7 years agomac80211: fix error in station state transitions during reconfig
Meenakshi Venkataraman [Wed, 30 May 2012 09:39:33 +0000]
mac80211: fix error in station state transitions during reconfig

commit bd34ab62a3297bd7685da11b0cbe05ae4cd8b02c upstream.

As part of hardware reconfig mac80211 tries
to restore the station state to its values
before the hardware reconfig, but it only
goes to the last-state - 1. Fix this
off-by-one error.

Signed-off-by: Meenakshi Venkataraman <meenakshi.venkataraman@intel.com>
Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiwlwifi: disable the buggy chain extension feature in HW
Emmanuel Grumbach [Wed, 6 Jun 2012 11:55:02 +0000]
iwlwifi: disable the buggy chain extension feature in HW

commit d012d04e4d6312ea157b6cf19e9689af934f5aa7 upstream.

This feature has been reported to be buggy and enabled by
default. We therefore need to disable it manually.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiwlwifi: don't mess up the SCD when removing a key
Emmanuel Grumbach [Wed, 6 Jun 2012 07:13:36 +0000]
iwlwifi: don't mess up the SCD when removing a key

commit d6ee27eb13beab94056e0de52d81220058ca2297 upstream.

When we remove a key, we put a key index which was supposed
to tell the fw that we are actually removing the key. But
instead the fw took that index as a valid index and messed
up the SRAM of the device.

This memory corruption on the device mangled the data of
the SCD. The impact on the user is that SCD queue 2 got
stuck after having removed keys.
The message is the log that was printed is:

Queue 2 stuck for 10000ms

This doesn't seem to fix the higher queues that get stuck
from time to time.

Reviewed-by: Meenakshi Venkataraman <meenakshi.venkataraman@intel.com>
Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiwlwifi: unregister LEDs if mac80211 registration fails
Johannes Berg [Tue, 5 Jun 2012 07:38:35 +0000]
iwlwifi: unregister LEDs if mac80211 registration fails

commit 0e1fa7ef25004b9c1a14147bce61c15c2f1c6744 upstream.

Otherwise the LEDs stick around and cause issues the
next time around since they're still there but not
really hooked up.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiwlwifi: disable WoWLAN if !CONFIG_PM_SLEEP
Johannes Berg [Mon, 4 Jun 2012 11:43:11 +0000]
iwlwifi: disable WoWLAN if !CONFIG_PM_SLEEP

commit fcb6ff5e2cb83e1de10631f6621f45ca3401bf61 upstream.

If CONFIG_PM_SLEEP is disabled, then iwlwifi doesn't
support suspend/resume handlers and thus mac80211
(correctly) refuses advertising WoWLAN. Disable
WoWLAN in the driver in this case.

Reported-by: Sebastian Kemper <sebastian_ml@gmx.net>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoBtrfs: fall back to non-inline if we don't have enough space
Josef Bacik [Wed, 23 May 2012 20:10:14 +0000]
Btrfs: fall back to non-inline if we don't have enough space

commit 2adcac1a7331d93a17285804819caa96070b231f upstream.

If cow_file_range_inline fails with ENOSPC we abort the transaction which
isn't very nice.  This really shouldn't be happening anyways but there's no
sense in making it a horrible error when we can easily just go allocate
normal data space for this stuff.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>
Acked-by: Chris Mason <chris.mason@fusionio.com>
Cc: Alexandre Oliva <oliva@lsd.ic.unicamp.br>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/ttm: Fix buffer object metadata accounting regression v2
Thomas Hellstrom [Tue, 12 Jun 2012 11:28:42 +0000]
drm/ttm: Fix buffer object metadata accounting regression v2

commit a393c730ab69617c3291a3b0b2a228c9be2fc28c upstream.

A regression was introduced in the 3.3 rc series, commit
"drm/ttm: simplify memory accounting for ttm user v2",
causing the metadata of buffer objects created using the ttm_bo_create()
function to be accounted twice.
That causes massive leaks with the vmwgfx driver running for example
SpecViewperf Catia-03 test 2, eventually killing the app.

Furthermore, the same commit introduces a regression where
metadata accounting is leaked if a buffer object is
initialized with an illegal size. This is also fixed with this commit.

v2: Fixed an error path and removed an unused variable.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agocrypto: aesni-intel - fix unaligned cbc decrypt for x86-32
Mathias Krause [Tue, 29 May 2012 23:43:08 +0000]
crypto: aesni-intel - fix unaligned cbc decrypt for x86-32

commit 7c8d51848a88aafdb68f42b6b650c83485ea2f84 upstream.

The 32 bit variant of cbc(aes) decrypt is using instructions requiring
128 bit aligned memory locations but fails to ensure this constraint in
the code. Fix this by loading the data into intermediate registers with
load unaligned instructions.

This fixes reported general protection faults related to aesni.

References: https://bugzilla.kernel.org/show_bug.cgi?id=43223
Reported-by: Daniel <garkein@mailueberfall.de>
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agohwrng: atmel-rng - fix race condition leading to repeated bits
Peter Korsgaard [Thu, 31 May 2012 10:53:08 +0000]
hwrng: atmel-rng - fix race condition leading to repeated bits

commit 121daad8fd1dce63076fa55aaedd5dc3f981b334 upstream.

Data valid gets cleared by reading the ISR (status register) and NOT from
reading ODATA (data register). A new data word can become available between
checking ISR and reading ODATA, causing us to reuse the same data word next
time atmel_trng_read() gets called, if that happens before the following
data word is ready.

With this fixed, rngtest no longer complains of 'Continous run' errors.
Before:

rngtest -c 1000 < /dev/hwrng
rngtest 3
Copyright (c) 2004 by Henrique de Moraes Holschuh
This is free software; see the source for copying conditions.  There is NO warr.

rngtest: starting FIPS tests...
rngtest: bits received from input: 20000032
rngtest: FIPS 140-2 successes: 923
rngtest: FIPS 140-2 failures: 77
rngtest: FIPS 140-2(2001-10-10) Monobit: 0
rngtest: FIPS 140-2(2001-10-10) Poker: 0
rngtest: FIPS 140-2(2001-10-10) Runs: 1
rngtest: FIPS 140-2(2001-10-10) Long run: 0
rngtest: FIPS 140-2(2001-10-10) Continuous run: 76
rngtest: input channel speed: (min=721.402; avg=46003.510; max=49321.338)Kibitss
rngtest: FIPS tests speed: (min=11.442; avg=12.714; max=12.801)Mibits/s
rngtest: Program run time: 1931860 microseconds

After:

rngtest -c 1000 < /dev/hwrng
rngtest 3
Copyright (c) 2004 by Henrique de Moraes Holschuh
This is free software; see the source for copying conditions.  There is NO warr.

rngtest: starting FIPS tests...
rngtest: bits received from input: 20000032
rngtest: FIPS 140-2 successes: 1000
rngtest: FIPS 140-2 failures: 0
rngtest: FIPS 140-2(2001-10-10) Monobit: 0
rngtest: FIPS 140-2(2001-10-10) Poker: 0
rngtest: FIPS 140-2(2001-10-10) Runs: 0
rngtest: FIPS 140-2(2001-10-10) Long run: 0
rngtest: FIPS 140-2(2001-10-10) Continuous run: 0
rngtest: input channel speed: (min=777.518; avg=36988.482; max=43115.342)Kibitss
rngtest: FIPS tests speed: (min=11.951; avg=12.715; max=12.887)Mibits/s
rngtest: Program run time: 2035543 microseconds

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Reported-by: George Pontis <GPontis@z9.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agox86, MCE, AMD: Make APIC LVT thresholding interrupt optional
Borislav Petkov [Mon, 16 Apr 2012 16:01:53 +0000]
x86, MCE, AMD: Make APIC LVT thresholding interrupt optional

commit f227d4306cf30e1d5b6f231e8ef9006c34f3d186 upstream.

Currently, the APIC LVT interrupt for error thresholding is implicitly
enabled. However, there are models in the F15h range which do not enable
it. Make the code machinery which sets up the APIC interrupt support
an optional setting and add an ->interrupt_capable member to the bank
representation mirroring that capability and enable the interrupt offset
programming only if it is true.

Simplify code and fixup comment style while at it.

Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Cc: Robert Richter <robert.richter@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agox86/uv: Fix UV2 BAU legacy mode
Cliff Wickman [Thu, 7 Jun 2012 13:31:40 +0000]
x86/uv: Fix UV2 BAU legacy mode

commit d5d2d2eea84b0d8450b082edbc3dbde41fb8bfd8 upstream.

The SGI Altix UV2 BAU (Broadcast Assist Unit) as used for
tlb-shootdown (selective broadcast mode) always uses UV2
broadcast descriptor format. There is no need to clear the
'legacy' (UV1) mode, because the hardware always uses UV2 mode
for selective broadcast.

But the BIOS uses general broadcast and legacy mode, and the
hardware pays attention to the legacy mode bit for general
broadcast. So the kernel must not clear that mode bit.

Signed-off-by: Cliff Wickman <cpw@sgi.com>
Link: http://lkml.kernel.org/r/E1SccoO-0002Lh-Cb@eag09.americas.sgi.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: fix the free blocks calculation for ext3 file systems w/ uninit_bg
Theodore Ts'o [Thu, 7 Jun 2012 22:56:06 +0000]
ext4: fix the free blocks calculation for ext3 file systems w/ uninit_bg

commit b0dd6b70f0fda17ae9762fbb72d98e40a4f66556 upstream.

Ext3 filesystems that are converted to use as many ext4 file system
features as possible will enable uninit_bg to speed up e2fsck times.
These file systems will have a native ext3 layout of inode tables and
block allocation bitmaps (as opposed to ext4's flex_bg layout).
Unfortunately, in these cases, when first allocating a block in an
uninitialized block group, ext4 would incorrectly calculate the number
of free blocks in that block group, and then errorneously report that
the file system was corrupt:

EXT4-fs error (device vdd): ext4_mb_generate_buddy:741: group 30, 32254 clusters in bitmap, 32258 in gd

This problem can be reproduced via:

    mke2fs -q -t ext4 -O ^flex_bg /dev/vdd 5g
    mount -t ext4 /dev/vdd /mnt
    fallocate -l 4600m /mnt/test

The problem was caused by a bone headed mistake in the check to see if a
particular metadata block was part of the block group.

Many thanks to Kees Cook for finding and bisecting the buggy commit
which introduced this bug (commit fd034a84e1, present since v3.2).

Reported-by: Sander Eikelenboom <linux@eikelenboom.it>
Reported-by: Kees Cook <keescook@chromium.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Tested-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agotimekeeping: Fix CLOCK_MONOTONIC inconsistency during leapsecond
John Stultz [Wed, 30 May 2012 17:54:57 +0000]
timekeeping: Fix CLOCK_MONOTONIC inconsistency during leapsecond

commit fad0c66c4bb836d57a5f125ecd38bed653ca863a upstream.

Commit 6b43ae8a61 (ntp: Fix leap-second hrtimer livelock) broke the
leapsecond update of CLOCK_MONOTONIC. The missing leapsecond update to
wall_to_monotonic causes discontinuities in CLOCK_MONOTONIC.

Adjust wall_to_monotonic when NTP inserted a leapsecond.

Reported-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Tested-by: Richard Cochran <richardcochran@gmail.com>
Link: http://lkml.kernel.org/r/1338400497-12420-1-git-send-email-john.stultz@linaro.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomodule_param: stop double-calling parameters.
Rusty Russell [Fri, 8 Jun 2012 05:28:13 +0000]
module_param: stop double-calling parameters.

commit ae82fdb1406ad41d68f07027fe31f2d35ba22a90 upstream.

Commit 026cee0086fe1df4cf74691cf273062cc769617d "params:
<level>_initcall-like kernel parameters" set old-style module
parameters to level 0.  And we call those level 0 calls where we used
to, early in start_kernel().

We also loop through the initcall levels and call the levelled
module_params before the corresponding initcall.  Unfortunately level
0 is early_init(), so we call the standard module_param calls twice.

(Turns out most things don't care, but at least ubi.mtd does).

Change the level to -1 for standard module_param calls.

Reported-by: Benoît Thébaudeau <benoit.thebaudeau@advansee.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agopowerpc: Fix kernel panic during kernel module load
Steffen Rumler [Wed, 6 Jun 2012 14:37:17 +0000]
powerpc: Fix kernel panic during kernel module load

commit 3c75296562f43e6fbc6cddd3de948a7b3e4e9bcf upstream.

This fixes a problem which can causes kernel oopses while loading
a kernel module.

According to the PowerPC EABI specification, GPR r11 is assigned
the dedicated function to point to the previous stack frame.
In the powerpc-specific kernel module loader, do_plt_call()
(in arch/powerpc/kernel/module_32.c), GPR r11 is also used
to generate trampoline code.

This combination crashes the kernel, in the case where the compiler
chooses to use a helper function for saving GPRs on entry, and the
module loader has placed the .init.text section far away from the
.text section, meaning that it has to generate a trampoline for
functions in the .init.text section to call the GPR save helper.
Because the trampoline trashes r11, references to the stack frame
using r11 can cause an oops.

The fix just uses GPR r12 instead of GPR r11 for generating the
trampoline code.  According to the statements from Freescale, this is
safe from an EABI perspective.

I've tested the fix for kernel 2.6.33 on MPC8541.

Signed-off-by: Steffen Rumler <steffen.rumler.ext@nsn.com>
[paulus@samba.org: reworded the description]
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agopowerpc/time: Sanity check of decrementer expiration is necessary
Paul Mackerras [Fri, 1 Jun 2012 08:13:43 +0000]
powerpc/time: Sanity check of decrementer expiration is necessary

commit 860aed25a1f0936d4852ab936252b47cd1e630f1 upstream.

This reverts 68568add2c ("powerpc/time: Remove unnecessary sanity check
of decrementer expiration").  We do need to check whether we have reached
the expiration time of the next event, because we sometimes get an early
decrementer interrupt, most notably when we set the decrementer to 1 in
arch_irq_work_raise().  The effect of not having the sanity check is that
if timer_interrupt() gets called early, we leave the decrementer set to
its maximum value, which means we then don't get any more decrementer
interrupts for about 4 seconds (or longer, depending on timebase
frequency).  I saw these pauses as a consequence of getting a stray
hypervisor decrementer interrupt left over from exiting a KVM guest.

This isn't quite a straight revert because of changes to the surrounding
code, but it restores the same algorithm as was previously used.

Acked-by: Anton Blanchard <anton@samba.org>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agobtree: fix tree corruption in btree_get_prev()
Roland Dreier [Thu, 7 Jun 2012 21:21:13 +0000]
btree: fix tree corruption in btree_get_prev()

commit cbf8ae32f66a9ceb8907ad9e16663c2a29e48990 upstream.

The memory the parameter __key points to is used as an iterator in
btree_get_prev(), so if we save off a bkey() pointer in retry_key and
then assign that to __key, we'll end up corrupting the btree internals
when we do eg

longcpy(__key, bkey(geo, node, i), geo->keylen);

to return the key value.  What we should do instead is use longcpy() to
copy the key value that retry_key points to __key.

This can cause a btree to get corrupted by seemingly read-only
operations such as btree_for_each_safe.

[akpm@linux-foundation.org: avoid the double longcpy()]
Signed-off-by: Roland Dreier <roland@purestorage.com>
Acked-by: Joern Engel <joern@logfs.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agochar/agp: add another Ironlake host bridge
Eugeni Dodonov [Wed, 6 Jun 2012 14:59:06 +0000]
char/agp: add another Ironlake host bridge

commit 67384fe3fd450536342330f684ea1f7dcaef8130 upstream.

This seems to come on Gigabyte H55M-S2V and was discovered through the
https://bugs.freedesktop.org/show_bug.cgi?id=50381 debugging.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=50381
Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/i915: Reset last_retired_head when resetting ring
Chris Wilson [Mon, 28 May 2012 21:33:02 +0000]
drm/i915: Reset last_retired_head when resetting ring

commit c3b20037926e607b6cdaecbf9d3103e2ca63bc31 upstream.

When we reset the ring control registers, including the HEAD and TAIL of
the ring, we also need to reset associated state. In this instance, we
were failing to reset the cached value of ring->last_retired_head and so
upon the first request for more space following a resume would
potentially (depending on a narrow race window) believe that the HEAD had
advanced much further than reality.

This is a regression from:

commit a71d8d94525e8fd855c0466fb586ae1cb008f3a2
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Feb 15 11:25:36 2012 +0000

    drm/i915: Record the tail at each request and use it to estimate the head

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/nouveau/disp: fix dithering not being enabled on some eDP macbooks
Ben Skeggs [Fri, 4 May 2012 14:39:21 +0000]
drm/nouveau/disp: fix dithering not being enabled on some eDP macbooks

commit a6a17859f1bdf607650ee055101f54c5f207762b upstream.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoLinux 3.4.2
Greg Kroah-Hartman [Sat, 9 Jun 2012 15:36:33 +0000]
Linux 3.4.2

7 years agoext4: don't set i_flags in EXT4_IOC_SETFLAGS
Tao Ma [Thu, 7 Jun 2012 23:04:19 +0000]
ext4: don't set i_flags in EXT4_IOC_SETFLAGS

commit b22b1f178f6799278d3178d894f37facb2085765 upstream.

Commit 7990696 uses the ext4_{set,clear}_inode_flags() functions to
change the i_flags automatically but fails to remove the error setting
of i_flags.  So we still have the problem of trashing state flags.
Fix this by removing the assignment.

Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoIA64: Add cmpxchg.h to exported userspace headers
maximilian attems [Wed, 16 May 2012 07:46:30 +0000]
IA64: Add cmpxchg.h to exported userspace headers

commit 98e4cff73a18af27f0d40d0b0d37f105dfc1994a upstream.

Fixes klibc build on ia64 after 85f8f7759e418c814ee2ceacf73eddb9bed39492.

Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: maximilian attems <max@stro.at>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/i915: enable vdd when switching off the eDP panel
Daniel Vetter [Sun, 20 May 2012 15:14:50 +0000]
drm/i915: enable vdd when switching off the eDP panel

commit 6cb49835da0426f69a2931bc2a0a8156344b0e41 upstream.

We have one bug report from a validation team that we get the eDP
panel sequencing still somewhat wrong: We need to enable VDD while
switching off the panel and backlight. Unfortunately that reporter
seems to have fallen off the earth :(

For another reporter this actually fixes a black panel issue because
without this the backlight/panel gets confused and doesn't light up
again.

v2: I've forgotten to remove the vdd_off call in panel_off which is
now bogus. This essentially reverts

commit 17038de5f16569a25343cf68668f3b657eafb00e
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Mon Apr 16 22:43:42 2012 +0100

    drm/i915/dp: Flush any outstanding work to turn the VDD off

v3: the current panel_off code forces off the vdd power, too. Which is
bogus and resulted in some funny warnings later on when we've tried to
do aux channel communications with just the vdd forced on. Fix this,
too.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=46312
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=43163
Tested-by: Vincent Frentzel <zcecc22@gmail.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/i915/dp: Flush any outstanding work to turn the VDD off
Chris Wilson [Mon, 16 Apr 2012 21:43:42 +0000]
drm/i915/dp: Flush any outstanding work to turn the VDD off

commit 17038de5f16569a25343cf68668f3b657eafb00e upstream.

As we may kick off a delayed workqueue task to switch of the VDD lines, we
need to complete that task prior to turning off the panel (which itself
depends upon VDD being off).

v2: Don't cancel the outstanding work as this may trigger a deadlock

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Keith Packard <keithp@keithp.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/i915: always use RPNSWREQ for turbo change requests
Jesse Barnes [Tue, 22 May 2012 16:30:33 +0000]
drm/i915: always use RPNSWREQ for turbo change requests

commit 89ba829e38bd500f438bc08af4229204c8ed7f35 upstream.

Media turbo requests can either use RPVSWREQ or RPNSWREQ to indicate
what the interrupt handler should do.  Since we only deal with the
latter in our turbo code, make the media engine use that for turbo
requests.

Tested-by: Joe Bloggsian <joebloggsian@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Reviewed-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agowl1251: fix oops on early interrupt
Grazvydas Ignotas [Fri, 18 May 2012 00:04:08 +0000]
wl1251: fix oops on early interrupt

commit f380f2c4a12e913356bd49f8790ec1063c4fe9f8 upstream.

This driver disables interrupt just after requesting it and enables it
later, after interface is up. However currently there is a time window
between request_irq() and disable_irq() where if interrupt arrives, the
driver oopses because it's not yet ready to process it. This can be
reproduced by inserting the module, associating and removing the module
multiple times.

Eliminate this race by setting IRQF_NOAUTOEN flag before request_irq().

Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomd: raid1/raid10: fix problem with merge_bvec_fn
NeilBrown [Thu, 31 May 2012 05:39:11 +0000]
md: raid1/raid10: fix problem with merge_bvec_fn

commit aba336bd1d46d6b0404b06f6915ed76150739057 upstream.

The new merge_bvec_fn which calls the corresponding function
in subsidiary devices requires that mddev->merge_check_needed
be set if any child has a merge_bvec_fn.

However were were only setting that when a device was hot-added,
not when a device was present from the start.

This bug was introduced in 3.4 so patch is suitable for 3.4.y
kernels.  However that are conflicts in raid10.c so a separate
patch will be needed for 3.4.y.

Reported-by: Sebastian Riemer <sebastian.riemer@profitbricks.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agovfs: Fix /proc/<tid>/fdinfo/<fd> file handling
Linus Torvalds [Mon, 4 Jun 2012 18:00:45 +0000]
vfs: Fix /proc/<tid>/fdinfo/<fd> file handling

commit 0640113be25d283e0ff77a9f041e1242182387f0 upstream.

Cyrill Gorcunov reports that I broke the fdinfo files with commit
30a08bf2d31d ("proc: move fd symlink i_mode calculations into
tid_fd_revalidate()"), and he's quite right.

The tid_fd_revalidate() function is not just used for the <tid>/fd
symlinks, it's also used for the <tid>/fdinfo/<fd> files, and the
permission model for those are different.

So do the dynamic symlink permission handling just for symlinks, making
the fdinfo files once more appear as the proper regular files they are.

Of course, Al Viro argued (probably correctly) that we shouldn't do the
symlink permission games at all, and make the symlinks always just be
the normal 'lrwxrwxrwx'.  That would have avoided this issue too, but
since somebody noticed that the permissions had changed (which was the
reason for that original commit 30a08bf2d31d in the first place), people
do apparently use this feature.

[ Basically, you can use the symlink permission data as a cheap "fdinfo"
  replacement, since you see whether the file is open for reading and/or
  writing by just looking at st_mode of the symlink.  So the feature
  does make sense, even if the pain it has caused means we probably
  shouldn't have done it to begin with. ]

Reported-and-tested-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoACPI battery: only refresh the sysfs files when pertinent information changes
Andy Whitcroft [Thu, 3 May 2012 13:48:26 +0000]
ACPI battery: only refresh the sysfs files when pertinent information changes

commit c5971456964290da7e98222892797b71ef793e62 upstream.

We only need to regenerate the sysfs files when the capacity units
change, avoid the update otherwise.

The origin of this issue is dates way back to 2.6.38:
da8aeb92d4853f37e281f11fddf61f9c7d84c3cd
(ACPI / Battery: Update information on info notification and resume)

Signed-off-by: Andy Whitcroft <apw@canonical.com>
Tested-by: Ralf Jung <post@ralfj.de>
Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiommu/amd: Fix deadlock in ppr-handling error path
Joerg Roedel [Fri, 1 Jun 2012 13:20:23 +0000]
iommu/amd: Fix deadlock in ppr-handling error path

commit eee53537c476c947bf7faa1c916d2f5a0ae8ec93 upstream.

In the error path of the ppr_notifer it can happen that the
iommu->lock is taken recursivly. This patch fixes the
problem by releasing the iommu->lock before any notifier is
invoked. This also requires to move the erratum workaround
for the ppr-log (interrupt may be faster than data in the log)
one function up.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiommu/amd: Cache pdev pointer to root-bridge
Joerg Roedel [Thu, 31 May 2012 15:38:11 +0000]
iommu/amd: Cache pdev pointer to root-bridge

commit c1bf94ec1e12d76838ad485158aecf208ebd8fb9 upstream.

At some point pci_get_bus_and_slot started to enable
interrupts. Since this function is used in the
amd_iommu_resume path it will enable interrupts on resume
which causes a warning. The fix will use a cached pointer
to the root-bridge to re-enable the IOMMU in case the BIOS
is broken.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon/kms: add new SI PCI ids
Alex Deucher [Tue, 5 Jun 2012 13:50:31 +0000]
drm/radeon/kms: add new SI PCI ids

commit 7aaa61b3476462b69f1ac7669fcca8d608ce3cb5 upstream.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon/kms: add new BTC PCI ids
Alex Deucher [Tue, 5 Jun 2012 13:50:30 +0000]
drm/radeon/kms: add new BTC PCI ids

commit a2bef8ce826dd1e787fd8ad9b6e0566ba59dab43 upstream.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon/kms: add new Palm, Sumo PCI ids
Alex Deucher [Tue, 5 Jun 2012 13:50:29 +0000]
drm/radeon/kms: add new Palm, Sumo PCI ids

commit 4a6991cc1fad514745b79181df3ace72d561e7aa upstream.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon/kms: add new Trinity PCI ids
Alex Deucher [Tue, 5 Jun 2012 13:50:28 +0000]
drm/radeon/kms: add new Trinity PCI ids

commit d430f7dbf7bd6aaaa40c0660b3204df8cf07b22b upstream.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon: fix vm deadlocks on cayman
Christian König [Sun, 3 Jun 2012 14:09:43 +0000]
drm/radeon: fix vm deadlocks on cayman

commit bb4091558228ff4a3e02328c931e683fc7f08722 upstream.

Locking mutex in different orders just screams for
deadlocks, and some testing showed that it is actually
quite easy to trigger them.

Signed-off-by: Christian König <deathsimple@vodafone.de>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon/audio: don't hardcode CRTC id
Rafał Miłecki [Mon, 4 Jun 2012 16:36:58 +0000]
drm/radeon/audio: don't hardcode CRTC id

commit 0aecb5a4ba1ea4167f31d1790eec027f1e658f2d upstream.

This is based on info released by AMD, should allow using audio in much
more cases.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoradix-tree: fix contiguous iterator
Konstantin Khlebnikov [Tue, 5 Jun 2012 17:36:33 +0000]
radix-tree: fix contiguous iterator

commit fffaee365fded09f9ebf2db19066065fa54323c3 upstream.

This patch fixes bug in macro radix_tree_for_each_contig().

If radix_tree_next_slot() sees NULL in next slot it returns NULL, but following
radix_tree_next_chunk() switches iterating into next chunk. As result iterating
becomes non-contiguous and breaks vfs "splice" and all its users.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Reported-and-bisected-by: Hans de Bruin <jmdebruin@xmsnet.nl>
Reported-and-bisected-by: Ondrej Zary <linux@rainbow-software.org>
Reported-bisected-and-tested-by: Toralf Förster <toralf.foerster@gmx.de>
Link: https://lkml.org/lkml/2012/6/5/64
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: remove mb_groups before tearing down the buddy_cache
Salman Qazi [Fri, 1 Jun 2012 03:52:14 +0000]
ext4: remove mb_groups before tearing down the buddy_cache

commit 95599968d19db175829fb580baa6b68939b320fb upstream.

We can't have references held on pages in the s_buddy_cache while we are
trying to truncate its pages and put the inode.  All the pages must be
gone before we reach clear_inode.  This can only be gauranteed if we
can prevent new users from grabbing references to s_buddy_cache's pages.

The original bug can be reproduced and the bug fix can be verified by:

while true; do mount -t ext4 /dev/ram0 /export/hda3/ram0; \
umount /export/hda3/ram0; done &

while true; do cat /proc/fs/ext4/ram0/mb_groups; done

Signed-off-by: Salman Qazi <sqazi@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: add ext4_mb_unload_buddy in the error path
Salman Qazi [Fri, 1 Jun 2012 03:51:27 +0000]
ext4: add ext4_mb_unload_buddy in the error path

commit 02b7831019ea4e7994968c84b5826fa8b248ffc8 upstream.

ext4_free_blocks fails to pair an ext4_mb_load_buddy with a matching
ext4_mb_unload_buddy when it fails a memory allocation.

Signed-off-by: Salman Qazi <sqazi@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: don't trash state flags in EXT4_IOC_SETFLAGS
Theodore Ts'o [Fri, 1 Jun 2012 03:46:01 +0000]
ext4: don't trash state flags in EXT4_IOC_SETFLAGS

commit 79906964a187c405db72a3abc60eb9b50d804fbc upstream.

In commit 353eb83c we removed i_state_flags with 64-bit longs, But
when handling the EXT4_IOC_SETFLAGS ioctl, we replace i_flags
directly, which trashes the state flags which are stored in the high
32-bits of i_flags on 64-bit platforms.  So use the the
ext4_{set,clear}_inode_flags() functions which use atomic bit
manipulation functions instead.

Reported-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: add missing save_error_info() to ext4_error()
Theodore Ts'o [Thu, 31 May 2012 03:00:16 +0000]
ext4: add missing save_error_info() to ext4_error()

commit f3fc0210c0fc91900766c995f089c39170e68305 upstream.

The ext4_error() function is missing a call to save_error_info().
Since this is the function which marks the file system as containing
an error, this oversight (which was introduced in 2.6.36) is quite
significant, and should be backported to older stable kernels with
high urgency.

Reported-by: Ken Sumrall <ksumrall@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: ksumrall@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: disallow hard-linked directory in ext4_lookup
Andreas Dilger [Mon, 28 May 2012 21:02:25 +0000]
ext4: disallow hard-linked directory in ext4_lookup

commit 7e936b737211e6b54e34b71a827e56b872e958d8 upstream.

A hard-linked directory to its parent can cause the VFS to deadlock,
and is a sign of a corrupted file system.  So detect this case in
ext4_lookup(), before the rmdir() lockup scenario can take place.

Signed-off-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: fix potential integer overflow in alloc_flex_gd()
Haogang Chen [Mon, 28 May 2012 18:21:55 +0000]
ext4: fix potential integer overflow in alloc_flex_gd()

commit 967ac8af4475ce45474800709b12137aa7634c77 upstream.

In alloc_flex_gd(), when flexbg_size is large, kmalloc size would
overflow and flex_gd->groups would point to a buffer smaller than
expected, causing OOB accesses when it is used.

Note that in ext4_resize_fs(), flexbg_size is calculated using
sbi->s_log_groups_per_flex, which is read from the disk and only bounded
to [1, 31]. The patch returns NULL for too large flexbg_size.

Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Haogang Chen <haogangchen@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: force ro mount if ext4_setup_super() fails
Eric Sandeen [Mon, 28 May 2012 18:17:25 +0000]
ext4: force ro mount if ext4_setup_super() fails

commit 7e84b6216467b84cd332c8e567bf5aa113fd2f38 upstream.

If ext4_setup_super() fails i.e. due to a too-high revision,
the error is logged in dmesg but the fs is not mounted RO as
indicated.

Tested by:

# mkfs.ext4 -r 4 /dev/sdb6
# mount /dev/sdb6 /mnt/test
# dmesg | grep "too high"
[164919.759248] EXT4-fs (sdb6): revision level too high, forcing read-only mode
# grep sdb6 /proc/mounts
/dev/sdb6 /mnt/test2 ext4 rw,seclabel,relatime,data=ordered 0 0

Reviewed-by: Andreas Dilger <adilger@whamcloud.com>
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoext4: fix potential NULL dereference in ext4_free_inodes_counts()
Dan Carpenter [Mon, 28 May 2012 18:16:57 +0000]
ext4: fix potential NULL dereference in ext4_free_inodes_counts()

commit bb3d132a24cd8bf5e7773b2d9f9baa58b07a7dae upstream.

The ext4_get_group_desc() function returns NULL on error, and
ext4_free_inodes_count() function dereferences it without checking.
There is a check on the next line, but it's too late.

Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoxfrm: take net hdr len into account for esp payload size calculation
Benjamin Poirier [Thu, 24 May 2012 11:32:38 +0000]
xfrm: take net hdr len into account for esp payload size calculation

[ Upstream commit 91657eafb64b4cb53ec3a2fbc4afc3497f735788 ]

Corrects the function that determines the esp payload size. The calculations
done in esp{4,6}_get_mtu() lead to overlength frames in transport mode for
certain mtu values and suboptimal frames for others.

According to what is done, mainly in esp{,6}_output() and tcp_mtu_to_mss(),
net_header_len must be taken into account before doing the alignment
calculation.

Signed-off-by: Benjamin Poirier <bpoirier@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoskb: avoid unnecessary reallocations in __skb_cow
Felix Fietkau [Tue, 29 May 2012 03:35:08 +0000]
skb: avoid unnecessary reallocations in __skb_cow

[ Upstream commit 617c8c11236716dcbda877e764b7bf37c6fd8063 ]

At the beginning of __skb_cow, headroom gets set to a minimum of
NET_SKB_PAD. This causes unnecessary reallocations if the buffer was not
cloned and the headroom is just below NET_SKB_PAD, but still more than the
amount requested by the caller.
This was showing up frequently in my tests on VLAN tx, where
vlan_insert_tag calls skb_cow_head(skb, VLAN_HLEN).

Locally generated packets should have enough headroom, and for forward
paths, we already have NET_SKB_PAD bytes of headroom, so we don't need to
add any extra space here.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agol2tp: fix oops in L2TP IP sockets for connect() AF_UNSPEC case
James Chapman [Tue, 29 May 2012 23:13:23 +0000]
l2tp: fix oops in L2TP IP sockets for connect() AF_UNSPEC case

[ Upstream commit c51ce49735c183ef2592db70f918ee698716276b ]

An application may call connect() to disconnect a socket using an
address with family AF_UNSPEC. The L2TP IP sockets were not handling
this case when the socket is not bound and an attempt to connect()
using AF_UNSPEC in such cases would result in an oops. This patch
addresses the problem by protecting the sk_prot->disconnect() call
against trying to unhash the socket before it is bound.

The patch also adds more checks that the sockaddr supplied to bind()
and connect() calls is valid.

 RIP: 0010:[<ffffffff82e133b0>]  [<ffffffff82e133b0>] inet_unhash+0x50/0xd0
 RSP: 0018:ffff88001989be28  EFLAGS: 00010293
 Stack:
  ffff8800407a8000 0000000000000000 ffff88001989be78 ffffffff82e3a249
  ffffffff82e3a050 ffff88001989bec8 ffff88001989be88 ffff8800407a8000
  0000000000000010 ffff88001989bec8 ffff88001989bea8 ffffffff82e42639
 Call Trace:
 [<ffffffff82e3a249>] udp_disconnect+0x1f9/0x290
 [<ffffffff82e42639>] inet_dgram_connect+0x29/0x80
 [<ffffffff82d012fc>] sys_connect+0x9c/0x100

Reported-by: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoipv6: fix incorrect ipsec fragment
Gao feng [Sat, 26 May 2012 01:30:53 +0000]
ipv6: fix incorrect ipsec fragment

[ Upstream commit 0c1833797a5a6ec23ea9261d979aa18078720b74 ]

Since commit ad0081e43a
"ipv6: Fragment locally generated tunnel-mode IPSec6 packets as needed"
the fragment of packets is incorrect.
because tunnel mode needs IPsec headers and trailer for all fragments,
while on transport mode it is sufficient to add the headers to the
first fragment and the trailer to the last.

so modify mtu and maxfraglen base on ipsec mode and if fragment is first
or last.

with my test,it work well(every fragment's size is the mtu)
and does not trigger slow fragment path.

Changes from v1:
though optimization, mtu_prev and maxfraglen_prev can be delete.
replace xfrm mode codes with dst_entry's new frag DST_XFRM_TUNNEL.
add fuction ip6_append_data_mtu to make codes clearer.

Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoipv4: fix the rcu race between free_fib_info and ip_route_output_slow
Yanmin Zhang [Wed, 23 May 2012 15:39:45 +0000]
ipv4: fix the rcu race between free_fib_info and ip_route_output_slow

[ Upstream commit e49cc0da7283088c5e03d475ffe2fdcb24a6d5b1 ]

We hit a kernel OOPS.

<3>[23898.789643] BUG: sleeping function called from invalid context at
/data/buildbot/workdir/ics/hardware/intel/linux-2.6/arch/x86/mm/fault.c:1103
<3>[23898.862215] in_atomic(): 0, irqs_disabled(): 0, pid: 10526, name:
Thread-6683
<4>[23898.967805] HSU serial 0000:00:05.1: 0000:00:05.2:HSU serial prevented me
to suspend...
<4>[23899.258526] Pid: 10526, comm: Thread-6683 Tainted: G        W
3.0.8-137685-ge7742f9 #1
<4>[23899.357404] HSU serial 0000:00:05.1: 0000:00:05.2:HSU serial prevented me
to suspend...
<4>[23899.904225] Call Trace:
<4>[23899.989209]  [<c1227f50>] ? pgtable_bad+0x130/0x130
<4>[23900.000416]  [<c1238c2a>] __might_sleep+0x10a/0x110
<4>[23900.007357]  [<c1228021>] do_page_fault+0xd1/0x3c0
<4>[23900.013764]  [<c18e9ba9>] ? restore_all+0xf/0xf
<4>[23900.024024]  [<c17c007b>] ? napi_complete+0x8b/0x690
<4>[23900.029297]  [<c1227f50>] ? pgtable_bad+0x130/0x130
<4>[23900.123739]  [<c1227f50>] ? pgtable_bad+0x130/0x130
<4>[23900.128955]  [<c18ea0c3>] error_code+0x5f/0x64
<4>[23900.133466]  [<c1227f50>] ? pgtable_bad+0x130/0x130
<4>[23900.138450]  [<c17f6298>] ? __ip_route_output_key+0x698/0x7c0
<4>[23900.144312]  [<c17f5f8d>] ? __ip_route_output_key+0x38d/0x7c0
<4>[23900.150730]  [<c17f63df>] ip_route_output_flow+0x1f/0x60
<4>[23900.156261]  [<c181de58>] ip4_datagram_connect+0x188/0x2b0
<4>[23900.161960]  [<c18e981f>] ? _raw_spin_unlock_bh+0x1f/0x30
<4>[23900.167834]  [<c18298d6>] inet_dgram_connect+0x36/0x80
<4>[23900.173224]  [<c14f9e88>] ? _copy_from_user+0x48/0x140
<4>[23900.178817]  [<c17ab9da>] sys_connect+0x9a/0xd0
<4>[23900.183538]  [<c132e93c>] ? alloc_file+0xdc/0x240
<4>[23900.189111]  [<c123925d>] ? sub_preempt_count+0x3d/0x50

Function free_fib_info resets nexthop_nh->nh_dev to NULL before releasing
fi. Other cpu might be accessing fi. Fixing it by delaying the releasing.

With the patch, we ran MTBF testing on Android mobile for 12 hours
and didn't trigger the issue.

Thank Eric for very detailed review/checking the issue.

Signed-off-by: Yanmin Zhang <yanmin_zhang@linux.intel.com>
Signed-off-by: Kun Jiang <kunx.jiang@intel.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoasix: allow full size 8021Q frames to be received
Eric Dumazet [Mon, 28 May 2012 22:31:41 +0000]
asix: allow full size 8021Q frames to be received

[ Upstream commit 9dae31009b1a00d926c6fe032d5a88099620adc3 ]

asix driver drops 8021Q full size frames because it doesn't take into
account VLAN header size.

Tested on AX88772 adapter.

Signed-off-by: Eric Dumazet <edumazet@google.com>
CC: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
CC: Allan Chou <allan@asix.com.tw>
CC: Trond Wuellner <trond@chromium.org>
CC: Grant Grundler <grundler@chromium.org>
CC: Paul Stewart <pstew@chromium.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/vmwgfx: Fix nasty write past alloced memory area
Thomas Hellstrom [Fri, 1 Jun 2012 13:48:04 +0000]
drm/vmwgfx: Fix nasty write past alloced memory area

commit 0824db38e515644f8d1bfd64adbd7cb2c6ea7c62 upstream.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/ttm: Fix spinlock imbalance
Thomas Hellstrom [Fri, 1 Jun 2012 13:39:11 +0000]
drm/ttm: Fix spinlock imbalance

commit a8ff3ee211fccf708e1911bbc096625453ebf759 upstream.

This imbalance may cause hangs when TTM is trying to swap out a buffer
that is already on the delayed delete list.

Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon: fix HD6790, HD6570 backend programming
Jerome Glisse [Thu, 31 May 2012 23:00:24 +0000]
drm/radeon: fix HD6790, HD6570 backend programming

commit 95c4b23ec4e2fa5604df229ddf134e31d7b3b378 upstream.

Without this bit sets we get broken rendering and
lockups.

fglrx sets this bit.

Bugs that should be fixed by this patch :
https://bugs.freedesktop.org/show_bug.cgi?id=49792
https://bugzilla.kernel.org/show_bug.cgi?id=43207
https://bugs.freedesktop.org/show_bug.cgi?id=39282

Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon: properly program gart on rv740, juniper, cypress, barts, hemlock
Alex Deucher [Thu, 31 May 2012 22:54:43 +0000]
drm/radeon: properly program gart on rv740, juniper, cypress, barts, hemlock

commit 0b8c30bc4943137a4a36b9cb059b1cc684f5d702 upstream.

Need to program an additional VM register.  This doesn't not currently
cause any problems, but allows us to program the proper backend
map in a subsequent patch which should improve performance on these
asics.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon: fix bank information in tiling config
Alex Deucher [Thu, 31 May 2012 22:53:36 +0000]
drm/radeon: fix bank information in tiling config

commit 29d654067a98c1cb8874c774e5fd799a038af8a6 upstream.

While there are cards with more than 8 mem banks, the max
number of banks from a tiling perspective is 8, so cap
the tiling config at 8 banks.

Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=43448

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon: fix regression in UMS CS ioctl
Alex Deucher [Wed, 30 May 2012 14:09:30 +0000]
drm/radeon: fix regression in UMS CS ioctl

commit 9b00147d9f2ba137ce74b66b768a8312be0b6781 upstream.

radeon_cs_parser_init is called by both the legacy UMS
CS ioctl and the KMS CS ioctl.  Protect KMS specific
pieces of the code by checking that rdev is not NULL.

Reported-by: Michael Burian <michael.burian@sbg.at>

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomtd: mxc_nand: move ecc strengh setup before nand_scan_tail
Sascha Hauer [Fri, 25 May 2012 14:22:42 +0000]
mtd: mxc_nand: move ecc strengh setup before nand_scan_tail

commit 4a43faf54e9173b6acce37cf7f053fc9515a2cdf upstream.

Since commit 6a918bade9dab40aaef80559bd1169c69e8d69cb, the mxc_nand driver
fails with:

Driver must set ecc.strength when using hardware ECC

This is because nand_scan_tail checks for correct ecc strength
settings, so we must set them up before nand_scan_tail.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomtd: block2mtd: fix recursive call of mtd_writev
Gabor Juhos [Wed, 23 May 2012 22:17:23 +0000]
mtd: block2mtd: fix recursive call of mtd_writev

commit 2e24e32e2759348c9290404abad4f729f791bfad upstream.

The 'mtd_writev' interface calls the function assigned
to the '_write' field of a given mtd device if that is
not NULL. The block2mtd driver sets the '_writev' field
to the 'mtd_writev' function itself and thus causes a
endless loop.

This is caused by 1dbebd32562b3c2caeca35960e5cb00bfcc12900
(mtd: harmonize mtd_writev usage).

Remove the assignment from the block2mtd driver to fix the
issue.

Signed-off-by: Gabor Juhos <juhosg@openwrt.org>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomtd: of_parts: fix breakage in Kconfig
Frank Svendsboe [Thu, 17 May 2012 20:43:09 +0000]
mtd: of_parts: fix breakage in Kconfig

commit 2e929d001e85126d9267de373d4b76014789661d upstream.

MTD_OF_PARTS and the default setting is not working due to using 'Y'
instead of 'y', introduced in commit
d6137badeff1ef64b4e0092ec249ebdeaeb3ff37. This made our board, and
possibly other boards using DTS defined partitions and not having
CONFIG_MTD_OF_PARTS=y defined in the defconfig, fail to mount root.

Signed-off-by: Frank Svendsboe <frank.svendsboe@gmail.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomtd: nand: fix scan_read_raw_oob
Dmitry Maluka [Fri, 11 May 2012 17:51:51 +0000]
mtd: nand: fix scan_read_raw_oob

commit 34a5704d91d6f8376a4c0a0143a1dd3eb3ccb37e upstream.

It seems there is a bug in scan_read_raw_oob() in nand_bbt.c which
should cause wrong functioning of NAND_BBT_SCANALLPAGES option.

Artem: the patch did not apply and I had to amend it a bit.

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agox86: Reset the debug_stack update counter
Steven Rostedt [Wed, 30 May 2012 15:43:19 +0000]
x86: Reset the debug_stack update counter

commit c0525a6972d3f1fb83058ef503e183475d6e4e26 upstream.

When an NMI goes off and it sees that it preempted the debug stack,
to keep the debug stack safe, it changes the IDT to point to one that
does not modify the stack on breakpoint (to allow breakpoints in NMIs).

But the variable that gets set to know to undo it on exit never gets
cleared on exit. Thus every NMI will reset it on exit the first time
it is done even if it does not need to be reset.

[ Added H. Peter Anvin's suggestion to use this_cpu_read/write ]

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agox86, x32, ptrace: Remove PTRACE_ARCH_PRCTL for x32
H.J. Lu [Tue, 22 May 2012 03:29:45 +0000]
x86, x32, ptrace: Remove PTRACE_ARCH_PRCTL for x32

commit bad1a753d4d4deb09d4bc0bac1dd4fc3298502e9 upstream.

When I added x32 ptrace to 3.4 kernel, I also include PTRACE_ARCH_PRCTL
support for x32 GDB  For ARCH_GET_FS/GS, it takes a pointer to int64.  But
at user level, ARCH_GET_FS/GS takes a pointer to int32.  So I have to add
x32 ptrace to glibc to handle it with a temporary int64 passed to kernel and
copy it back to GDB as int32.  Roland suggested that PTRACE_ARCH_PRCTL
is obsolete and x32 GDB should use fs_base and gs_base fields of
user_regs_struct instead.

Accordingly, remove PTRACE_ARCH_PRCTL completely from the x32 code to
avoid possible memory overrun when pointer to int32 is passed to
kernel.

Link: http://lkml.kernel.org/r/CAMe9rOpDzHfS7NH7m1vmD9QRw8SSj4Sc%2BaNOgcWm_WJME2eRsQ@mail.gmail.com
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agofec_mpc52xx: fix timestamp filtering
Stephan Gatzka [Sat, 2 Jun 2012 03:04:06 +0000]
fec_mpc52xx: fix timestamp filtering

commit 9ca3cc6f3026946ba655e863ca2096339e667639 upstream.

skb_defer_rx_timestamp was called with a freshly allocated skb but must
be called with rskb instead.

Signed-off-by: Stephan Gatzka <stephan@gatzka.org>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agovfs: increment iversion when a file is truncated
Dmitry Kasatkin [Tue, 29 May 2012 18:02:21 +0000]
vfs: increment iversion when a file is truncated

commit 799243a389bde0de10fa21ca1ca453d2fe538b85 upstream.

When a file is truncated with truncate()/ftruncate() and then closed,
iversion is not updated.  This patch uses ATTR_SIZE flag as an indication
to increment iversion.

Mimi said:

On fput(), i_version is used to detect and flag files that have changed
and need to be re-measured in the IMA measurement policy.  When a file
is truncated with truncate()/ftruncate() and then closed, i_version is
not updated.  As a result, although the file has changed, it will not be
re-measured and added to the IMA measurement list on subsequent access.

Signed-off-by: Dmitry Kasatkin <dmitry.kasatkin@intel.com>
Acked-by: Mimi Zohar <zohar@us.ibm.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agovfs: umount_tree() might be called on subtree that had never made it
Al Viro [Wed, 30 May 2012 02:03:48 +0000]
vfs: umount_tree() might be called on subtree that had never made it

commit 63d37a84ab6004c235314ffd7a76c5eb28c2fae0 upstream.

__mnt_make_shortterm() in there undoes the effect of __mnt_make_longterm()
we'd done back when we set ->mnt_ns non-NULL; it should not be done to
vfsmounts that had never gone through commit_tree() and friends.  Kudos to
lczerner for catching that one...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoslub: fix a memory leak in get_partial_node()
Joonsoo Kim [Wed, 16 May 2012 15:13:02 +0000]
slub: fix a memory leak in get_partial_node()

commit 02d7633fa567be7bf55a993b79d2a31b95ce2227 upstream.

In the case which is below,

1. acquire slab for cpu partial list
2. free object to it by remote cpu
3. page->freelist = t

then memory leak is occurred.

Change acquire_slab() not to zap freelist when it works for cpu partial list.
I think it is a sufficient solution for fixing a memory leak.

Below is output of 'slabinfo -r kmalloc-256'
when './perf stat -r 30 hackbench 50 process 4000 > /dev/null' is done.

***Vanilla***
Sizes (bytes)     Slabs              Debug                Memory
------------------------------------------------------------------------
Object :     256  Total  :     468   Sanity Checks : Off  Total: 3833856
SlabObj:     256  Full   :     111   Redzoning     : Off  Used : 2004992
SlabSiz:    8192  Partial:     302   Poisoning     : Off  Loss : 1828864
Loss   :       0  CpuSlab:      55   Tracking      : Off  Lalig:       0
Align  :       8  Objects:      32   Tracing       : Off  Lpadd:       0

***Patched***
Sizes (bytes)     Slabs              Debug                Memory
------------------------------------------------------------------------
Object :     256  Total  :     300   Sanity Checks : Off  Total: 2457600
SlabObj:     256  Full   :     204   Redzoning     : Off  Used : 2348800
SlabSiz:    8192  Partial:      33   Poisoning     : Off  Loss :  108800
Loss   :       0  CpuSlab:      63   Tracking      : Off  Lalig:       0
Align  :       8  Objects:      32   Tracing       : Off  Lpadd:       0

Total and loss number is the impact of this patch.

Acked-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Joonsoo Kim <js1304@gmail.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoALSA: usb-audio: fix rate_list memory leak
Clemens Ladisch [Fri, 18 May 2012 16:00:43 +0000]
ALSA: usb-audio: fix rate_list memory leak

commit 5cd5d7c44990658df6ab49f6253c39617c53b03d upstream.

The array of sample rates is reallocated every time when opening
the PCM device, but was freed only once when unplugging the device.

Reported-by: "Alexander E. Patrakov" <patrakov@gmail.com>
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomm: fix vma_resv_map() NULL pointer
Dave Hansen [Wed, 30 May 2012 14:51:07 +0000]
mm: fix vma_resv_map() NULL pointer

commit 4523e1458566a0e8ecfaff90f380dd23acc44d27 upstream.

hugetlb_reserve_pages() can be used for either normal file-backed
hugetlbfs mappings, or MAP_HUGETLB.  In the MAP_HUGETLB, semi-anonymous
mode, there is not a VMA around.  The new call to resv_map_put() assumed
that there was, and resulted in a NULL pointer dereference:

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000030
  IP: vma_resv_map+0x9/0x30
  PGD 141453067 PUD 1421e1067 PMD 0
  Oops: 0000 [#1] PREEMPT SMP
  ...
  Pid: 14006, comm: trinity-child6 Not tainted 3.4.0+ #36
  RIP: vma_resv_map+0x9/0x30
  ...
  Process trinity-child6 (pid: 14006, threadinfo ffff8801414e0000, task ffff8801414f26b0)
  Call Trace:
    resv_map_put+0xe/0x40
    hugetlb_reserve_pages+0xa6/0x1d0
    hugetlb_file_setup+0x102/0x2c0
    newseg+0x115/0x360
    ipcget+0x1ce/0x310
    sys_shmget+0x5a/0x60
    system_call_fastpath+0x16/0x1b

This was reported by Dave Jones, but was reproducible with the
libhugetlbfs test cases, so shame on me for not running them in the
first place.

With this, the oops is gone, and the output of libhugetlbfs's
run_tests.py is identical to plain 3.4 again.

[ Marked for stable, since this was introduced by commit c50ac050811d
  ("hugetlb: fix resv_map leak in error path") which was also marked for
  stable ]

Reported-by: Dave Jones <davej@redhat.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agokbuild: install kernel-page-flags.h
Ulrich Drepper [Tue, 29 May 2012 22:06:30 +0000]
kbuild: install kernel-page-flags.h

commit 9295b7a07c859a42346221b5839be0ae612333b0 upstream.

Programs using /proc/kpageflags need to know about the various flags.  The
<linux/kernel-page-flags.h> provides them and the comments in the file
indicate that it is supposed to be used by user-level code.  But the file
is not installed.

Install the headers and mark the unstable flags as out-of-bounds.  The
page-type tool is also adjusted to not duplicate the definitions

Signed-off-by: Ulrich Drepper <drepper@gmail.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/i915: no lvds quirk for HP t5740e Thin Client
Jan-Benedict Glaw [Tue, 22 May 2012 13:21:53 +0000]
drm/i915: no lvds quirk for HP t5740e Thin Client

commit 3347111999870c37eab1b969e90af9fdaf0334ba upstream.

This box has DisplayPort and VGA, but no LVDS. Product specs are at
http://h10010.www1.hp.com/wwpc/us/en/sm/WF25a/12454-12454-321959-338927-3640406-4282707.html?dnr=1
and dmidecode output can be found at http://www.getslash.de/bug_attachments/dmidecode-t5740e.txt

Signed-off-by: Jan-Benedict Glaw <jbglaw@getslash.de>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/i915: wait for a vblank to pass after tv detect
Daniel Vetter [Tue, 22 May 2012 19:41:25 +0000]
drm/i915: wait for a vblank to pass after tv detect

commit bf2125e2f7e931b50a6c76ba0435ba001409ccbf upstream.

Otherwise the hw will get confused and result in a black screen.

This regression has been most likely introduce in

commit 974b93315b2213b74a42a87e8a9d4fc8c0dbe90c
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sun Sep 5 00:44:20 2010 +0100

    drm/i915/tv: Poll for DAC state change

That commit replace the first msleep(20) with a busy-loop, but failed
to keep the 2nd msleep around. Later on we've replaced all these
msleep(20) by proper vblanks.

For reference also see the commit in xf86-video-intel:

commit 1142be53eb8d2ee8a9b60ace5d49f0ba27332275
Author: Jesse Barnes <jbarnes@hobbes.lan>
Date:   Mon Jun 9 08:52:59 2008 -0700

    Fix TV programming:  add vblank wait after TV_CTL writes

    Fxies FDO bug #14000; we need to wait for vblank after
    writing TV_CTL or following "DPMS on" calls may not actually enable the output.

v2: As suggested by Chris Wilson, add a small comment to ensure that
no one accidentally removes this vblank wait again - there really
seems to be no sane explanation for why we need it, but it is
required.

Launchpad: https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/763688
Reported-and-Tested-by: Robert Lowery <rglowery@exemail.com.au>
Cc: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/i915: Adding TV Out Missing modes.
Rodrigo Vivi [Tue, 22 May 2012 18:23:24 +0000]
drm/i915: Adding TV Out Missing modes.

commit 9589919fb3d269d4202a112b197468c7db1f97a3 upstream.

These 2 modes were removed by mistake during a clean up.
So, now it is time to add them back. For further info about
supported mode and standard timing table please check:
VOL_3_display_registers_updated.pdf at intellinuxgraphics.org.

Note that this regression has been introduce in

commit 55a6713b3f30a5024056027e9dbf03ac8f13bfc9
Author: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Date:   Thu Dec 15 14:47:33 2011 -0200

    drm/i915: Removing TV Out modes.

and this commit partially reverts it by re-adding the wrongly removed
modes.

Reported-by: Robert Lowery <rglowery@exemail.com.au>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
[danvet: Pimped commit message to cite the commit that introduced this
regression.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/i915: properly handle interlaced bit for sdvo dtd conversion
Daniel Vetter [Sat, 12 May 2012 20:22:58 +0000]
drm/i915: properly handle interlaced bit for sdvo dtd conversion

commit 59d92bfa5f0cdf57f82f5181b0ad6af75c3fdf41 upstream.

We've simply ignored this, which isn't too great. With this, interlaced
1080i works on my HDMI screen connected through sdvo. For no apparent
reason anything else still doesn't work as it should.

While at it, give these magic numbers in the dtd proper names and
add a comment that they match with EDID detailed timings.

v2: Actually use the right bit for interlaced.

Tested-by: Peter Ross <pross@xvid.org>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon: fix typo in trinity tiling setup
Alex Deucher [Fri, 25 May 2012 02:55:15 +0000]
drm/radeon: fix typo in trinity tiling setup

commit 1f73cca799d29df80de3e8f1f1c488485467577a upstream.

Using the wrong union.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agodrm/radeon: fix XFX quirk
Alex Deucher [Wed, 23 May 2012 15:48:59 +0000]
drm/radeon: fix XFX quirk

commit 1ebf169ad4dc68f18cc0dab35163b0f324fc6c41 upstream.

Only override the ddc bus if the connector doesn't have
a valid one.  The existing code overrode the ddc bus for
all connectors even if it had ddc bus.

Fixes ddc on another XFX card with the same pci ids that
was broken by the quirk overwriting the correct ddc bus.

Reported-by: Mehdi Aqadjani Memar <m.aqadjanimemar@student.ru.nl>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiommu/amd: Add workaround for event log erratum
Joerg Roedel [Thu, 12 Apr 2012 12:12:00 +0000]
iommu/amd: Add workaround for event log erratum

commit 3d06fca8d2aa3543030e40b95f1d62f9f5a03540 upstream.

Due to a recent erratum it can happen that the head pointer
of the event-log is updated before the actual event-log
entry is written. This patch implements the recommended
workaround.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiommu/amd: Check for the right TLP prefix bit
Joerg Roedel [Thu, 12 Apr 2012 10:49:26 +0000]
iommu/amd: Check for the right TLP prefix bit

commit a3b93121430c7b46c2895a7744261be107ccdf7f upstream.

Unfortunatly the PRI spec changed and moved the
TLP-prefix-required bit to a different location. This patch
makes the necessary change in the AMD IOMMU driver.
Regressions are not expected because all hardware
implementing the PRI capability sets this bit to zero
anyway.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agosunrpc: fix loss of task->tk_status after rpc_delay call in xprt_alloc_slot
Trond Myklebust [Sat, 19 May 2012 16:12:53 +0000]
sunrpc: fix loss of task->tk_status after rpc_delay call in xprt_alloc_slot

commit 1afeaf5c29aa07db25760d2fbed5c08a3aec3498 upstream.

xprt_alloc_slot will call rpc_delay() to make the task wait a bit before
retrying when it gets back an -ENOMEM error from xprt_dynamic_alloc_slot.
The problem is that rpc_delay will clear the task->tk_status, causing
call_reserveresult to abort the task.

The solution is simply to let call_reserveresult handle the ENOMEM error
directly.

Reported-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agohugetlb: fix resv_map leak in error path
Dave Hansen [Tue, 29 May 2012 22:06:46 +0000]
hugetlb: fix resv_map leak in error path

commit c50ac050811d6485616a193eb0f37bfbd191cc89 upstream.

When called for anonymous (non-shared) mappings, hugetlb_reserve_pages()
does a resv_map_alloc().  It depends on code in hugetlbfs's
vm_ops->close() to release that allocation.

However, in the mmap() failure path, we do a plain unmap_region() without
the remove_vma() which actually calls vm_ops->close().

This is a decent fix.  This leak could get reintroduced if new code (say,
after hugetlb_reserve_pages() in hugetlbfs_file_mmap()) decides to return
an error.  But, I think it would have to unroll the reservation anyway.

Christoph's test case:

http://marc.info/?l=linux-mm&m=133728900729735

This patch applies to 3.4 and later.  A version for earlier kernels is at
https://lkml.org/lkml/2012/5/22/418.

Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reported-by: Christoph Lameter <cl@linux.com>
Tested-by: Christoph Lameter <cl@linux.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoNFSv4: Map NFS4ERR_SHARE_DENIED into an EACCES error instead of EIO
Trond Myklebust [Mon, 28 May 2012 15:36:28 +0000]
NFSv4: Map NFS4ERR_SHARE_DENIED into an EACCES error instead of EIO

commit fb13bfa7e1bcfdcfdece47c24b62f1a1cad957e9 upstream.

If a file OPEN is denied due to a share lock, the resulting
NFS4ERR_SHARE_DENIED is currently mapped to the default EIO.
This patch adds a more appropriate mapping, and brings Linux
into line with what Solaris 10 does.

See https://bugzilla.kernel.org/show_bug.cgi?id=43286

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoNFS: kmalloc() doesn't return an ERR_PTR()
Dan Carpenter [Mon, 14 May 2012 19:45:28 +0000]
NFS: kmalloc() doesn't return an ERR_PTR()

commit 5abc03cd919535c61b813f2319cb38326a41e810 upstream.

Obviously we should check for NULL here instead of IS_ERR().

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agox86, amd, xen: Avoid NULL pointer paravirt references
Konrad Rzeszutek Wilk [Wed, 30 May 2012 22:23:56 +0000]
x86, amd, xen: Avoid NULL pointer paravirt references

commit 1ab46fd319bcf1fcd9fb6311727d532b580e4eba upstream.

Stub out MSR methods that aren't actually needed.  This fixes a crash
as Xen Dom0 on AMD Trinity systems.  A bigger patch should be added to
remove the paravirt machinery completely for the methods which
apparently have no users!

Reported-by: Andre Przywara <andre.przywara@amd.com>
Link: http://lkml.kernel.org/r/20120530222356.GA28417@andromeda.dapyr.net
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoath9k: fix a use-after-free-bug when ath_tx_setup_buffer() fails
Felix Fietkau [Thu, 24 May 2012 12:32:20 +0000]
ath9k: fix a use-after-free-bug when ath_tx_setup_buffer() fails

commit 81357a281dcc454841532c46b30e6f2ba12b73ea upstream.

ath_tx_setup_buffer() can fail if there is no ath_buf left, or if mapping DMA
failed. In this case it frees the skb passed to it.
If ath_tx_setup_buffer is called from ath_tx_form_aggr, the skb is still
linked into the tid buffer list and must be dequeued before being released.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoixp4xx: fix compilation by adding gpiolib support
Richard Cochran [Wed, 23 May 2012 16:19:51 +0000]
ixp4xx: fix compilation by adding gpiolib support

commit 9dde0ae3769875ec1370cb316e50c54b57d52c1a upstream.

Once again, ixp4xx no longer even compiles. This patch fixes the issue
by converting over to gpiolib. This patch was first made by Imre and
posted by Marc, and I added in Russell's suggestion to empty the gpio
header file.

This fix should also go for 3.1, 3.2, 3.3, and 3.4.

Signed-off-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomac80211: fix ADDBA declined after suspend with wowlan
Eyal Shapira [Tue, 29 May 2012 09:00:22 +0000]
mac80211: fix ADDBA declined after suspend with wowlan

commit 7b21aea04d084916ac4e0e8852dcc9cd60ec0d1d upstream.

WLAN_STA_BLOCK_BA is set while suspending but doesn't get cleared
when resuming in case of wowlan. This causes further ADDBA requests
received to be rejected. Fix it by clearing it in the wowlan path
as well.

Signed-off-by: Eyal Shapira <eyal@wizery.com>
Reviewed-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoMIPS: BCM63XX: Add missing include for bcm63xx_gpio.h
Jonas Gorski [Tue, 15 May 2012 15:47:52 +0000]
MIPS: BCM63XX: Add missing include for bcm63xx_gpio.h

commit 442209f31dafef9fde852858e1ce566b675b720d upstream.

bcm63xx_gpio.h uses macros defined in bcm63xx_cpu.h without including it,
leading to the following build failure:

CC [M] drivers/mmc/core/cd-gpio.o
In file included from arch/mips/include/asm/mach-bcm63xx/gpio.h:4:0,
                 from arch/mips/include/asm/gpio.h:4,
                 from include/linux/gpio.h:30,
                 from drivers/mmc/core/cd-gpio.c:12:

arch/mips/include/asm/mach-bcm63xx/bcm63xx_gpio.h: In function 'bcm63xx_gpio_count':
arch/mips/include/asm/mach-bcm63xx/bcm63xx_gpio.h:10:2: error: implicit declaration of function 'bcm63xx_get_cpu_id'
arch/mips/include/asm/mach-bcm63xx/bcm63xx_gpio.h:11:7: error: 'BCM6358_CPU_ID' undeclared (first use in this function)
arch/mips/include/asm/mach-bcm63xx/bcm63xx_gpio.h:11:7: note: each undeclared identifier is reported only once for each function it appears in
arch/mips/include/asm/mach-bcm63xx/bcm63xx_gpio.h:13:7: error: 'BCM6338_CPU_ID' undeclared (first use in this function)
arch/mips/include/asm/mach-bcm63xx/bcm63xx_gpio.h:15:7: error: 'BCM6345_CPU_ID' undeclared (first use in this function)
arch/mips/include/asm/mach-bcm63xx/bcm63xx_gpio.h:17:7: error: 'BCM6368_CPU_ID' undeclared (first use in this function)
arch/mips/include/asm/mach-bcm63xx/bcm63xx_gpio.h:19:7: error: 'BCM6348_CPU_ID' undeclared (first use in this function)

make[7]: *** [drivers/mmc/core/cd-gpio.o] Error 1

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Cc: linux-mips@linux-mips.org
Cc: Maxime Bizon <mbizon@freebox.fr>
Cc: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agosolos-pci: Fix DMA support
David Woodhouse [Thu, 24 May 2012 04:58:27 +0000]
solos-pci: Fix DMA support

commit b4bd8ad9bb311e8536f726f7a633620ccd358cde upstream.

DMA support has finally made its way to the top of the TODO list, having
realised that a Geode using MMIO can't keep up with two ADSL2+ lines
each running at 21Mb/s.

This patch fixes a couple of bugs in the DMA support in the driver, so
once the corresponding FPGA update is complete and tested everything
should work properly.

We weren't storing the currently-transmitting skb, so we were never
unmapping it and never freeing/popping it when the TX was done.
And the addition of pci_set_master() is fairly self-explanatory.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoPARISC: fix TLB fault path on PA2.0 narrow systems
James Bottomley [Mon, 21 May 2012 06:49:01 +0000]
PARISC: fix TLB fault path on PA2.0 narrow systems

commit 2f649c1f6f0fef445ce79a19b79e5ce8fe9d7f19 upstream.

commit 5e185581d7c46ddd33cd9c01106d1fc86efb9376
Author: James Bottomley <JBottomley@Parallels.com>

    [PARISC] fix PA1.1 oops on boot

Didn't quite fix the crash on boot.  It moved it from PA1.1 processors to
PA2.0 narrow kernels.  The final fix is to make sure the [id]tlb_miss_20 paths
also work.  Even on narrow systems, these paths require using the wide
instructions becuase the tlb insertion format is wide.  Fix this by
conditioning the dep[wd],z on whether we're being called from _11 or _20[w]
paths.

Tested-by: Helge Deller <deller@gmx.de>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoPARISC: fix boot failure on 32-bit systems caused by branch stubs placed before ...
John David Anglin [Thu, 17 May 2012 14:34:34 +0000]
PARISC: fix boot failure on 32-bit systems caused by branch stubs placed before .text

commit ed5fb2471b7060767957fb964eb1aaec71533ab1 upstream.

In certain configurations, the resulting kernel becomes too large to boot
because the linker places the long branch stubs for the merged .text section
at the very start of the image.  As a result, the initial transfer of control
jumps to an unexpected location.  Fix this by placing the head text in a
separate section so the stubs for .text are not at the start of the image.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agocifs: fix oops while traversing open file list (try #4)
Shirish Pargaonkar [Mon, 21 May 2012 14:20:12 +0000]
cifs: fix oops while traversing open file list (try #4)

commit 2c0c2a08bed7a3b791f88d09d16ace56acb3dd98 upstream.

While traversing the linked list of open file handles, if the identfied
file handle is invalid, a reopen is attempted and if it fails, we
resume traversing where we stopped and cifs can oops while accessing
invalid next element, for list might have changed.

So mark the invalid file handle and attempt reopen if no
valid file handle is found in rest of the list.
If reopen fails, move the invalid file handle to the end of the list
and start traversing the list again from the begining.
Repeat this four times before giving up and returning an error if
file reopen keeps failing.

Signed-off-by: Shirish Pargaonkar <shirishpargaonkar@gmail.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agocifs: Include backup intent search flags during searches {try #2)
Shirish Pargaonkar [Tue, 15 May 2012 15:19:16 +0000]
cifs: Include backup intent search flags during searches {try #2)

commit 2608bee744a92d60d15ff4e6e0b913d8b406aedd upstream.

As observed and suggested by Tushar Gosavi...

---------
readdir calls these function to send TRANS2_FIND_FIRST and
TRANS2_FIND_NEXT command to the server. The current cifs module is
not specifying CIFS_SEARCH_BACKUP_SEARCH flag while sending these
command when backupuid/backupgid is specified. This can be resolved
by specifying CIFS_SEARCH_BACKUP_SEARCH flag.
---------

Reported-and-Tested-by: Tushar Gosavi <tugosavi@in.ibm.com>
Signed-off-by: Shirish Pargaonkar <shirishpargaonkar@gmail.com>
Acked-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Steve French <sfrench@us.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiwlwifi: do not use shadow registers by default
Meenakshi Venkataraman [Wed, 16 May 2012 20:35:59 +0000]
iwlwifi: do not use shadow registers by default

commit 66a770729a5cdd24efed8afa5258f81232d8bba2 upstream.

Shadow registers in the device are meant to
allow the driver to update certain device
registers without needing to wake up all
components of the device. However, using
this feature in the device causes
communication between the driver and the
device to become unreliable, resulting in
host command timeouts.

Disable this feature by default till a fix is
available for the bug.

Signed-off-by: Meenakshi Venkataraman <meenakshi.venkataraman@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agoiwlwifi: update BT traffic load states correctly
Meenakshi Venkataraman [Wed, 16 May 2012 20:35:57 +0000]
iwlwifi: update BT traffic load states correctly

commit 882dde8eb0d49ce0f853f8f4084dde56a21fe55f upstream.

When BT traffic load changes from its
previous state, a new LQ command needs to be
sent down to the firmware. This needs to
be done only once per change. The state
variable that keeps track of this change is
last_bt_traffic_load. However, it was not
being updated when the change had been
handled. Not updating this variable was
causing a flood of advanced BT config
commands to be sent to the firmware. Fix
this.

Signed-off-by: Meenakshi Venkataraman <meenakshi.venkataraman@intel.com>
Signed-off-by: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomm: fix faulty initialization in vmalloc_init()
KyongHo [Tue, 29 May 2012 22:06:49 +0000]
mm: fix faulty initialization in vmalloc_init()

commit dbda591d920b4c7692725b13e3f68ecb251e9080 upstream.

The transfer of ->flags causes some of the static mapping virtual
addresses to be prematurely freed (before the mapping is removed) because
VM_LAZY_FREE gets "set" if tmp->flags has VM_IOREMAP set.  This might
cause subsequent vmalloc/ioremap calls to fail because it might allocate
one of the freed virtual address ranges that aren't unmapped.

va->flags has different types of flags from tmp->flags.  If a region with
VM_IOREMAP set is registered with vm_area_add_early(), it will be removed
by __purge_vmap_area_lazy().

Fix vmalloc_init() to correctly initialize vmap_area for the given
vm_struct.

Also initialise va->vm.  If it is not set, find_vm_area() for the early
vm regions will always fail.

Signed-off-by: KyongHo Cho <pullip.cho@samsung.com>
Cc: "Olav Haugan" <ohaugan@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomm: consider all swapped back pages in used-once logic
Michal Hocko [Tue, 29 May 2012 22:06:45 +0000]
mm: consider all swapped back pages in used-once logic

commit e48982734ea0500d1eba4f9d96195acc5406cad6 upstream.

Commit 645747462435 ("vmscan: detect mapped file pages used only once")
made mapped pages have another round in inactive list because they might
be just short lived and so we could consider them again next time.  This
heuristic helps to reduce pressure on the active list with a streaming
IO worklods.

This patch fixes a regression introduced by this commit for heavy shmem
based workloads because unlike Anon pages, which are excluded from this
heuristic because they are usually long lived, shmem pages are handled
as a regular page cache.

This doesn't work quite well, unfortunately, if the workload is mostly
backed by shmem (in memory database sitting on 80% of memory) with a
streaming IO in the background (backup - up to 20% of memory).  Anon
inactive list is full of (dirty) shmem pages when watermarks are hit.
Shmem pages are kept in the inactive list (they are referenced) in the
first round and it is hard to reclaim anything else so we reach lower
scanning priorities very quickly which leads to an excessive swap out.

Let's fix this by excluding all swap backed pages (they tend to be long
lived wrt.  the regular page cache anyway) from used-once heuristic and
rather activate them if they are referenced.

The customer's workload is shmem backed database (80% of RAM) and they
are measuring transactions/s with an IO in the background (20%).
Transactions touch more or less random rows in the table.  The
transaction rate fell by a factor of 3 (in the worst case) because of
commit 64574746.  This patch restores the previous numbers.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomm: fix NULL ptr deref when walking hugepages
Sasha Levin [Tue, 29 May 2012 22:06:15 +0000]
mm: fix NULL ptr deref when walking hugepages

commit 08fa29d916c6e271ad13978cd993e7238c68db97 upstream.

A missing validation of the value returned by find_vma() could cause a
NULL ptr dereference when walking the pagetable.

This is triggerable from usermode by a simple user by trying to read a
page info out of /proc/pid/pagemap which doesn't exist.

Introduced by commit 025c5b2451e4 ("thp: optimize away unnecessary page
table locking").

Signed-off-by: Sasha Levin <levinsasha928@gmail.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: 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>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

7 years agomm/fork: fix overflow in vma length when copying mmap on clone
Siddhesh Poyarekar [Tue, 29 May 2012 22:06:22 +0000]
mm/fork: fix overflow in vma length when copying mmap on clone

commit 7edc8b0ac16cbaed7cb4ea4c6b95ce98d2997e84 upstream.

The vma length in dup_mmap is calculated and stored in a unsigned int,
which is insufficient and hence overflows for very large maps (beyond
16TB). The following program demonstrates this:

#include <stdio.h>
#include <unistd.h>
#include <sys/mman.h>

#define GIG 1024 * 1024 * 1024L
#define EXTENT 16393

int main(void)
{
        int i, r;
        void *m;
        char buf[1024];

        for (i = 0; i < EXTENT; i++) {
                m = mmap(NULL, (size_t) 1 * 1024 * 1024 * 1024L,
                         PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);

                if (m == (void *)-1)
                        printf("MMAP Failed: %d\n", m);
                else
                        printf("%d : MMAP returned %p\n", i, m);

                r = fork();

                if (r == 0) {
                        printf("%d: successed\n", i);
                        return 0;
                } else if (r < 0)
                        printf("FORK Failed: %d\n", r);
                else if (r > 0)
                        wait(NULL);
        }
        return 0;
}

Increase the storage size of the result to unsigned long, which is
sufficient for storing the difference between addresses.

Signed-off-by: Siddhesh Poyarekar <siddhesh.poyarekar@gmail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
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>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>