing device or not.

Example: Assume we have an ext3 filesystem on /dev/sda1 mounted on /.
There's a device inode A described by a path "/dev/sdb" on this
filesystem. This inode will be dirtied against backing device "8:0"
after this patch. bdev filesystem contains block device inode B coupled
with our inode A. When someone modifies a page of /dev/sdb, it's B that
gets dirtied and the dirtying happens against the backing device "8:16".
Thus both inodes get filed to a correct bdi list.

Cc: stable@kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

char: Mark /dev/zero and /dev/kmem as not capable of writeback

These devices don't do any writeback but their device inodes still can get
dirty so mark bdi appropriately so that bdi code does the right thing and files
inodes to lists of bdi carrying the device inodes.

Cc: stable@kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

bdi: Initialize noop_backing_dev_info properly

Properly initialize this backing dev info so that writeback code does not
barf when getting to it e.g. via sb->s_bdi.

Cc: stable@kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

sparc: Prevent no-handler signal syscall restart recursion.

Explicitly clear the "in-syscall" bit when we have no signal
handler and back up the program counters to back up the system
call.

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

sparc: Don't mask signal when we can't setup signal frame.

Don't invoke the signal handler tracehook in that situation
either.

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'for-linus/i2c/2636-rc5' of git://git.fluff.org/bjdooks/linux

* 'for-linus/i2c/2636-rc5' of git://git.fluff.org/bjdooks/linux:
  i2c-omap: Make sure i2c bus is free before setting it to idle

fs: {lock,unlock}_flocks() stubs to prepare for BKL removal

The lock structs are currently protected by the BKL, but are accessed by
code in fs/locks.c and misc file system and DLM code.  These stubs will
allow all users to switch to the new interface before the implementation
is changed to a spinlock.

Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Sage Weil <sage@newdream.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

i2c-omap: Make sure i2c bus is free before setting it to idle

If the i2c bus receives an interrupt with both BB (bus busy) and
ARDY (register access ready) statuses set during the tranfer of the last message
the bus was put to idle while still busy.

This caused bus to timeout.

Signed-off-by: Mathias Nyman <mathias.nyman@nokia.com>
Acked-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>

Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched: Fix nohz balance kick
  sched: Fix user time incorrectly accounted as system time on 32-bit

Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  hw breakpoints: Fix pid namespace bug
  x86: Fix instruction breakpoint encoding
  oprofile: Add Support for Intel CPU Family 6 / Model 22 (Intel Celeron 540)
  kprobes: Fix Kconfig dependency

Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  ceph: select CRYPTO
  ceph: check mapping to determine if FILE_CACHE cap is used
  ceph: only send one flushsnap per cap_snap per mds session
  ceph: fix cap_snap and realm split
  ceph: stop sending FLUSHSNAPs when we hit a dirty capsnap
  ceph: correctly set 'follows' in flushsnap messages
  ceph: fix dn offset during readdir_prepopulate
  ceph: fix file offset wrapping at 4GB on 32-bit archs
  ceph: fix reconnect encoding for old servers
  ceph: fix pagelist kunmap tail
  ceph: fix null pointer deref on anon root dentry release

Merge branch 'drm-intel-fixes' of git://git./linux/kernel/git/ickle/drm-intel

* 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ickle/drm-intel:
  drm/i915: Hold a reference to the object whilst unbinding the eviction list
  drm/i915,agp/intel: Add second set of PCI-IDs for B43
  drm/i915: Fix Sandybridge fence registers
  drm/i915/crt: Downgrade warnings for hotplug failures
  drm/i915: Ensure that the crtcinfo is populated during mode_fixup()

Merge git://git./linux/kernel/git/rusty/linux-2.6-for-linus

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  lguest: update comments to reflect LHCALL_LOAD_GDT_ENTRY.
  virtio: console: Prevent userspace from submitting NULL buffers
  virtio: console: Fix poll blocking even though there is data to read

sched: Fix nohz balance kick

There's a situation where the nohz balancer will try to wake itself:

cpu-x is idle which is also ilb_cpu
got a scheduler tick during idle
and the nohz_kick_needed() in trigger_load_balance() checks for
rq_x->nr_running which might not be zero (because of someone waking a
task on this rq etc) and this leads to the situation of the cpu-x
sending a kick to itself.

And this can cause a lockup.

Avoid this by not marking ourself eligible for kicking.

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1284400941.2684.19.camel@sbsiddha-MOBL3.sc.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

cfq-iosched: fix a kernel OOPs when usb key is inserted

Mike reported a kernel crash when a usb key hotplug is performed while all
kernel thrads are not in a root cgroup and are running in one of the child
cgroups of blkio controller.

BUG: unable to handle kernel NULL pointer dereference at 0000002c
IP: [<c11c7b08>] cfq_get_queue+0x232/0x412
*pde = 00000000
Oops: 0000 [#1] PREEMPT
last sysfs file: /sys/devices/pci0000:00/0000:00:1d.7/usb2/2-1/2-1:1.0/host3/scsi_host/host3/uevent

[..]
Pid: 30039, comm: scsi_scan_3 Not tainted 2.6.35.2-fg.roam #1 Volvi2                         /Aspire 4315
EIP: 0060:[<c11c7b08>] EFLAGS: 00010086 CPU: 0
EIP is at cfq_get_queue+0x232/0x412
EAX: f705f9c0 EBX: e977abac ECX: 00000000 EDX: 00000000
ESI: f00da400 EDI: f00da4ec EBP: e977a800 ESP: dff8fd00
 DS: 007b ES: 007b FS: 0000 GS: 0000 SS: 0068
Process scsi_scan_3 (pid: 30039, ti=dff8e000 task=f6b6c9a0 task.ti=dff8e000)
Stack:
 00000000 00000000 00000001 01ff0000 f00da508 00000000 f00da524 f00da540
<0> e7994940 dd631750 f705f9c0 e977a820 e977ac44 f00da4d0 00000001 f6b6c9a0
<0> 00000010 00008010 0000000b 00000000 00000001 e977a800 dd76fac0 00000246
Call Trace:
 [<c11c7f10>] ? cfq_set_request+0x228/0x34c
 [<c11c7ce8>] ? cfq_set_request+0x0/0x34c
 [<c11bb3b9>] ? elv_set_request+0xf/0x1c
 [<c11bdd51>] ? get_request+0x1ad/0x22f
 [<c11bddf2>] ? get_request_wait+0x1f/0x11a
 [<c11d013b>] ? kvasprintf+0x33/0x3b
 [<c127b537>] ? scsi_execute+0x1d/0x103
 [<c127b675>] ? scsi_execute_req+0x58/0x83
 [<c127c391>] ? scsi_probe_and_add_lun+0x188/0x7c2
 [<c12718c6>] ? attribute_container_add_device+0x15/0xfa
 [<c11c95d1>] ? kobject_get+0xf/0x13
 [<c126d1db>] ? get_device+0x10/0x14
 [<c127be93>] ? scsi_alloc_target+0x217/0x24d
 [<c127cbd8>] ? __scsi_scan_target+0x95/0x480
 [<c10204eb>] ? dequeue_entity+0x14/0x1fe
 [<c1020491>] ? update_curr+0x165/0x1ab
 [<c1020491>] ? update_curr+0x165/0x1ab
 [<c127d00d>] ? scsi_scan_channel+0x4a/0x76
 [<c127d0b0>] ? scsi_scan_host_selected+0x77/0xad
 [<c127d13c>] ? do_scan_async+0x0/0x11a
 [<c127d137>] ? do_scsi_scan_host+0x51/0x56
 [<c127d13c>] ? do_scan_async+0x0/0x11a
 [<c127d14a>] ? do_scan_async+0xe/0x11a
 [<c127d13c>] ? do_scan_async+0x0/0x11a
 [<c10354c5>] ? kthread+0x5e/0x63
 [<c1035467>] ? kthread+0x0/0x63
 [<c1002af6>] ? kernel_thread_helper+0x6/0x10
Code: 44 24 1c 54 83 44 24 18 54 83 fa 03 75 94 8b 06 c7 86 64 02 00 00 01 00 00 00 83 e0 03 09 f0 89 06 8b 44 24 28 8b 90 58 01 00 00 <8b> 42 2c 85 c0 75 03 8b 42 08 8d 54 24 48 52 8d 4c 24 50 51 68
EIP: [<c11c7b08>] cfq_get_queue+0x232/0x412 SS:ESP 0068:dff8fd00
CR2: 000000000000002c
---[ end trace 9a88306573f69b12 ]---

The problem here is that we don't have bdi->dev information available when
thread does some IO.  Hence when dev_name() tries to access bdi->dev, it
crashes.

This problem does not happen if kernel threads are in root group as root
group is statically allocated at device initialization time and we don't
hit this piece of code.

Fix it by delaying the filling of major and minor number information of
device in blk_group.  Initially a blk_group is created with 0 as device
information and this information is filled later once some more IO comes
in from same group.

Reported-by: Mike Kazantsev <mk.fraggod@gmail.com>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

block: fix blk_rq_map_kern bio direction flag

This bug was introduced in 7b6d91daee5cac6402186ff224c3af39d79f4a0e
"block: unify flags for struct bio and struct request"

Cc: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

cciss: freeing uninitialized data on error path

The "h->scatter_list" is allocated inside a for loop.  If any of those
allocations fail, then the rest of the list is uninitialized data.  When
we free it we should start from the top and free backwards so that we
don't call kfree() on uninitialized pointers.

Also if the allocation for "h->scatter_list" fails then we would get an
Oops here.  I should have noticed this when I send: 4ee69851c "cciss:
handle allocation failure."  but I didn't.  Sorry about that.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

Merge remote branch 'linus' into drm-intel-fixes

sparc64: Fix race in signal instruction flushing.

If another cpu does a very wide munmap() on the signal frame area,
it can tear down the page table hierarchy from underneath us.

Borrow an idea from the 64-bit fault path's get_user_insn(), and
disable cross call interrupts during the page table traversal
to lock them in place while we operate.

Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>

percpu: fix pcpu_last_unit_cpu

pcpu_first/last_unit_cpu are used to track which cpu has the first and
last units assigned.  This in turn is used to determine the span of a
chunk for man/unmap cache flushes and whether an address belongs to
the first chunk or not in per_cpu_ptr_to_phys().

When the number of possible CPUs isn't power of two, a chunk may
contain unassigned units towards the end of a chunk.  The logic to
determine pcpu_last_unit_cpu was incorrect when there was an unused
unit at the end of a chunk.  It failed to ignore the unused unit and
assigned the unused marker NR_CPUS to pcpu_last_unit_cpu.

This was discovered through kdump failure which was caused by
malfunctioning per_cpu_ptr_to_phys() on a kvm setup with 50 possible
CPUs by CAI Qian.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: CAI Qian <caiqian@redhat.com>
Cc: stable@kernel.org

lguest: update comments to reflect LHCALL_LOAD_GDT_ENTRY.

We used to have a hypercall which reloaded the entire GDT, then we
switched to one which loaded a single entry (to match the IDT code).

Some comments were not updated, so fix them.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Reported by: Eviatar Khen <eviatarkhen@gmail.com>

virtio: console: Prevent userspace from submitting NULL buffers

A userspace could submit a buffer with 0 length to be written to the
host.  Prevent such a situation.

This was not needed previously, but recent changes in the way write()
works exposed this condition to trigger a virtqueue event to the host,
causing a NULL buffer to be sent across.

Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
CC: stable@kernel.org

virtio: console: Fix poll blocking even though there is data to read

I found this while working on a Linux agent for spice, the symptom I was
seeing was select blocking on the spice vdagent virtio serial port even
though there were messages queued up there.

virtio_console's port_fops_poll checks port->inbuf != NULL to determine
if read won't block. However if an application reads enough bytes from
inbuf through port_fops_read, to empty the current port->inbuf,
port->inbuf will be NULL even though there may be buffers left in the
virtqueue.

This causes poll() to block even though there is data to be read,
this patch fixes this by using will_read_block(port) instead of the
port->inbuf != NULL check.

Signed-off-By: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: stable@kernel.org

Linux 2.6.36-rc5

Merge git://git./linux/kernel/git/gregkh/staging-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging-2.6:
  Staging: vt6655: fix buffer overflow
  Revert: "Staging: batman-adv: Adding netfilter-bridge hooks"

Merge git://git./linux/kernel/git/gregkh/usb-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6:
  USB: musb: MAINTAINERS: Fix my mail address
  USB: serial/mos*: prevent reading uninitialized stack memory
  USB: otg: twl4030: fix phy initialization(v1)
  USB: EHCI: Disable langwell/penwell LPM capability
  usb: musb_debugfs: don't use the struct file private_data field with seq_files

Merge git://git./linux/kernel/git/gregkh/tty-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty-2.6:
  serial: mfd: fix bug in serial_hsu_remove()
  serial: amba-pl010: fix set_ldisc

Staging: vt6655: fix buffer overflow

"param->u.wpa_associate.wpa_ie_len" comes from the user.  We should
check it so that the copy_from_user() doesn't overflow the buffer.

Also further down in the function, we assume that if
"param->u.wpa_associate.wpa_ie_len" is set then "abyWPAIE[0]" is
initialized.  To make that work, I changed the test here to say that if
"wpa_ie_len" is set then "wpa_ie" has to be a valid pointer or we return
-EINVAL.

Oddly, we only use the first element of the abyWPAIE[] array.  So I
suspect there may be some other issues in this function.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

Revert: "Staging: batman-adv: Adding netfilter-bridge hooks"

This reverts commit 96d592ed599434d2d5f339a1d282871bc6377d2c.

The netfilter hook seems to be misused and may leak skbs in situations
when NF_HOOK returns NF_STOLEN. It may not filter everything as
expected. Also the ethernet bridge tables are not yet capable to
understand batman-adv packet correctly.

It was only added for testing purposes and can be removed again.

Reported-by: Vasiliy Kulikov <segooon@gmail.com>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

serial: mfd: fix bug in serial_hsu_remove()

Medfield HSU driver deal with 4 pci devices(3 uart ports + 1 dma controller),
so in pci remove func, we need handle them differently

Signed-off-by: Feng Tang <feng.tang@intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

serial: amba-pl010: fix set_ldisc

Commit d87d9b7d1 ("tty: serial - fix tty referencing in set_ldisc") changed
set_ldisc to take ldisc number as parameter. This patch fixes AMBA PL010 driver
according the new prototype.

Signed-off-by: Mika Westerberg <mika.westerberg@iki.fi>
Cc: Alan Cox <alan@linux.intel.com>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

USB: musb: MAINTAINERS: Fix my mail address

If we don't, contributors to musb and any USB OMAP
code will be sending mails to an unexistent inbox.

Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

USB: serial/mos*: prevent reading uninitialized stack memory

The TIOCGICOUNT device ioctl in both mos7720.c and mos7840.c allows
unprivileged users to read uninitialized stack memory, because the
"reserved" member of the serial_icounter_struct struct declared on the
stack is not altered or zeroed before being copied back to the user.
This patch takes care of it.

Signed-off-by: Dan Rosenberg <dan.j.rosenberg@gmail.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

USB: otg: twl4030: fix phy initialization(v1)

Commit 461c317705eca5cac09a360f488715927fd0a927(into 2.6.36-v3)
is put forward to power down phy if no usb cable is connected,
but does introduce the two issues below:

1), phy is not into work state if usb cable is connected
with PC during poweron, so musb device mode is not usable
in such case, follows the reasons:
-twl4030_phy_resume is not called, so
regulators are not enabled
i2c access are not enabled
usb mode not configurated

2), The kernel warings[1] of regulators 'unbalanced disables'
is caused if poweron without usb cable connected
with PC or b-device.

This patch fixes the two issues above:
-power down phy only if no usb cable is connected with PC
and b-device
-do phy initialization(via __twl4030_phy_resume) if usb cable
is connected with PC(vbus event) or another b-device(ID event) in
twl4030_usb_probe.

This patch also doesn't put VUSB3V1 LDO into active mode in
twl4030_usb_ldo_init until VBUS/ID change detected, so we can
save more power consumption than before.

This patch is verified OK on Beagle board either connected with
usb cable or not when poweron.

[1]. warnings of 'unbalanced disables' of regulators.
[root@OMAP3EVM /]# dmesg
------------[ cut here ]------------
WARNING: at drivers/regulator/core.c:1357 _regulator_disable+0x38/0x128()
unbalanced disables for VUSB1V8
Modules linked in:
Backtrace:
[<c0030c48>] (dump_backtrace+0x0/0x110) from [<c034f5a8>] (dump_stack+0x18/0x1c)
 r7:c78179d8 r6:c01ed6b8 r5:c0410822 r4:0000054d
[<c034f590>] (dump_stack+0x0/0x1c) from [<c0057da8>] (warn_slowpath_common+0x54/0x6c)
[<c0057d54>] (warn_slowpath_common+0x0/0x6c) from [<c0057e64>] (warn_slowpath_fmt+0x38/0x40)
 r9:00000000 r8:00000000 r7:c78e6608 r6:00000000 r5:fffffffb
 r4:c78e6c00
[<c0057e2c>] (warn_slowpath_fmt+0x0/0x40) from [<c01ed6b8>] (_regulator_disable+0x38/0x128)
 r3:c0410e53 r2:c0410ad5
[<c01ed680>] (_regulator_disable+0x0/0x128) from [<c01ed87c>] (regulator_disable+0x24/0x38)
 r7:c78e6608 r6:00000000 r5:c78e6c40 r4:c78e6c00
[<c01ed858>] (regulator_disable+0x0/0x38) from [<c02382dc>] (twl4030_phy_power+0x15c/0x17c)
 r5:c78595c0 r4:00000000
[<c0238180>] (twl4030_phy_power+0x0/0x17c) from [<c023831c>] (twl4030_phy_suspend+0x20/0x2c)
 r6:00000000 r5:c78595c0 r4:c78595c0
[<c02382fc>] (twl4030_phy_suspend+0x0/0x2c) from [<c0238638>] (twl4030_usb_irq+0x11c/0x16c)
 r5:c78595c0 r4:00000040
[<c023851c>] (twl4030_usb_irq+0x0/0x16c) from [<c034ec18>] (twl4030_usb_probe+0x2c4/0x32c)
 r6:00000000 r5:00000000 r4:c78595c0
[<c034e954>] (twl4030_usb_probe+0x0/0x32c) from [<c02152a0>] (platform_drv_probe+0x20/0x24)
 r7:00000000 r6:c047d49c r5:c78e6608 r4:c047d49c
[<c0215280>] (platform_drv_probe+0x0/0x24) from [<c0214244>] (driver_probe_device+0xd0/0x190)
[<c0214174>] (driver_probe_device+0x0/0x190) from [<c02143d4>] (__device_attach+0x44/0x48)
 r7:00000000 r6:c78e6608 r5:c78e6608 r4:c047d49c
[<c0214390>] (__device_attach+0x0/0x48) from [<c0213694>] (bus_for_each_drv+0x50/0x90)
 r5:c0214390 r4:00000000
[<c0213644>] (bus_for_each_drv+0x0/0x90) from [<c0214474>] (device_attach+0x70/0x94)
 r6:c78e663c r5:c78e6608 r4:c78e6608
[<c0214404>] (device_attach+0x0/0x94) from [<c02134fc>] (bus_probe_device+0x2c/0x48)
 r7:00000000 r6:00000002 r5:c78e6608 r4:c78e6600
[<c02134d0>] (bus_probe_device+0x0/0x48) from [<c0211e48>] (device_add+0x340/0x4b4)
[<c0211b08>] (device_add+0x0/0x4b4) from [<c021597c>] (platform_device_add+0x110/0x16c)
[<c021586c>] (platform_device_add+0x0/0x16c) from [<c0220cb0>] (add_numbered_child+0xd8/0x118)
 r7:00000000 r6:c045f15c r5:c78e6600 r4:00000000
[<c0220bd8>] (add_numbered_child+0x0/0x118) from [<c001c618>] (twl_probe+0x3a4/0x72c)
[<c001c274>] (twl_probe+0x0/0x72c) from [<c02601ac>] (i2c_device_probe+0x7c/0xa4)
[<c0260130>] (i2c_device_probe+0x0/0xa4) from [<c0214244>] (driver_probe_device+0xd0/0x190)
 r5:c7856e20 r4:c047c860
[<c0214174>] (driver_probe_device+0x0/0x190) from [<c02143d4>] (__device_attach+0x44/0x48)
 r7:c7856e04 r6:c7856e20 r5:c7856e20 r4:c047c860
[<c0214390>] (__device_attach+0x0/0x48) from [<c0213694>] (bus_for_each_drv+0x50/0x90)
 r5:c0214390 r4:00000000
[<c0213644>] (bus_for_each_drv+0x0/0x90) from [<c0214474>] (device_attach+0x70/0x94)
 r6:c7856e54 r5:c7856e20 r4:c7856e20
[<c0214404>] (device_attach+0x0/0x94) from [<c02134fc>] (bus_probe_device+0x2c/0x48)
 r7:c7856e04 r6:c78fd048 r5:c7856e20 r4:c7856e20
[<c02134d0>] (bus_probe_device+0x0/0x48) from [<c0211e48>] (device_add+0x340/0x4b4)
[<c0211b08>] (device_add+0x0/0x4b4) from [<c0211fd8>] (device_register+0x1c/0x20)
[<c0211fbc>] (device_register+0x0/0x20) from [<c0260aa8>] (i2c_new_device+0xec/0x150)
 r5:c7856e00 r4:c7856e20
[<c02609bc>] (i2c_new_device+0x0/0x150) from [<c0260dc0>] (i2c_register_adapter+0xa0/0x1c4)
 r7:00000000 r6:c78fd078 r5:c78fd048 r4:c781d5c0
[<c0260d20>] (i2c_register_adapter+0x0/0x1c4) from [<c0260f80>] (i2c_add_numbered_adapter+0x9c/0xb4)
 r7:00000a28 r6:c04600a8 r5:c78fd048 r4:00000000
[<c0260ee4>] (i2c_add_numbered_adapter+0x0/0xb4) from [<c034efa4>] (omap_i2c_probe+0x324/0x3e8)
 r5:00000000 r4:c78fd000
[<c034ec80>] (omap_i2c_probe+0x0/0x3e8) from [<c02152a0>] (platform_drv_probe+0x20/0x24)
[<c0215280>] (platform_drv_probe+0x0/0x24) from [<c0214244>] (driver_probe_device+0xd0/0x190)
[<c0214174>] (driver_probe_device+0x0/0x190) from [<c021436c>] (__driver_attach+0x68/0x8c)
 r7:c78b2140 r6:c047e214 r5:c04600e4 r4:c04600b0
[<c0214304>] (__driver_attach+0x0/0x8c) from [<c021399c>] (bus_for_each_dev+0x50/0x84)
 r7:c78b2140 r6:c047e214 r5:c0214304 r4:00000000
[<c021394c>] (bus_for_each_dev+0x0/0x84) from [<c0214068>] (driver_attach+0x20/0x28)
 r6:c047e214 r5:c047e214 r4:c00270d0
[<c0214048>] (driver_attach+0x0/0x28) from [<c0213274>] (bus_add_driver+0xa8/0x228)
[<c02131cc>] (bus_add_driver+0x0/0x228) from [<c02146a4>] (driver_register+0xb0/0x13c)
[<c02145f4>] (driver_register+0x0/0x13c) from [<c0215744>] (platform_driver_register+0x4c/0x60)
 r9:00000000 r8:c001f688 r7:00000013 r6:c005b6fc r5:c00083dc
r4:c00270d0
[<c02156f8>] (platform_driver_register+0x0/0x60) from [<c001f69c>] (omap_i2c_init_driver+0x14/0x1c)
[<c001f688>] (omap_i2c_init_driver+0x0/0x1c) from [<c002c460>] (do_one_initcall+0xd0/0x1a4)
[<c002c390>] (do_one_initcall+0x0/0x1a4) from [<c0008478>] (kernel_init+0x9c/0x154)
[<c00083dc>] (kernel_init+0x0/0x154) from [<c005b6fc>] (do_exit+0x0/0x688)
 r5:c00083dc r4:00000000
---[ end trace 1b75b31a2719ed1d ]---

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Cc: Felipe Balbi <me@felipebalbi.com>
Cc: Anand Gadiyar <gadiyar@ti.com>
Cc: Mike Frysinger <vapier@gentoo.org>
Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

USB: EHCI: Disable langwell/penwell LPM capability

We have to do so due to HW limitation.

Signed-off-by: Alek Du <alek.du@intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

usb: musb_debugfs: don't use the struct file private_data field with seq_files

seq_files use the private_data field of a file struct for storing a seq_file structure,
data should be stored in seq_file's own private field (e.g. file->private_data->private)
Otherwise seq_release() will free the private data when the file is closed.

Signed-off-by: Mathias Nyman <mathias.nyman@nokia.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

drm/i915: Hold a reference to the object whilst unbinding the eviction list

During heavy aperture thrashing we may be forced to wait upon several active
objects during eviction. The active list may be the last reference to
these objects and so the action of waiting upon one of them may cause
another to be freed (and itself unbound). To prevent the object
disappearing underneath us, we need to acquire and hold a reference
whilst unbinding.

This should fix the reported page refcount OOPS:

kernel BUG at drivers/gpu/drm/i915/i915_gem.c:1444!
...
RIP: 0010:[<ffffffffa0093026>]  [<ffffffffa0093026>] i915_gem_object_put_pages+0x25/0xf5 [i915]
Call Trace:
 [<ffffffffa009481d>] i915_gem_object_unbind+0xc5/0x1a7 [i915]
 [<ffffffffa0098ab2>] i915_gem_evict_something+0x3bd/0x409 [i915]
 [<ffffffffa0027923>] ? drm_gem_object_lookup+0x27/0x57 [drm]
 [<ffffffffa0093bc3>] i915_gem_object_bind_to_gtt+0x1d3/0x279 [i915]
 [<ffffffffa0095b30>] i915_gem_object_pin+0xa3/0x146 [i915]
 [<ffffffffa0027948>] ? drm_gem_object_lookup+0x4c/0x57 [drm]
 [<ffffffffa00961bc>] i915_gem_do_execbuffer+0x50d/0xe32 [i915]

Reported-by: Shawn Starr <shawn.starr@rogers.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=18902
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

frv: double syscall restarts, syscall restart in sigreturn()

We need to make sure that only the first do_signal() to be handled on
the way out syscall will bother with syscall restarts; additionally, the
check on the "signal has user handler" path had been wrong - compare
with restart prevention in sigreturn()...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

frv: handling of restart into restart_syscall is fscked

do_signal() should place the syscall number in gr7, not gr8 when
handling ERESTART_WOULDBLOCK.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

frv: avoid infinite loop of SIGSEGV delivery

Use force_sigsegv() rather than force_sig(SIGSEGV, ...) as the former
resets the SEGV handler pointer which will kill the process, rather than
leaving it open to an infinite loop if the SEGV handler itself caused a
SEGV signal.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

frv: fix address verification holes in setup_frame/setup_rt_frame

a) sa_handler might be maliciously set to point to kernel memory;
   blindly dereferencing it in FDPIC case is a Bad Idea(tm).

b) I'm not sure you need that set_fs(USER_DS) there at all, but if you
   do, you'd better do it *before* checking the frame you've decided to
   use with access_ok(), lest sigaltstack() becomes a convenient
   roothole.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

frv: restart_block.fn needs to be reset on sigreturn

Reset restart_block.fn on executing a sigreturn such that any currently
pending system call restarts will be forced to return -EINTR.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: further fix swapin race condition

Commit 4969c1192d15 ("mm: fix swapin race condition") is now agreed to
be incomplete.  There's a race, not very much less likely than the
original race envisaged, in which it is further necessary to check that
the swapcache page's swap has not changed.

Here's the reasoning: cast in terms of reuse_swap_page(), but probably
could be reformulated to rely on try_to_free_swap() instead, or on
swapoff+swapon.

A, faults into do_swap_page(): does page1 = lookup_swap_cache(swap1) and
comes through the lock_page(page1).

B, a racing thread of the same process, faults on the same address: does
page1 = lookup_swap_cache(swap1) and now waits in lock_page(page1), but
for whatever reason is unlucky not to get the lock any time soon.

A carries on through do_swap_page(), a write fault, but cannot reuse the
swap page1 (another reference to swap1).  Unlocks the page1 (but B
doesn't get it yet), does COW in do_wp_page(), page2 now in that pte.

C, perhaps the parent of A+B, comes in and write faults the same swap
page1 into its mm, reuse_swap_page() succeeds this time, swap1 is freed.

kswapd comes in after some time (B still unlucky) and swaps out some
pages from A+B and C: it allocates the original swap1 to page2 in A+B,
and some other swap2 to the original page1 now in C.  But does not
immediately free page1 (actually it couldn't: B holds a reference),
leaving it in swap cache for now.

B at last gets the lock on page1, hooray! Is PageSwapCache(page1)? Yes.
Is pte_same(*page_table, orig_pte)? Yes, because page2 has now been
given the swap1 which page1 used to have.  So B proceeds to insert page1
into A+B's page_table, though its content now belongs to C, quite
different from what A wrote there.

B ought to have checked that page1's swap was still swap1.

Signed-off-by: Hugh Dickins <hughd@google.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>