pandora-kernel.git
11 years agoALSA: hda - Fix invalid connections in VT1802 codec
Takashi Iwai [Wed, 7 Nov 2012 09:37:48 +0000 (10:37 +0100)]
ALSA: hda - Fix invalid connections in VT1802 codec

commit ef4da45828603df57e5e21b8aa21a66ce309f79b upstream.

VT1802 codec provides the invalid connection lists of NID 0x24 and
0x33 containing the routes to a non-exist widget 0x3e.  This confuses
the auto-parser.  Fix it up in the driver by overriding these
connections.

Reported-by: Massimo Del Fedele <max@veneto.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: hda - Fix empty DAC filling in patch_via.c
Takashi Iwai [Wed, 7 Nov 2012 09:32:47 +0000 (10:32 +0100)]
ALSA: hda - Fix empty DAC filling in patch_via.c

commit 5b3761954dac2d1393beef8210eb8cee81d16b8d upstream.

In via_auto_fill_adc_nids(), the parser tries to fill dac_nids[] at
the point of the current line-out (i).  When no valid path is found
for this output, this results in dac = 0, thus it creates a hole in
dac_nids[].  This confuses is_empty_dac() and trims the detected DAC
in later reference.

This patch fixes the bug by appending DAC properly to dac_nids[] in
via_auto_fill_adc_nids().

Reported-by: Massimo Del Fedele <max@veneto.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrm: restore open_count if drm_setup fails
Ilija Hadzic [Mon, 29 Oct 2012 17:35:00 +0000 (17:35 +0000)]
drm: restore open_count if drm_setup fails

commit 0f1cb1bd94a9c967cd4ad3de51cfdabe61eb5dcc upstream.

If drm_setup (called at first open) fails, the whole
open call has failed, so we should not keep the
open_count incremented.

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Thomas Hellstrom <thellstrom@vmware.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agohwmon: (w83627ehf) Force initial bank selection
Jean Delvare [Mon, 5 Nov 2012 20:54:39 +0000 (21:54 +0100)]
hwmon: (w83627ehf) Force initial bank selection

commit 3300fb4f88688029fff8dfb9ec0734f6e4cba3e7 upstream.

Don't assume bank 0 is selected at device probe time. This may not be
the case. Force bank selection at first register access to guarantee
that we read the right registers upon driver loading.

Signed-off-by: Jean Delvare <khali@linux-fr.org>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: hda - Force to reset IEC958 status bits for AD codecs
Takashi Iwai [Mon, 5 Nov 2012 11:32:46 +0000 (12:32 +0100)]
ALSA: hda - Force to reset IEC958 status bits for AD codecs

commit ae24c3191ba2ab03ec6b4be323e730e00404b4b6 upstream.

Several bug reports suggest that the forcibly resetting IEC958 status
bits is required for AD codecs to get the SPDIF output working
properly after changing streams.

Original fix credit to Javeed Shaikh.

BugLink: https://bugs.launchpad.net/ubuntu/+source/alsa-driver/+bug/359361
Reported-by: Robin Kreis <r.kreis@uni-bremen.de>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: HDA: Fix digital microphone on CS420x
Daniel J Blueman [Sun, 4 Nov 2012 05:19:03 +0000 (13:19 +0800)]
ALSA: HDA: Fix digital microphone on CS420x

commit 16337e028a6dae9fbdd718c0d42161540a668ff3 upstream.

Correctly enable the digital microphones with the right bits in the
right coeffecient registers on Cirrus CS4206/7 codecs. It also
prevents misconfiguring ADC1/2.

This fixes the digital mic on the Macbook Pro 10,1/Retina.

Based-on-patch-by: Alexander Stein <alexander.stein@systec-electronic.com>
Signed-off-by: Daniel J Blueman <daniel@quora.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: hda: Cirrus: Fix coefficient index for beep configuration
Alexander Stein [Thu, 1 Nov 2012 12:42:37 +0000 (13:42 +0100)]
ALSA: hda: Cirrus: Fix coefficient index for beep configuration

commit 5a83b4b5a391f07141b157ac9daa51c409e71ab5 upstream.

Signed-off-by: Alexander Stein <alexander.stein@systec-electronic.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agocifs: fix potential buffer overrun in cifs.idmap handling code
Jeff Layton [Sat, 3 Nov 2012 13:37:28 +0000 (09:37 -0400)]
cifs: fix potential buffer overrun in cifs.idmap handling code

commit 36960e440ccf94349c09fb944930d3bfe4bc473f upstream.

The userspace cifs.idmap program generally works with the wbclient libs
to generate binary SIDs in userspace. That program defines the struct
that holds these values as having a max of 15 subauthorities. The kernel
idmapping code however limits that value to 5.

When the kernel copies those values around though, it doesn't sanity
check the num_subauths value handed back from userspace or from the
server. It's possible therefore for userspace to hand us back a bogus
num_subauths value (or one that's valid, but greater than 5) that could
cause the kernel to walk off the end of the cifs_sid->sub_auths array.

Fix this by defining a new routine for copying sids and using that in
all of the places that copy it. If we end up with a sid that's longer
than expected then this approach will just lop off the "extra" subauths,
but that's basically what the code does today already. Better approaches
might be to fix this code to reject SIDs with >5 subauths, or fix it
to handle the subauths array dynamically.

At the same time, change the kernel to check the length of the data
returned by userspace. If it's shorter than struct cifs_sid, reject it
and return -EIO. If that happens we'll end up with fields that are
basically uninitialized.

Long term, it might make sense to redefine cifs_sid using a flexarray at
the end, to allow for variable-length subauth lists, and teach the code
to handle the case where the subauths array being passed in from
userspace is shorter than 5 elements.

Note too, that I don't consider this a security issue since you'd need
a compromised cifs.idmap program. If you have that, you can do all sorts
of nefarious stuff. Still, this is probably reasonable for stable.

Reviewed-by: Shirish Pargaonkar <shirishpargaonkar@gmail.com>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoDRM/Radeon: Fix Load Detection on legacy primary DAC.
Egbert Eich [Wed, 24 Oct 2012 16:29:49 +0000 (18:29 +0200)]
DRM/Radeon: Fix Load Detection on legacy primary DAC.

commit 83325d072185899b706de2956170b246585aaec9 upstream.

An uninitialized variable led to broken load detection.

Signed-off-by: Egbert Eich <eich@suse.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoiscsi-target: Fix missed wakeup race in TX thread
Roland Dreier [Wed, 31 Oct 2012 16:16:46 +0000 (09:16 -0700)]
iscsi-target: Fix missed wakeup race in TX thread

commit d5627acba9ae584cf4928af19f7ddf5f6837de32 upstream.

The sleeping code in iscsi_target_tx_thread() is susceptible to the classic
missed wakeup race:

 - TX thread finishes handle_immediate_queue() and handle_response_queue(),
   thinks both queues are empty.
 - Another thread adds a queue entry and does wake_up_process(), which does
   nothing because the TX thread is still awake.
 - TX thread does schedule_timeout() and sleeps forever.

In practice this can kill an iSCSI connection if for example an initiator
does single-threaded writes and the target misses the wakeup window when
queueing an R2T; in this case the connection will be stuck until the
initiator loses patience and does some task management operation (or kills
the connection entirely).

Fix this by converting to wait_event_interruptible(), which does not
suffer from this sort of race.

Signed-off-by: Roland Dreier <roland@purestorage.com>
Cc: Andy Grover <agrover@redhat.com>
Cc: Hannes Reinecke <hare@suse.de>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agotarget: Avoid integer overflow in se_dev_align_max_sectors()
Roland Dreier [Wed, 31 Oct 2012 16:16:45 +0000 (09:16 -0700)]
target: Avoid integer overflow in se_dev_align_max_sectors()

commit 3e03989b5868acf69a391a424dc71fcd6cc48167 upstream.

The expression (max_sectors * block_size) might overflow a u32
(indeed, since iblock sets max_hw_sectors to UINT_MAX, it is
guaranteed to overflow and end up with a much-too-small result in many
common cases).  Fix this by doing an equivalent calculation that
doesn't require multiplication.

While we're touching this code, avoid splitting a printk format across
two lines and use pr_info(...) instead of printk(KERN_INFO ...).

Signed-off-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agotarget: Don't return success from module_init() if setup fails
Roland Dreier [Wed, 31 Oct 2012 16:16:44 +0000 (09:16 -0700)]
target: Don't return success from module_init() if setup fails

commit 0d0f9dfb31e0a6c92063e235417b42df185b3275 upstream.

If the call to core_dev_release_virtual_lun0() fails, then nothing
sets ret to anything other than 0, so even though everything is
torn down and freed, target_core_init_configfs() will seem to succeed
and the module will be loaded.  Fix this by passing the return value
on up the chain.

Signed-off-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agonfs: Show original device name verbatim in /proc/*/mount{s,info}
Ben Hutchings [Sun, 21 Oct 2012 18:23:52 +0000 (19:23 +0100)]
nfs: Show original device name verbatim in /proc/*/mount{s,info}

commit 97a54868262da1629a3e65121e65b8e8c4419d9f upstream.

Since commit c7f404b ('vfs: new superblock methods to override
/proc/*/mount{s,info}'), nfs_path() is used to generate the mounted
device name reported back to userland.

nfs_path() always generates a trailing slash when the given dentry is
the root of an NFS mount, but userland may expect the original device
name to be returned verbatim (as it used to be).  Make this
canonicalisation optional and change the callers accordingly.

[jrnieder@gmail.com: use flag instead of bool argument]
Reported-and-tested-by: Chris Hiestand <chiestand@salk.edu>
Reference: http://bugs.debian.org/669314
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
[bwh: Backported to 3.2:
 - Adjust context
 - nfs_show_devname() still takes a pointer to struct vfsmount]

11 years agonfsv3: Make v3 mounts fail with ETIMEDOUTs instead EIO on mountd timeouts
Scott Mayhew [Tue, 16 Oct 2012 17:22:19 +0000 (13:22 -0400)]
nfsv3: Make v3 mounts fail with ETIMEDOUTs instead EIO on mountd timeouts

commit acce94e68a0f346115fd41cdc298197d2d5a59ad upstream.

In very busy v3 environment, rpc.mountd can respond to the NULL
procedure but not the MNT procedure in a timely manner causing
the MNT procedure to time out. The problem is the mount system
call returns EIO which causes the mount to fail, instead of
ETIMEDOUT, which would cause the mount to be retried.

This patch sets the RPC_TASK_SOFT|RPC_TASK_TIMEOUT flags to
the rpc_call_sync() call in nfs_mount() which causes
ETIMEDOUT to be returned on timed out connections.

Signed-off-by: Steve Dickson <steved@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoNFS: fix bug in legacy DNS resolver.
NeilBrown [Wed, 31 Oct 2012 01:16:01 +0000 (12:16 +1100)]
NFS: fix bug in legacy DNS resolver.

commit 8d96b10639fb402357b75b055b1e82a65ff95050 upstream.

The DNS resolver's use of the sunrpc cache involves a 'ttl' number
(relative) rather that a timeout (absolute).  This confused me when
I wrote
  commit c5b29f885afe890f953f7f23424045cdad31d3e4
     "sunrpc: use seconds since boot in expiry cache"

and I managed to break it.  The effect is that any TTL is interpreted
as 0, and nothing useful gets into the cache.

This patch removes the use of get_expiry() - which really expects an
expiry time - and uses get_uint() instead, treating the int correctly
as a ttl.

This fixes a regression that has been present since 2.6.37, causing
certain NFS accesses in certain environments to incorrectly fail.

Reported-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoNFSv4: nfs4_locku_done must release the sequence id
Trond Myklebust [Mon, 29 Oct 2012 22:53:23 +0000 (18:53 -0400)]
NFSv4: nfs4_locku_done must release the sequence id

commit 2b1bc308f492589f7d49012ed24561534ea2be8c upstream.

If the state recovery machinery is triggered by the call to
nfs4_async_handle_error() then we can deadlock.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoNFSv4.1: We must release the sequence id when we fail to get a session slot
Trond Myklebust [Mon, 29 Oct 2012 22:37:40 +0000 (18:37 -0400)]
NFSv4.1: We must release the sequence id when we fail to get a session slot

commit 2240a9e2d013d8269ea425b73e1d7a54c7bc141f upstream.

If we do not release the sequence id in cases where we fail to get a
session slot, then we can deadlock if we hit a recovery scenario.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
[bwh: Backported to 3.2:
 - Adjust context
 - nfs4_setup_sequence() has an additional 'cache_reply' parameter]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoNFS: Wait for session recovery to finish before returning
Bryan Schumaker [Tue, 30 Oct 2012 20:06:35 +0000 (16:06 -0400)]
NFS: Wait for session recovery to finish before returning

commit 399f11c3d872bd748e1575574de265a6304c7c43 upstream.

Currently, we will schedule session recovery and then return to the
caller of nfs4_handle_exception.  This works for most cases, but causes
a hang on the following test case:

Client Server
------ ------
Open file over NFS v4.1
Write to file
Expire client
Try to lock file

The server will return NFS4ERR_BADSESSION, prompting the client to
schedule recovery.  However, the client will continue placing lock
attempts and the open recovery never seems to be scheduled.  The
simplest solution is to wait for session recovery to run before retrying
the lock.

Signed-off-by: Bryan Schumaker <bjschuma@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoxen/mmu: Use Xen specific TLB flush instead of the generic one.
Konrad Rzeszutek Wilk [Wed, 31 Oct 2012 16:38:31 +0000 (12:38 -0400)]
xen/mmu: Use Xen specific TLB flush instead of the generic one.

commit 95a7d76897c1e7243d4137037c66d15cbf2cce76 upstream.

As Mukesh explained it, the MMUEXT_TLB_FLUSH_ALL allows the
hypervisor to do a TLB flush on all active vCPUs. If instead
we were using the generic one (which ends up being xen_flush_tlb)
we end up making the MMUEXT_TLB_FLUSH_LOCAL hypercall. But
before we make that hypercall the kernel will IPI all of the
vCPUs (even those that were asleep from the hypervisor
perspective). The end result is that we needlessly wake them
up and do a TLB flush when we can just let the hypervisor
do it correctly.

This patch gives around 50% speed improvement when migrating
idle guest's from one host to another.

Oracle-bug: 14630170

Tested-by: Jingjie Jiang <jingjie.jiang@oracle.com>
Suggested-by: Mukesh Rathor <mukesh.rathor@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoHID: microsoft: fix invalid rdesc for 3k kbd
Jiri Slaby [Fri, 19 Oct 2012 11:28:46 +0000 (13:28 +0200)]
HID: microsoft: fix invalid rdesc for 3k kbd

commit 3ccc60f9d8c39180c205dba1a020735bda1b2491 upstream.

Microsoft Digital Media Keyboard 3000 has two interfaces, and the
second one has a report descriptor with a bug. The second collection
says:
05 01 -- global; usage page -- 01 -- Generic Desktop Controls
09 80 -- local; usage -- 80 -- System Control
a1 01 -- main; collection -- 01 -- application

85 03 -- global; report ID -- 03
19 00 -- local; Usage Minimum -- 00
29 ff -- local; Usage Maximum -- ff
15 00 -- global; Logical Minimum -- 0
26 ff 00 -- global; Logical Maximum -- ff
81 00 -- main; input

c0 -- main; End Collection

I.e. it makes us think that there are all kinds of usages of system
control. That the keyboard is a not only a keyboard, but also a
joystick, mouse, gamepad, keypad, etc. The same as for the Wireless
Desktop Receiver, this should be Physical Min/Max. So fix that
appropriately.

References: https://bugzilla.novell.com/show_bug.cgi?id=776834
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoInput: tsc40 - remove wrong announcement of pressure support
Rolf Eike Beer [Wed, 31 Oct 2012 06:39:10 +0000 (23:39 -0700)]
Input: tsc40 - remove wrong announcement of pressure support

commit 32ed1911fc79908d704023317d4ddeb3883fd07e upstream.

The tsc40 driver announces it supports the pressure event, but will never
send one. The announcement will cause tslib to wait for such events and
sending all touch events with a pressure of 0. Removing the announcement
will make tslib fall back to emulating the pressure on touch events so
everything works as expected.

Signed-off-by: Rolf Eike Beer <eike-kernel@sf-tec.de>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agomodule: fix out-by-one error in kallsyms
Rusty Russell [Thu, 25 Oct 2012 00:19:25 +0000 (10:49 +1030)]
module: fix out-by-one error in kallsyms

commit 59ef28b1f14899b10d6b2682c7057ca00a9a3f47 upstream.

Masaki found and patched a kallsyms issue: the last symbol in a
module's symtab wasn't transferred.  This is because we manually copy
the zero'th entry (which is always empty) then copy the rest in a loop
starting at 1, though from src[0].  His fix was minimal, I prefer to
rewrite the loops in more standard form.

There are two loops: one to get the size, and one to copy.  Make these
identical: always count entry 0 and any defined symbol in an allocated
non-init section.

This bug exists since the following commit was introduced.
   module: reduce symbol table for loaded modules (v2)
   commit: 4a4962263f07d14660849ec134ee42b63e95ea9a

LKML: http://lkml.org/lkml/2012/10/24/27
Reported-by: Masaki Kimura <masaki.kimura.kz@hitachi.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
[bwh: Backported to 3.2: we're still using a bitmap to compress the string
 table]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoath9k: Test for TID only in BlockAcks while checking tx status
Sven Eckelmann [Mon, 29 Oct 2012 12:25:20 +0000 (13:25 +0100)]
ath9k: Test for TID only in BlockAcks while checking tx status

commit 6fe7cc71bbf3a0bc28c9cec3c00bc11e81344412 upstream.

The ath9k xmit functions for AMPDUs can send frames as non-aggregate in case
only one frame is currently available. The client will then answer using a
normal Ack instead of a BlockAck. This acknowledgement has no TID stored and
therefore the hardware is not able to provide us the corresponding TID.

The TID set by the hardware in the tx status descriptor has to be seen as
undefined and not as a valid TID value for normal acknowledgements. Doing
otherwise results in a massive amount of retransmissions and stalls of
connections.

Users may experience low bandwidth and complete connection stalls in
environments with transfers using multiple TIDs.

This regression was introduced in b11b160defc48e4daa283f785192ea3a23a51f8e
("ath9k: validate the TID in the tx status information").

Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Acked-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agort2800: validate step value for temperature compensation
Stanislaw Gruszka [Thu, 25 Oct 2012 07:51:39 +0000 (09:51 +0200)]
rt2800: validate step value for temperature compensation

commit bf7e1abe434ba9e22e8dc04a4cba4ab504b788b8 upstream.

Some hardware has correct (!= 0xff) value of tssi_bounds[4] in the
EEPROM, but step is equal to 0xff. This results on ridiculous delta
calculations and completely broke TX power settings.

Reported-and-tested-by: Pavel Lucik <pavel.lucik@gmail.com>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Acked-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoxen/gntdev: don't leak memory from IOCTL_GNTDEV_MAP_GRANT_REF
David Vrabel [Wed, 24 Oct 2012 11:39:02 +0000 (12:39 +0100)]
xen/gntdev: don't leak memory from IOCTL_GNTDEV_MAP_GRANT_REF

commit a67baeb77375199bbd842fa308cb565164dd1f19 upstream.

map->kmap_ops allocated in gntdev_alloc_map() wasn't freed by
gntdev_put_map().

Add a gntdev_free_map() helper function to free everything allocated
by gntdev_alloc_map().

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agofloppy: properly handle failure on add_disk loop
Herton Ronaldo Krzesinski [Mon, 27 Aug 2012 23:56:54 +0000 (20:56 -0300)]
floppy: properly handle failure on add_disk loop

commit d60e7ec18c3fb2cbf90969ccd42889eb2d03aef9 upstream.

On floppy initialization, if something failed inside the loop we call
add_disk, there was no cleanup of previous iterations in the error
handling.

Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agofloppy: do put_disk on current dr if blk_init_queue fails
Herton Ronaldo Krzesinski [Mon, 27 Aug 2012 23:56:52 +0000 (20:56 -0300)]
floppy: do put_disk on current dr if blk_init_queue fails

commit 238ab78469c6ab7845b43d5061cd3c92331b2452 upstream.

If blk_init_queue fails, we do not call put_disk on the current dr
(dr is decremented first in the error handling loop).

Reviewed-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoath9k: fix stale pointers potentially causing access to free'd skbs
Felix Fietkau [Thu, 25 Oct 2012 22:31:11 +0000 (00:31 +0200)]
ath9k: fix stale pointers potentially causing access to free'd skbs

commit 8c6e30936a7893a85f6222084f0f26aceb81137a upstream.

bf->bf_next is only while buffers are chained as part of an A-MPDU
in the tx queue. When a tid queue is flushed (e.g. on tearing down
an aggregation session), frames can be enqueued again as normal
transmission, without bf_next being cleared. This can lead to the
old pointer being dereferenced again later.

This patch might fix crashes and "Failed to stop TX DMA!" messages.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agomac80211: make sure data is accessible in EAPOL check
Johannes Berg [Thu, 25 Oct 2012 22:41:23 +0000 (00:41 +0200)]
mac80211: make sure data is accessible in EAPOL check

commit 6dbda2d00d466225f9db1dc695ff852443f28832 upstream.

The code to allow EAPOL frames even when the station
isn't yet marked associated needs to check that the
incoming frame is long enough and due to paged RX it
also can't assume skb->data contains the right data,
it must use skb_copy_bits(). Fix this to avoid using
data that doesn't really exist.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agomac80211: verify that skb data is present
Johannes Berg [Thu, 25 Oct 2012 22:36:40 +0000 (00:36 +0200)]
mac80211: verify that skb data is present

commit 9b395bc3be1cebf0144a127c7e67d56dbdac0930 upstream.

A number of places in the mesh code don't check that
the frame data is present and in the skb header when
trying to access. Add those checks and the necessary
pskb_may_pull() calls. This prevents accessing data
that doesn't actually exist.

To do this, export ieee80211_get_mesh_hdrlen() to be
able to use it in mac80211.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agomac80211: check management frame header length
Johannes Berg [Thu, 25 Oct 2012 22:33:36 +0000 (00:33 +0200)]
mac80211: check management frame header length

commit 4a4f1a5808c8bb0b72a4f6e5904c53fb8c9cd966 upstream.

Due to pskb_may_pull() checking the skb length, all
non-management frames are checked on input whether
their 802.11 header is fully present. Also add that
check for management frames and remove a check that
is now duplicate. This prevents accessing skb data
beyond the frame end.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agowireless: drop invalid mesh address extension frames
Johannes Berg [Thu, 25 Oct 2012 19:51:59 +0000 (21:51 +0200)]
wireless: drop invalid mesh address extension frames

commit 7dd111e8ee10cc6816669eabcad3334447673236 upstream.

The mesh header can have address extension by a 4th
or a 5th and 6th address, but never both. Drop such
frames in 802.11 -> 802.3 conversion along with any
frames that have the wrong extension.

Reviewed-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agomac80211: fix SSID copy on IBSS JOIN
Antonio Quartulli [Fri, 26 Oct 2012 16:54:25 +0000 (18:54 +0200)]
mac80211: fix SSID copy on IBSS JOIN

commit badecb001a310408d3473b1fc2ed5aefd0bc92a9 upstream.

The 'ssid' field of the cfg80211_ibss_params is a u8 pointer and
its length is likely to be less than IEEE80211_MAX_SSID_LEN most
of the time.

This patch fixes the ssid copy in ieee80211_ibss_join() by using
the SSID length to prevent it from reading beyond the string.

Signed-off-by: Antonio Quartulli <ordex@autistici.org>
[rewrapped commit message, small rewording]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agomac80211: don't inspect Sequence Control field on control frames
Javier Cardona [Thu, 25 Oct 2012 18:10:18 +0000 (11:10 -0700)]
mac80211: don't inspect Sequence Control field on control frames

commit f7fbf70ee9db6da6033ae50d100e017ac1f26555 upstream.

Per IEEE Std. 802.11-2012, Sec 8.2.4.4.1, the sequence Control field is
not present in control frames.  We noticed this problem when processing
Block Ack Requests.

Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: Javier Lopez <jlopex@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agomac80211: Only process mesh config header on frames that RA_MATCH
Javier Cardona [Wed, 24 Oct 2012 19:43:30 +0000 (12:43 -0700)]
mac80211: Only process mesh config header on frames that RA_MATCH

commit 555cb715be8ef98b8ec362b23dfc254d432a35b1 upstream.

Doing otherwise is wrong, and may wreak havoc on the mpp tables,
specially if the frame is encrypted.

Reported-by: Chaoxing Lin <Chaoxing.Lin@ultra-3eti.com>
Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
[bwh: Backported to 3.2: we have a large block conditional on
 IEEE80211_RX_RA_MATCH rather than a goto conditional on the opposite,
 so delete the condition]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoARM: at91/i2c: change id to let i2c-gpio work
Bo Shen [Mon, 15 Oct 2012 09:30:27 +0000 (17:30 +0800)]
ARM: at91/i2c: change id to let i2c-gpio work

commit 7840487cd6298f9f931103b558290d8d98d41c49 upstream.

The i2c core driver will turn the platform device ID to busnum
When using platfrom device ID as -1, it means dynamically assigned
the busnum. When writing code, we need to make sure the busnum,
and call i2c_register_board_info(int busnum, ...) to register device
if using -1, we do not know the value of busnum

In order to solve this issue, set the platform device ID as a fix number
Here using 0 to match the busnum used in i2c_regsiter_board_info()

Signed-off-by: Bo Shen <voice.shen@atmel.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Acked-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
Acked-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoARM: at91: at91sam9g10: fix SOC type detection
Ivan Shugov [Wed, 24 Oct 2012 09:02:44 +0000 (11:02 +0200)]
ARM: at91: at91sam9g10: fix SOC type detection

commit 3d9a0183dd3423353e9e363bcc261c1220d05f9f upstream.

Newer at91sam9g10 SoC revision can't be detected, so the kernel can't boot with
this kind of kernel panic:
"AT91: Impossible to detect the SOC type"

CPU: ARM926EJ-S [41069265] revision 5 (ARMv5TEJ), cr=00053177
CPU: VIVT data cache, VIVT instruction cache
Machine: Atmel AT91SAM9G10-EK
Ignoring tag cmdline (using the default kernel command line)
bootconsole [earlycon0] enabled
Memory policy: ECC disabled, Data cache writeback
Kernel panic - not syncing: AT91: Impossible to detect the SOC type
[<c00133d4>] (unwind_backtrace+0x0/0xe0) from [<c02366dc>] (panic+0x78/0x1cc)
[<c02366dc>] (panic+0x78/0x1cc) from [<c02fa35c>] (at91_map_io+0x90/0xc8)
[<c02fa35c>] (at91_map_io+0x90/0xc8) from [<c02f9860>] (paging_init+0x564/0x6d0)
[<c02f9860>] (paging_init+0x564/0x6d0) from [<c02f7914>] (setup_arch+0x464/0x704)
[<c02f7914>] (setup_arch+0x464/0x704) from [<c02f44f8>] (start_kernel+0x6c/0x2d4)
[<c02f44f8>] (start_kernel+0x6c/0x2d4) from [<20008040>] (0x20008040)

The reason for this is that the Debug Unit Chip ID Register has changed between
Engineering Sample and definitive revision of the SoC. Changing the check of
cidr to socid will address the problem. We do not integrate this check to the
list just above because we also have to make sure that the extended id is
disregarded.

Signed-off-by: Ivan Shugov <ivan.shugov@gmail.com>
[nicolas.ferre@atmel.com: change commit message]
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Acked-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agocrypto: cryptd - disable softirqs in cryptd_queue_worker to prevent data corruption
Jussi Kivilinna [Sun, 21 Oct 2012 17:42:28 +0000 (20:42 +0300)]
crypto: cryptd - disable softirqs in cryptd_queue_worker to prevent data corruption

commit 9efade1b3e981f5064f9db9ca971b4dc7557ae42 upstream.

cryptd_queue_worker attempts to prevent simultaneous accesses to crypto
workqueue by cryptd_enqueue_request using preempt_disable/preempt_enable.
However cryptd_enqueue_request might be called from softirq context,
so add local_bh_disable/local_bh_enable to prevent data corruption and
panics.

Bug report at http://marc.info/?l=linux-crypto-vger&m=134858649616319&w=2

v2:
 - Disable software interrupts instead of hardware interrupts

Reported-by: Gurucharan Shetty <gurucharan.shetty@gmail.com>
Signed-off-by: Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrm/i915: clear the entire sdvo infoframe buffer
Daniel Vetter [Sun, 21 Oct 2012 10:52:39 +0000 (12:52 +0200)]
drm/i915: clear the entire sdvo infoframe buffer

commit b6e0e543f75729f207b9c72b0162ae61170635b2 upstream.

Like in the case of native hdmi, which is fixed already in

commit adf00b26d18e1b3570451296e03bcb20e4798cdd
Author: Paulo Zanoni <paulo.r.zanoni@intel.com>
Date:   Tue Sep 25 13:23:34 2012 -0300

    drm/i915: make sure we write all the DIP data bytes

we need to clear the entire sdvo buffer to avoid upsetting the
display.

Since infoframe buffer writing is now a bit more elaborate, extract it
into it's own function. This will be useful if we ever get around to
properly update the ELD for sdvo. Also #define proper names for the
two buffer indexes with fixed usage.

v2: Cite the right commit above, spotted by Paulo Zanoni.

v3: I'm too stupid to paste the right commit.

v4: Ben Hutchings noticed that I've failed to handle an underflow in
my loop logic, breaking it for i >= length + 8. Since I've just lost C
programmer license, use his solution. Also, make the frustrated 0-base
buffer size a notch more clear.

Reported-and-tested-by: Jürg Billeter <j@bitron.ch>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=25732
Cc: Paulo Zanoni <przanoni@gmail.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrm/i915: fixup infoframe support for sdvo
Daniel Vetter [Sat, 12 May 2012 18:22:00 +0000 (20:22 +0200)]
drm/i915: fixup infoframe support for sdvo

commit 81014b9d0b55fb0b48f26cd2a943359750d532db upstream.

At least the worst offenders:
- SDVO specifies that the encoder should compute the ecc. Testing also
  shows that we must not send the ecc field, so copy the dip_infoframe
  struct to a temporay place and avoid the ecc field. This way the avi
  infoframe is exactly 17 bytes long, which agrees with what the spec
  mandates as a minimal storage capacity (with the ecc field it would
  be 18 bytes).
- Only 17 when sending the avi infoframe. The SDVO spec explicitly
  says that sending more data than what the device announces results
  in undefined behaviour.
- Add __attribute__((packed)) to the avi and spd infoframes, for
  otherwise they're wrongly aligned. Noticed because the avi infoframe
  ended up being 18 bytes large instead of 17. We haven't noticed this
  yet because we don't use the uint16_t fields yet (which are the only
  ones that would be wrongly aligned).

This regression has been introduce by

3c17fe4b8f40a112a85758a9ab2aebf772bdd647 is the first bad commit
commit 3c17fe4b8f40a112a85758a9ab2aebf772bdd647
Author: David Härdeman <david@hardeman.nu>
Date:   Fri Sep 24 21:44:32 2010 +0200

    i915: enable AVI infoframe for intel_hdmi.c [v4]

Patch tested on my g33 with a sdvo hdmi adaptor.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=25732
Tested-by: Peter Ross <pross@xvid.org> (G35 SDVO-HDMI)
Reviewed-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrm/i915: fix overlay on i830M
Daniel Vetter [Mon, 22 Oct 2012 10:55:55 +0000 (12:55 +0200)]
drm/i915: fix overlay on i830M

commit a9193983f4f292a82a00c72971c17ec0ee8c6c15 upstream.

The overlay on the i830M has a peculiar failure mode: It works the
first time around after boot-up, but consistenly hangs the second time
it's used.

Chris Wilson has dug out a nice errata:

"1.5.12 Clock Gating Disable for Display Register
Address Offset: 06200h–06203h

"Bit 3
Ovrunit Clock Gating Disable.
0 = Clock gating controlled by unit enabling logic
1 = Disable clock gating function
DevALM Errata ALM049: Overlay Clock Gating Must be Disabled:  Overlay
& L2 Cache clock gating must be disabled in order to prevent device
hangs when turning off overlay.SW must turn off Ovrunit clock gating
(6200h) and L2 Cache clock gating (C8h)."

Now I've nowhere found that 0xc8 register and hence couldn't apply the
l2 cache workaround. But I've remembered that part of the magic that
the OVERLAY_ON/OFF commands are supposed to do is to rearrange cache
allocations so that the overlay scaler has some scratch space.

And while pondering how that could explain the hang the 2nd time we
enable the overlay, I've remembered that the old ums overlay code did
_not_ issue the OVERLAY_OFF cmd.

And indeed, disabling the OFF cmd results in the overlay working
flawlessly, so I guess we can workaround the lack of the above
workaround by simply never disabling the overlay engine once it's
enabled.

Note that we have the first part of the above w/a already implemented
in i830_init_clock_gating - leave that as-is to avoid surprises.

v2: Add a comment in the code.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=47827
Tested-by: Rhys <rhyspuk@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
[bwh: Backported to 3.2:
 - Adjust context
 - s/intel_ring_emit(ring, /OUT_RING(/]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agocfg80211: fix antenna gain handling
Felix Fietkau [Wed, 17 Oct 2012 11:56:19 +0000 (13:56 +0200)]
cfg80211: fix antenna gain handling

commit c4a9fafc77a5318f5ed26c509bbcddf03e18c201 upstream.

No driver initializes chan->max_antenna_gain to something sensible, and
the only place where it is being used right now is inside ath9k. This
leads to ath9k potentially using less tx power than it can use, which can
decrease performance/range in some rare cases.

Rather than going through every single driver, this patch initializes
chan->orig_mag in wiphy_register(), ignoring whatever value the driver
left in there. If a driver for some reason wishes to limit it independent
from regulatory rulesets, it can do so internally.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agogpio-timberdale: fix a potential wrapping issue
Dan Carpenter [Thu, 11 Oct 2012 06:56:35 +0000 (09:56 +0300)]
gpio-timberdale: fix a potential wrapping issue

commit d79550a7bc35c16476ebdc27c78378d8093390ec upstream.

->last_ier is an unsigned long but the high bits can't be used int the
original code because the shift wraps.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agortnetlink: fix rtnl_calcit() and rtnl_dump_ifinfo()
Eric Dumazet [Sun, 4 Mar 2012 12:32:10 +0000 (12:32 +0000)]
rtnetlink: fix rtnl_calcit() and rtnl_dump_ifinfo()

commit a4b64fbe482c7766f7925f03067fc637716bfa3f upstream.

nlmsg_parse() might return an error, so test its return value before
potential random memory accesses.

Errors introduced in commit 115c9b81928 (rtnetlink: Fix problem with
buffer allocation)

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agortnetlink: Fix problem with buffer allocation
Greg Rose [Tue, 21 Feb 2012 21:54:48 +0000 (16:54 -0500)]
rtnetlink: Fix problem with buffer allocation

commit 115c9b81928360d769a76c632bae62d15206a94a upstream.

Implement a new netlink attribute type IFLA_EXT_MASK.  The mask
is a 32 bit value that can be used to indicate to the kernel that
certain extended ifinfo values are requested by the user application.
At this time the only mask value defined is RTEXT_FILTER_VF to
indicate that the user wants the ifinfo dump to send information
about the VFs belonging to the interface.

This patch fixes a bug in which certain applications do not have
large enough buffers to accommodate the extra information returned
by the kernel with large numbers of SR-IOV virtual functions.
Those applications will not send the new netlink attribute with
the interface info dump request netlink messages so they will
not get unexpectedly large request buffers returned by the kernel.

Modifies the rtnl_calcit function to traverse the list of net
devices and compute the minimum buffer size that can hold the
info dumps of all matching devices based upon the filter passed
in via the new netlink attribute filter mask.  If no filter
mask is sent then the buffer allocation defaults to NLMSG_GOODSIZE.

With this change it is possible to add yet to be defined netlink
attributes to the dump request which should make it fairly extensible
in the future.

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.2: drop the change in do_setlink() that reverts
 commit f18da14565819ba43b8321237e2426a2914cc2ef, which we never applied]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: mos7840: remove unused variable
Johan Hovold [Thu, 8 Nov 2012 17:28:59 +0000 (18:28 +0100)]
USB: mos7840: remove unused variable

Fix warning about unused variable introduced by commit e681b66f2e19fa
("USB: mos7840: remove invalid disconnect handling") upstream.

A subsequent fix which removed the disconnect function got rid of the
warning but that one was only backported to v3.6.

Reported-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrm/radeon: fix typo in evergreen_mc_resume()
Alex Deucher [Mon, 5 Nov 2012 16:34:58 +0000 (16:34 +0000)]
drm/radeon: fix typo in evergreen_mc_resume()

commit 695ddeb457584a602f2ba117d08ce37cf6ec1589 upstream.

Add missing index that may have led us to enabling
more crtcs than necessary.

May also fix:
https://bugs.freedesktop.org/show_bug.cgi?id=56139

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: Avoid endless sleep after disconnect
Takashi Iwai [Tue, 16 Oct 2012 14:43:39 +0000 (16:43 +0200)]
ALSA: Avoid endless sleep after disconnect

commit 0914f7961babbf28aaa2f19b453951fb4841c03f upstream.

When disconnect callback is called, each component should wake up
sleepers and check card->shutdown flag for avoiding the endless sleep
blocking the proper resource release.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: Add a reference counter to card instance
Takashi Iwai [Tue, 16 Oct 2012 11:05:59 +0000 (13:05 +0200)]
ALSA: Add a reference counter to card instance

commit a0830dbd4e42b38aefdf3fb61ba5019a1a99ea85 upstream.

For more strict protection for wild disconnections, a refcount is
introduced to the card instance, and let it up/down when an object is
referred via snd_lookup_*() in the open ops.

The free-after-last-close check is also changed to check this refcount
instead of the empty list, too.

Reported-by: Matthieu CASTET <matthieu.castet@parrot.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: usb-audio: Fix races at disconnection in mixer_quirks.c
Takashi Iwai [Mon, 15 Oct 2012 10:40:37 +0000 (12:40 +0200)]
ALSA: usb-audio: Fix races at disconnection in mixer_quirks.c

commit 888ea7d5ac6815ba16b3b3a20f665a92c7af6724 upstream.

Similar like the previous commit, cover with chip->shutdown_rwsem
and chip->shutdown checks.

Reported-by: Matthieu CASTET <matthieu.castet@parrot.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: usb-audio: Use rwsem for disconnect protection
Takashi Iwai [Mon, 15 Oct 2012 10:16:02 +0000 (12:16 +0200)]
ALSA: usb-audio: Use rwsem for disconnect protection

commit 34f3c89fda4fba9fe689db22253ca8db2f5e6386 upstream.

Replace mutex with rwsem for codec->shutdown protection so that
concurrent accesses are allowed.

Also add the protection to snd_usb_autosuspend() and
snd_usb_autoresume(), too.

Reported-by: Matthieu CASTET <matthieu.castet@parrot.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: usb-audio: Fix races at disconnection
Takashi Iwai [Fri, 12 Oct 2012 13:12:55 +0000 (15:12 +0200)]
ALSA: usb-audio: Fix races at disconnection

commit 978520b75f0a1ce82b17e1e8186417250de6d545 upstream.

Close some races at disconnection of a USB audio device by adding the
chip->shutdown_mutex and chip->shutdown check at appropriate places.

The spots to put bandaids are:
- PCM prepare, hw_params and hw_free
- where the usb device is accessed for communication or get speed, in
 mixer.c and others; the device speed is now cached in subs->speed
 instead of accessing to chip->dev

The accesses in PCM open and close don't need the mutex protection
because these are already handled in the core PCM disconnection code.

The autosuspend/autoresume codes are still uncovered by this patch
because of possible mutex deadlocks.  They'll be covered by the
upcoming change to rwsem.

Also the mixer codes are untouched, too.  These will be fixed in
another patch, too.

Reported-by: Matthieu CASTET <matthieu.castet@parrot.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: PCM: Fix some races at disconnection
Takashi Iwai [Fri, 12 Oct 2012 13:07:34 +0000 (15:07 +0200)]
ALSA: PCM: Fix some races at disconnection

commit 9b0573c07f278e9888c352aa9724035c75784ea0 upstream.

Fix races at PCM disconnection:
- while a PCM device is being opened or closed
- while the PCM state is being changed without lock in prepare,
  hw_params, hw_free ops

Reported-by: Matthieu CASTET <matthieu.castet@parrot.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agonfsd: add get_uint for u32's
J. Bruce Fields [Tue, 12 Jun 2012 20:54:16 +0000 (16:54 -0400)]
nfsd: add get_uint for u32's

commit a007c4c3e943ecc054a806c259d95420a188754b upstream.

I don't think there's a practical difference for the range of values
these interfaces should see, but it would be safer to be unambiguous.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrm/nouveau: headless mode by default if pci class != vga display
Ben Skeggs [Wed, 31 Oct 2012 02:16:06 +0000 (12:16 +1000)]
drm/nouveau: headless mode by default if pci class != vga display

commit e412e95a268fa8544858ebfe066826b290430d51 upstream.

This is to prevent nouveau from taking over the console on headless boards
such as Tesla.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrm/nouveau: fix suspend/resume when in headless mode
Ben Skeggs [Thu, 1 Nov 2012 06:16:31 +0000 (16:16 +1000)]
drm/nouveau: fix suspend/resume when in headless mode

commit 9430738d80223a1cd791a2baa74fa170d3df1262 upstream.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
[jrnieder@gmail.com: backport to 3.2: make fbcon suspend/resume
 handling conditional in a vague hope that this will approximate what
 the original does for 3.3+]
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrm/nouveau: silence modesetting spam on pre-gf8 chipsets
Ben Skeggs [Sun, 28 Oct 2012 23:03:07 +0000 (09:03 +1000)]
drm/nouveau: silence modesetting spam on pre-gf8 chipsets

commit cee59f15a60cc6269a25e3f6fbf1a577d6ab8115 upstream.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agox86: Remove the ancient and deprecated disable_hlt() and enable_hlt() facility
Len Brown [Thu, 29 Mar 2012 21:49:17 +0000 (14:49 -0700)]
x86: Remove the ancient and deprecated disable_hlt() and enable_hlt() facility

commit f6365201d8a21fb347260f89d6e9b3e718d63c70 upstream.

The X86_32-only disable_hlt/enable_hlt mechanism was used by the
32-bit floppy driver. Its effect was to replace the use of the
HLT instruction inside default_idle() with cpu_relax() - essentially
it turned off the use of HLT.

This workaround was commented in the code as:

 "disable hlt during certain critical i/o operations"

 "This halt magic was a workaround for ancient floppy DMA
  wreckage. It should be safe to remove."

H. Peter Anvin additionally adds:

 "To the best of my knowledge, no-hlt only existed because of
  flaky power distributions on 386/486 systems which were sold to
  run DOS.  Since DOS did no power management of any kind,
  including HLT, the power draw was fairly uniform; when exposed
  to the much hhigher noise levels you got when Linux used HLT
  caused some of these systems to fail.

  They were by far in the minority even back then."

Alan Cox further says:

 "Also for the Cyrix 5510 which tended to go castors up if a HLT
  occurred during a DMA cycle and on a few other boxes HLT during
  DMA tended to go astray.

  Do we care ? I doubt it. The 5510 was pretty obscure, the 5520
  fixed it, the 5530 is probably the oldest still in any kind of
  use."

So, let's finally drop this.

Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Josh Boyer <jwboyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Stephen Hemminger <shemminger@vyatta.com
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-3rhk9bzf0x9rljkv488tloib@git.kernel.org
[ If anyone cares then alternative instruction patching could be
  used to replace HLT with a one-byte NOP instruction. Much simpler. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
[bwh: Backported to 3.2: adjust filename, context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoBluetooth: Avoid calling undefined smp_conn_security()
Ben Hutchings [Fri, 16 Nov 2012 16:46:46 +0000 (16:46 +0000)]
Bluetooth: Avoid calling undefined smp_conn_security()

Commit ff03261adc8b4bdd8291f1783c079b53a892b429 ('Bluetooth: Fix
sending a HCI Authorization Request over LE links', commit
d8343f125710fb596f7a88cd756679f14f4e77b9 upstream) added an
call to smp_conn_security() from hci_conn_security().  The
former is only defined if CONFIG_BT_L2CAP=y.

This is not required in mainline since commit
f1e91e1640d808d332498a6b09b2bcd01462eff9 ('Bluetooth: Always compile
SCO and L2CAP in Bluetooth Core') removed that option.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoLinux 3.2.33 v3.2.33
Ben Hutchings [Tue, 30 Oct 2012 23:27:11 +0000 (23:27 +0000)]
Linux 3.2.33

11 years agoAdd CDC-ACM support for the CX93010-2x UCMxx USB Modem
Jean-Christian de Rivaz [Wed, 10 Oct 2012 12:49:02 +0000 (12:49 +0000)]
Add CDC-ACM support for the CX93010-2x UCMxx USB Modem

commit e7d491a19d3e3aac544070293891a2542ae0c565 upstream.

This USB V.92/V.32bis Controllered Modem have the USB vendor ID 0x0572
and device ID 0x1340. It need the NO_UNION_NORMAL quirk to be recognized.

Reference:
http://www.conexant.com/servlets/DownloadServlet/DSH-201723-005.pdf?docid=1725&revid=5
See idVendor and idProduct in table 6-1. Device Descriptors

Signed-off-by: Jean-Christian de Rivaz <jc@eclis.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: emu10k1: add chip details for E-mu 1010 PCIe card
Maxim Kachur [Wed, 17 Oct 2012 16:18:10 +0000 (18:18 +0200)]
ALSA: emu10k1: add chip details for E-mu 1010 PCIe card

commit 10f571d09106c3eb85951896522c9650596eff2e upstream.

Add chip details for E-mu 1010 PCIe card. It has the same
chip as found in E-mu 1010b but it uses different PCI id.

Signed-off-by: Maxim Kachur <mcdebugger@duganet.ru>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoALSA: ac97 - Fix missing NULL check in snd_ac97_cvol_new()
Takashi Iwai [Thu, 11 Oct 2012 14:43:40 +0000 (16:43 +0200)]
ALSA: ac97 - Fix missing NULL check in snd_ac97_cvol_new()

commit 733a48e5ae5bf28b046fad984d458c747cbb8c21 upstream.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=44721

Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agotpm: Propagate error from tpm_transmit to fix a timeout hang
Peter Huewe [Thu, 27 Sep 2012 14:09:33 +0000 (16:09 +0200)]
tpm: Propagate error from tpm_transmit to fix a timeout hang

commit abce9ac292e13da367bbd22c1f7669f988d931ac upstream.

tpm_write calls tpm_transmit without checking the return value and
assigns the return value unconditionally to chip->pending_data, even if
it's an error value.
This causes three bugs.

So if we write to /dev/tpm0 with a tpm_param_size bigger than
TPM_BUFSIZE=0x1000 (e.g. 0x100a)
and a bufsize also bigger than TPM_BUFSIZE (e.g. 0x100a)
tpm_transmit returns -E2BIG which is assigned to chip->pending_data as
-7, but tpm_write returns that TPM_BUFSIZE bytes have been successfully
been written to the TPM, altough this is not true (bug #1).

As we did write more than than TPM_BUFSIZE bytes but tpm_write reports
that only TPM_BUFSIZE bytes have been written the vfs tries to write
the remaining bytes (in this case 10 bytes) to the tpm device driver via
tpm_write which then blocks at

 /* cannot perform a write until the read has cleared
 either via tpm_read or a user_read_timer timeout */
 while (atomic_read(&chip->data_pending) != 0)
 msleep(TPM_TIMEOUT);

for 60 seconds, since data_pending is -7 and nobody is able to
read it (since tpm_read luckily checks if data_pending is greater than
0) (#bug 2).

After that the remaining bytes are written to the TPM which are
interpreted by the tpm as a normal command. (bug #3)
So if the last bytes of the command stream happen to be a e.g.
tpm_force_clear this gets accidentally sent to the TPM.

This patch fixes all three bugs, by propagating the error code of
tpm_write and returning -E2BIG if the input buffer is too big,
since the response from the tpm for a truncated value is bogus anyway.
Moreover it returns -EBUSY to userspace if there is a response ready to be
read.

Signed-off-by: Peter Huewe <peter.huewe@infineon.com>
Signed-off-by: Kent Yoder <key@linux.vnet.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agonetfilter: nf_conntrack: fix racy timer handling with reliable events
Pablo Neira Ayuso [Wed, 29 Aug 2012 16:25:49 +0000 (16:25 +0000)]
netfilter: nf_conntrack: fix racy timer handling with reliable events

commit 5b423f6a40a0327f9d40bc8b97ce9be266f74368 upstream.

Existing code assumes that del_timer returns true for alive conntrack
entries. However, this is not true if reliable events are enabled.
In that case, del_timer may return true for entries that were
just inserted in the dying list. Note that packets / ctnetlink may
hold references to conntrack entries that were just inserted to such
list.

This patch fixes the issue by adding an independent timer for
event delivery. This increases the size of the ecache extension.
Still we can revisit this later and use variable size extensions
to allocate this area on demand.

Tested-by: Oliver Smith <olipro@8.c.9.b.0.7.4.0.1.0.0.2.ip6.arpa>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoRevert "lockd: use rpc client's cl_nodename for id encoding"
Ben Hutchings [Sun, 28 Oct 2012 18:43:50 +0000 (18:43 +0000)]
Revert "lockd: use rpc client's cl_nodename for id encoding"

This reverts 5ff39e971c87ea9f4c4c7b253898abafa960e32b which was commit
303a7ce92064c285a04c870f2dc0192fdb2968cb upstream.

It is not necessary for kernel versions without per-netns RPC clients.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agofs/compat_ioctl.c: VIDEO_SET_SPU_PALETTE missing error check
Kees Cook [Thu, 25 Oct 2012 20:38:16 +0000 (13:38 -0700)]
fs/compat_ioctl.c: VIDEO_SET_SPU_PALETTE missing error check

commit 12176503366885edd542389eed3aaf94be163fdb upstream.

The compat ioctl for VIDEO_SET_SPU_PALETTE was missing an error check
while converting ioctl arguments.  This could lead to leaking kernel
stack contents into userspace.

Patch extracted from existing fix in grsecurity.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: David Miller <davem@davemloft.net>
Cc: Brad Spengler <spender@grsecurity.net>
Cc: PaX Team <pageexec@freemail.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agogen_init_cpio: avoid stack overflow when expanding
Kees Cook [Thu, 25 Oct 2012 20:38:14 +0000 (13:38 -0700)]
gen_init_cpio: avoid stack overflow when expanding

commit 20f1de659b77364d55d4e7fad2ef657e7730323f upstream.

Fix possible overflow of the buffer used for expanding environment
variables when building file list.

In the extremely unlikely case of an attacker having control over the
environment variables visible to gen_init_cpio, control over the
contents of the file gen_init_cpio parses, and gen_init_cpio was built
without compiler hardening, the attacker can gain arbitrary execution
control via a stack buffer overflow.

  $ cat usr/crash.list
  file foo ${BIG}${BIG}${BIG}${BIG}${BIG}${BIG} 0755 0 0
  $ BIG=$(perl -e 'print "A" x 4096;') ./usr/gen_init_cpio usr/crash.list
  *** buffer overflow detected ***: ./usr/gen_init_cpio terminated

This also replaces the space-indenting with tabs.

Patch based on existing fix extracted from grsecurity.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Michal Marek <mmarek@suse.cz>
Cc: Brad Spengler <spender@grsecurity.net>
Cc: PaX Team <pageexec@freemail.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agodrivers/rtc/rtc-imxdi.c: add missing spin lock initialization
Jan Luebbe [Thu, 25 Oct 2012 20:38:11 +0000 (13:38 -0700)]
drivers/rtc/rtc-imxdi.c: add missing spin lock initialization

commit fee0de7791f967c2c5f0d43eb7b7261761b45e64 upstream.

Signed-off-by: Jan Luebbe <jlu@pengutronix.de>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Roland Stigge <stigge@antcom.de>
Cc: Grant Likely <grant.likely@secretlab.ca>
Tested-by: Roland Stigge <stigge@antcom.de>
Cc: Sascha Hauer <kernel@pengutronix.de>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agogenalloc: stop crashing the system when destroying a pool
Thadeu Lima de Souza Cascardo [Thu, 25 Oct 2012 20:37:51 +0000 (13:37 -0700)]
genalloc: stop crashing the system when destroying a pool

commit eedce141cd2dad8d0cefc5468ef41898949a7031 upstream.

The genalloc code uses the bitmap API from include/linux/bitmap.h and
lib/bitmap.c, which is based on long values.  Both bitmap_set from
lib/bitmap.c and bitmap_set_ll, which is the lockless version from
genalloc.c, use BITMAP_LAST_WORD_MASK to set the first bits in a long in
the bitmap.

That one uses (1 << bits) - 1, 0b111, if you are setting the first three
bits.  This means that the API counts from the least significant bits
(LSB from now on) to the MSB.  The LSB in the first long is bit 0, then.
The same works for the lookup functions.

The genalloc code uses longs for the bitmap, as it should.  In
include/linux/genalloc.h, struct gen_pool_chunk has unsigned long
bits[0] as its last member.  When allocating the struct, genalloc should
reserve enough space for the bitmap.  This should be a proper number of
longs that can fit the amount of bits in the bitmap.

However, genalloc allocates an integer number of bytes that fit the
amount of bits, but may not be an integer amount of longs.  9 bytes, for
example, could be allocated for 70 bits.

This is a problem in itself if the Least Significat Bit in a long is in
the byte with the largest address, which happens in Big Endian machines.
This means genalloc is not allocating the byte in which it will try to
set or check for a bit.

This may end up in memory corruption, where genalloc will try to set the
bits it has not allocated.  In fact, genalloc may not set these bits
because it may find them already set, because they were not zeroed since
they were not allocated.  And that's what causes a BUG when
gen_pool_destroy is called and check for any set bits.

What really happens is that genalloc uses kmalloc_node with __GFP_ZERO
on gen_pool_add_virt.  With SLAB and SLUB, this means the whole slab
will be cleared, not only the requested bytes.  Since struct
gen_pool_chunk has a size that is a multiple of 8, and slab sizes are
multiples of 8, we get lucky and allocate and clear the right amount of
bytes.

Hower, this is not the case with SLOB or with older code that did memset
after allocating instead of using __GFP_ZERO.

So, a simple module as this (running 3.6.0), will cause a crash when
rmmod'ed.

  [root@phantom-lp2 foo]# cat foo.c
  #include <linux/kernel.h>
  #include <linux/module.h>
  #include <linux/init.h>
  #include <linux/genalloc.h>

  MODULE_LICENSE("GPL");
  MODULE_VERSION("0.1");

  static struct gen_pool *foo_pool;

  static __init int foo_init(void)
  {
          int ret;
          foo_pool = gen_pool_create(10, -1);
          if (!foo_pool)
                  return -ENOMEM;
          ret = gen_pool_add(foo_pool, 0xa0000000, 32 << 10, -1);
          if (ret) {
                  gen_pool_destroy(foo_pool);
                  return ret;
          }
          return 0;
  }

  static __exit void foo_exit(void)
  {
          gen_pool_destroy(foo_pool);
  }

  module_init(foo_init);
  module_exit(foo_exit);
  [root@phantom-lp2 foo]# zcat /proc/config.gz | grep SLOB
  CONFIG_SLOB=y
  [root@phantom-lp2 foo]# insmod ./foo.ko
  [root@phantom-lp2 foo]# rmmod foo
  ------------[ cut here ]------------
  kernel BUG at lib/genalloc.c:243!
  cpu 0x4: Vector: 700 (Program Check) at [c0000000bb0e7960]
      pc: c0000000003cb50c: .gen_pool_destroy+0xac/0x110
      lr: c0000000003cb4fc: .gen_pool_destroy+0x9c/0x110
      sp: c0000000bb0e7be0
     msr: 8000000000029032
    current = 0xc0000000bb0e0000
    paca    = 0xc000000006d30e00   softe: 0        irq_happened: 0x01
      pid   = 13044, comm = rmmod
  kernel BUG at lib/genalloc.c:243!
  [c0000000bb0e7ca0d000000004b00020 .foo_exit+0x20/0x38 [foo]
  [c0000000bb0e7d20c0000000000dff98 .SyS_delete_module+0x1a8/0x290
  [c0000000bb0e7e30c0000000000097d4 syscall_exit+0x0/0x94
  --- Exception: c00 (System Call) at 000000800753d1a0
  SP (fffd0b0e640) is in userspace

Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@linux.vnet.ibm.com>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: Benjamin Gaignard <benjamin.gaignard@stericsson.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agomm: fix XFS oops due to dirty pages without buffers on s390
Jan Kara [Thu, 25 Oct 2012 20:37:31 +0000 (13:37 -0700)]
mm: fix XFS oops due to dirty pages without buffers on s390

commit ef5d437f71afdf4afdbab99213add99f4b1318fd upstream.

On s390 any write to a page (even from kernel itself) sets architecture
specific page dirty bit.  Thus when a page is written to via buffered
write, HW dirty bit gets set and when we later map and unmap the page,
page_remove_rmap() finds the dirty bit and calls set_page_dirty().

Dirtying of a page which shouldn't be dirty can cause all sorts of
problems to filesystems.  The bug we observed in practice is that
buffers from the page get freed, so when the page gets later marked as
dirty and writeback writes it, XFS crashes due to an assertion
BUG_ON(!PagePrivate(page)) in page_buffers() called from
xfs_count_page_state().

Similar problem can also happen when zero_user_segment() call from
xfs_vm_writepage() (or block_write_full_page() for that matter) set the
hardware dirty bit during writeback, later buffers get freed, and then
page unmapped.

Fix the issue by ignoring s390 HW dirty bit for page cache pages of
mappings with mapping_cap_account_dirty().  This is safe because for
such mappings when a page gets marked as writeable in PTE it is also
marked dirty in do_wp_page() or do_page_fault().  When the dirty bit is
cleared by clear_page_dirty_for_io(), the page gets writeprotected in
page_mkclean().  So pagecache page is writeable if and only if it is
dirty.

Thanks to Hugh Dickins for pointing out mapping has to have
mapping_cap_account_dirty() for things to work and proposing a cleaned
up variant of the patch.

The patch has survived about two hours of running fsx-linux on tmpfs
while heavily swapping and several days of running on out build machines
where the original problem was triggered.

Signed-off-by: Jan Kara <jack@suse.cz>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[bwh: Backported to 3.2: adjust context; in particular there is no local
 'anon' in page_remove_rmap()]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agofreezer: exec should clear PF_NOFREEZE along with PF_KTHREAD
Oleg Nesterov [Thu, 25 Oct 2012 20:28:12 +0000 (22:28 +0200)]
freezer: exec should clear PF_NOFREEZE along with PF_KTHREAD

commit b40a79591ca918e7b91b0d9b6abd5d00f2e88c19 upstream.

flush_old_exec() clears PF_KTHREAD but forgets about PF_NOFREEZE.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
[bwh: Backported to 3.2: PF_FORKNOEXEC is cleared elsewhere]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoxhci: Fix potential NULL ptr deref in command cancellation.
Sarah Sharp [Tue, 16 Oct 2012 20:17:43 +0000 (13:17 -0700)]
xhci: Fix potential NULL ptr deref in command cancellation.

commit 43a09f7fb01fa1e091416a2aa49b6c666458c1ee upstream.

The command cancellation code doesn't check whether find_trb_seg()
couldn't find the segment that contains the TRB to be canceled.  This
could cause a NULL pointer deference later in the function when next_trb
is called.  It's unlikely to happen unless something is wrong with the
command ring pointers, so add some debugging in case it happens.

This patch should be backported to stable kernels as old as 3.0, that
contain the commit b63f4053cc8aa22a98e3f9a97845afe6c15d0a0d "xHCI:
handle command after aborting the command ring".

Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoehci: Add yet-another Lucid nohandoff pci quirk
Anisse Astier [Tue, 9 Oct 2012 10:22:37 +0000 (12:22 +0200)]
ehci: Add yet-another Lucid nohandoff pci quirk

commit 8daf8b6086f9d575200cd0aa3797e26137255609 upstream.

Board name changed on another shipping Lucid tablet.

Signed-off-by: Anisse Astier <anisse@astier.eu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoehci: fix Lucid nohandoff pci quirk to be more generic with BIOS versions
Anisse Astier [Tue, 9 Oct 2012 10:22:36 +0000 (12:22 +0200)]
ehci: fix Lucid nohandoff pci quirk to be more generic with BIOS versions

commit c323dc023b9501e5d09582ec7efd1d40a9001d99 upstream.

BIOS vendors keep changing the BIOS versions. Only match the beginning
of the string to match all Lucid tablets with board name M11JB.

Signed-off-by: Anisse Astier <anisse@astier.eu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: mos7840: remove invalid disconnect handling
Johan Hovold [Thu, 25 Oct 2012 16:56:33 +0000 (18:56 +0200)]
USB: mos7840: remove invalid disconnect handling

commit e681b66f2e19fadbe8a7e2a17900978cb6bc921f upstream.

Remove private zombie flag used to signal disconnect and to prevent
control urb from being submitted from interrupt urb completion handler.

The control urb will not be re-submitted as both the control urb and the
interrupt urb is killed on disconnect.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: mos7840: remove NULL-urb submission
Johan Hovold [Thu, 25 Oct 2012 16:56:32 +0000 (18:56 +0200)]
USB: mos7840: remove NULL-urb submission

commit 28c3ae9a8cf45f439c9a0779ebd0256e2ae72813 upstream.

The private int_urb is never allocated so the submission from the
control completion handler will always fail. Remove this odd piece of
broken code.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: mos7840: fix port-device leak in error path
Johan Hovold [Thu, 25 Oct 2012 11:35:10 +0000 (13:35 +0200)]
USB: mos7840: fix port-device leak in error path

commit 3eb55cc4ed88eee3b5230f66abcdbd2a91639eda upstream.

The driver set the usb-serial port pointers to NULL on errors in attach,
effectively preventing usb-serial core from decrementing the port ref
counters and releasing the port devices and associated data.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: mos7840: fix urb leak at release
Johan Hovold [Thu, 25 Oct 2012 11:35:09 +0000 (13:35 +0200)]
USB: mos7840: fix urb leak at release

commit 65a4cdbb170e4ec1a7fa0e94936d47e24a17b0e8 upstream.

Make sure control urb is freed at release.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: sierra: fix memory leak in probe error path
Johan Hovold [Thu, 25 Oct 2012 08:29:18 +0000 (10:29 +0200)]
USB: sierra: fix memory leak in probe error path

commit 084817d79399ab5ccab2f90a148b0369912a8369 upstream.

Move interface data allocation to attach so that it is deallocated on
errors in usb-serial probe.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: sierra: fix memory leak in attach error path
Johan Hovold [Thu, 25 Oct 2012 08:29:17 +0000 (10:29 +0200)]
USB: sierra: fix memory leak in attach error path

commit 7e41f9bcdd2e813ea2a3c40db291d87ea06b559f upstream.

Make sure port private data is deallocated on errors in attach.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: mct_u232: fix broken close
Johan Hovold [Thu, 25 Oct 2012 08:29:14 +0000 (10:29 +0200)]
USB: mct_u232: fix broken close

commit 5260e458f5eff269a43e4f1e9c47186c57b88ddb upstream.

Make sure generic close is called at close.

The driver relies on the generic write implementation but did not call
generic close.

Note that the call to kill the read urb is not redundant, as mct_u232
uses an interrupt urb from the second port as the read urb and that
generic close therefore fails to kill it.

Compile-only tested.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: opticon: fix memory leak in error path
Johan Hovold [Thu, 25 Oct 2012 08:29:12 +0000 (10:29 +0200)]
USB: opticon: fix memory leak in error path

commit acbf0e5263de563e25f7c104868e4490b9e72b13 upstream.

Fix memory leak in write error path.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: opticon: fix DMA from stack
Johan Hovold [Thu, 25 Oct 2012 08:29:11 +0000 (10:29 +0200)]
USB: opticon: fix DMA from stack

commit ea0dbebffe118724cd4df7d9b071ea8ee48d48f0 upstream.

Make sure to allocate the control-message buffer dynamically as some
platforms cannot do DMA from stack.

Note that only the first byte of the old buffer was used.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: whiteheat: fix memory leak in error path
Johan Hovold [Thu, 25 Oct 2012 08:29:01 +0000 (10:29 +0200)]
USB: whiteheat: fix memory leak in error path

commit c129197c99550d356cf5f69b046994dd53cd1b9d upstream.

Make sure command buffer is deallocated in case of errors during attach.

Cc: <support@connecttech.com>
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agovhost: fix mergeable bufs on BE hosts
Michael S. Tsirkin [Wed, 24 Oct 2012 18:37:51 +0000 (20:37 +0200)]
vhost: fix mergeable bufs on BE hosts

commit 910a578f7e9400a78a3b13aba0b4d2df16a2cb05 upstream.

We copy head count to a 16 bit field, this works by chance on LE but on
BE guest gets 0. Fix it up.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Alexander Graf <agraf@suse.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agosysfs: sysfs_pathname/sysfs_add_one: Use strlcat() instead of strcat()
Geert Uytterhoeven [Sat, 29 Sep 2012 20:23:19 +0000 (22:23 +0200)]
sysfs: sysfs_pathname/sysfs_add_one: Use strlcat() instead of strcat()

commit 66081a72517a131430dcf986775f3268aafcb546 upstream.

The warning check for duplicate sysfs entries can cause a buffer overflow
when printing the warning, as strcat() doesn't check buffer sizes.
Use strlcat() instead.

Since strlcat() doesn't return a pointer to the passed buffer, unlike
strcat(), I had to convert the nested concatenation in sysfs_add_one() to
an admittedly more obscure comma operator construct, to avoid emitting code
for the concatenation if CONFIG_BUG is disabled.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoDrivers: hv: Cleanup error handling in vmbus_open()
K. Y. Srinivasan [Fri, 12 Oct 2012 20:22:42 +0000 (13:22 -0700)]
Drivers: hv: Cleanup error handling in vmbus_open()

commit 1392550240aaa72ce3a094a38bd23525cd67ce60 upstream.

Fix a memory leak  in the error handling path in the function vmbus_open().

Signed-off-by: K. Y. Srinivasan <kys@microsoft.com>
Reviewed-by: Haiyang Zhang <haiyangz@microsoft.com>
Reported-by: Jason Wang <jasowang@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agousb-storage: add unusual_devs entry for Casio EX-N1 digital camera
Michael Shigorin [Mon, 22 Oct 2012 09:18:56 +0000 (12:18 +0300)]
usb-storage: add unusual_devs entry for Casio EX-N1 digital camera

commit d7870af7e2e3a91b462075ec1ca669b482215187 upstream.

This commit sets removable subclass for Casio EX-N1 digital camera.

The patch has been tested within an ALT Linux kernel:
http://git.altlinux.org/people/led/packages/?p=kernel-image-3.0.git;a=commitdiff;h=c0fd891836e89fe0c93a4d536a59216d90e4e3e7

See also https://bugzilla.kernel.org/show_bug.cgi?id=49221

Signed-off-by: Oleksandr Chumachenko <ledest@gmail.com>
Signed-off-by: Michael Shigorin <mike@osdn.org.ua>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoUSB: serial: Fix memory leak in sierra_release()
Lennart Sorensen [Wed, 24 Oct 2012 14:23:09 +0000 (10:23 -0400)]
USB: serial: Fix memory leak in sierra_release()

commit f7bc5051667b74c3861f79eed98c60d5c3b883f7 upstream.

I found a memory leak in sierra_release() (well sierra_probe() I guess)
that looses 8 bytes each time the driver releases a device.

Signed-off-by: Len Sorensen <lsorense@csclub.uwaterloo.ca>
Acked-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoRevert "ath9k_hw: Updated AR9003 tx gain table for 5GHz"
Felix Fietkau [Wed, 17 Oct 2012 11:50:43 +0000 (13:50 +0200)]
Revert "ath9k_hw: Updated AR9003 tx gain table for 5GHz"

commit 73b26df5fa1a6245d6fc982362518b620bc7c2fe upstream.

This reverts commit a240dc7b3c7463bd60cf0a9b2a90f52f78aae0fd.

This commit is reducing tx power by at least 10 db on some devices,
e.g. the Buffalo WZR-HP-G450H.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Cc: rmanohar@qca.qualcomm.com
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoSUNRPC: Get rid of the xs_error_report socket callback
Trond Myklebust [Mon, 22 Oct 2012 21:14:36 +0000 (17:14 -0400)]
SUNRPC: Get rid of the xs_error_report socket callback

commit f878b657ce8e7d3673afe48110ec208a29e38c4a upstream.

Chris Perl reports that we're seeing races between the wakeup call in
xs_error_report and the connect attempts. Basically, Chris has shown
that in certain circumstances, the call to xs_error_report causes the
rpc_task that is responsible for reconnecting to wake up early, thus
triggering a disconnect and retry.

Since the sk->sk_error_report() calls in the socket layer are always
followed by a tcp_done() in the cases where we care about waking up
the rpc_tasks, just let the state_change callbacks take responsibility
for those wake ups.

Reported-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Tested-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoSUNRPC: Prevent races in xs_abort_connection()
Trond Myklebust [Tue, 23 Oct 2012 21:50:07 +0000 (17:50 -0400)]
SUNRPC: Prevent races in xs_abort_connection()

commit 4bc1e68ed6a8b59be8a79eb719be515a55c7bc68 upstream.

The call to xprt_disconnect_done() that is triggered by a successful
connection reset will trigger another automatic wakeup of all tasks
on the xprt->pending rpc_wait_queue. In particular it will cause an
early wake up of the task that called xprt_connect().

All we really want to do here is clear all the socket-specific state
flags, so we split that functionality out of xs_sock_mark_closed()
into a helper that can be called by xs_abort_connection()

Reported-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Tested-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoRevert "SUNRPC: Ensure we close the socket on EPIPE errors too..."
Trond Myklebust [Tue, 23 Oct 2012 15:40:02 +0000 (11:40 -0400)]
Revert "SUNRPC: Ensure we close the socket on EPIPE errors too..."

commit b9d2bb2ee537424a7f855e1f93eed44eb9ee0854 upstream.

This reverts commit 55420c24a0d4d1fce70ca713f84aa00b6b74a70e.
Now that we clear the connected flag when entering TCP_CLOSE_WAIT,
the deadlock described in this commit is no longer possible.
Instead, the resulting call to xs_tcp_shutdown() can interfere
with pending reconnection attempts.

Reported-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Tested-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoSUNRPC: Clear the connect flag when socket state is TCP_CLOSE_WAIT
Trond Myklebust [Tue, 23 Oct 2012 15:35:47 +0000 (11:35 -0400)]
SUNRPC: Clear the connect flag when socket state is TCP_CLOSE_WAIT

commit d0bea455dd48da1ecbd04fedf00eb89437455fdc upstream.

This is needed to ensure that we call xprt_connect() upon the next
call to call_connect().

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Tested-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoamd64_edac:__amd64_set_scrub_rate(): avoid overindexing scrubrates[]
Andrew Morton [Tue, 23 Oct 2012 21:09:39 +0000 (14:09 -0700)]
amd64_edac:__amd64_set_scrub_rate(): avoid overindexing scrubrates[]

commit 168bfeef7bba3f9784f7540b053e4ac72b769ce9 upstream.

If none of the elements in scrubrates[] matches, this loop will cause
__amd64_set_scrub_rate() to incorrectly use the n+1th element.

As the function is designed to use the final scrubrates[] element in the
case of no match, we can fix this bug by simply terminating the array
search at the n-1th element.

Boris: this code is fragile anyway, see here why:
http://marc.info/?l=linux-kernel&m=135102834131236&w=2

It will be rewritten more robustly soonish.

Reported-by: Denis Kirjanov <kirjanov@gmail.com>
Cc: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoarch/tile: avoid generating .eh_frame information in modules
Chris Metcalf [Fri, 19 Oct 2012 15:43:11 +0000 (11:43 -0400)]
arch/tile: avoid generating .eh_frame information in modules

commit 627072b06c362bbe7dc256f618aaa63351f0cfe6 upstream.

The tile tool chain uses the .eh_frame information for backtracing.
The vmlinux build drops any .eh_frame sections at link time, but when
present in kernel modules, it causes a module load failure due to the
presence of unsupported pc-relative relocations.  When compiling to
use compiler feedback support, the compiler by default omits .eh_frame
information, so we don't see this problem.  But when not using feedback,
we need to explicitly suppress the .eh_frame.

Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agocpufreq / powernow-k8: Remove usage of smp_processor_id() in preemptible code
Andreas Herrmann [Mon, 22 Oct 2012 22:55:10 +0000 (00:55 +0200)]
cpufreq / powernow-k8: Remove usage of smp_processor_id() in preemptible code

commit e4df1cbcc1f329e53a1fff7450b2229e0addff20 upstream.

Commit 6889125b8b4e09c5e53e6ecab3433bed1ce198c9
(cpufreq/powernow-k8: workqueue user shouldn't migrate the kworker to another CPU)
causes powernow-k8 to trigger a preempt warning, e.g.:

  BUG: using smp_processor_id() in preemptible [00000000] code: cpufreq/3776
  caller is powernowk8_target+0x20/0x49
  Pid: 3776, comm: cpufreq Not tainted 3.6.0 #9
  Call Trace:
   [<ffffffff8125b447>] debug_smp_processor_id+0xc7/0xe0
   [<ffffffff814877e7>] powernowk8_target+0x20/0x49
   [<ffffffff81482b02>] __cpufreq_driver_target+0x82/0x8a
   [<ffffffff81484fc6>] cpufreq_governor_performance+0x4e/0x54
   [<ffffffff81482c50>] __cpufreq_governor+0x8c/0xc9
   [<ffffffff81482e6f>] __cpufreq_set_policy+0x1a9/0x21e
   [<ffffffff814839af>] store_scaling_governor+0x16f/0x19b
   [<ffffffff81484f16>] ? cpufreq_update_policy+0x124/0x124
   [<ffffffff8162b4a5>] ? _raw_spin_unlock_irqrestore+0x2c/0x49
   [<ffffffff81483640>] store+0x60/0x88
   [<ffffffff811708c0>] sysfs_write_file+0xf4/0x130
   [<ffffffff8111243b>] vfs_write+0xb5/0x151
   [<ffffffff811126e0>] sys_write+0x4a/0x71
   [<ffffffff816319a9>] system_call_fastpath+0x16/0x1b

Fix this by by always using work_on_cpu().

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agousb hub: send clear_tt_buffer_complete events when canceling TT clear work
Octavian Purdila [Mon, 1 Oct 2012 19:21:12 +0000 (22:21 +0300)]
usb hub: send clear_tt_buffer_complete events when canceling TT clear work

commit 3b6054da68f9b0d5ed6a7ed0f42a79e61904352c upstream.

There is a race condition in the USB hub code with regard to handling
TT clear requests that can get the HCD driver in a deadlock. Usually
when an TT clear request is scheduled it will be executed immediately:

<7>[    6.077583] usb 2-1.3: unlink qh1-0e01/f4d4db00 start 0 [1/2 us]
<3>[    6.078041] usb 2-1: clear tt buffer port 3, a3 ep2 t04048d82
<7>[    6.078299] hub_tt_work:731
<7>[    9.309089] usb 2-1.5: link qh1-0e01/f4d506c0 start 0 [1/2 us]
<7>[    9.324526] ehci_hcd 0000:00:1d.0: reused qh f4d4db00 schedule
<7>[    9.324539] usb 2-1.3: link qh1-0e01/f4d4db00 start 0 [1/2 us]
<7>[    9.341530] usb 1-1.1: link qh4-0e01/f397aec0 start 2 [1/2 us]
<7>[   10.116159] usb 2-1.3: unlink qh1-0e01/f4d4db00 start 0 [1/2 us]
<3>[   10.116459] usb 2-1: clear tt buffer port 3, a3 ep2 t04048d82
<7>[   10.116537] hub_tt_work:731

However, if a suspend operation is triggered before hub_tt_work is
scheduled, hub_quiesce will cancel the work without notifying the HCD
driver:

<3>[   35.033941] usb 2-1: clear tt buffer port 3, a3 ep2 t04048d80
<5>[   35.034022] sd 0:0:0:0: [sda] Stopping disk
<7>[   35.034039] hub 2-1:1.0: hub_suspend
<7>[   35.034067] usb 2-1: unlink qh256-0001/f3b1ab00 start 1 [1/0 us]
<7>[   35.035085] hub 1-0:1.0: hub_suspend
<7>[   35.035102] usb usb1: bus suspend, wakeup 0
<7>[   35.035106] ehci_hcd 0000:00:1a.0: suspend root hub
<7>[   35.035298] hub 2-0:1.0: hub_suspend
<7>[   35.035313] usb usb2: bus suspend, wakeup 0
<7>[   35.035315] ehci_hcd 0000:00:1d.0: suspend root hub
<6>[   35.250017] PM: suspend of devices complete after 216.979 msecs
<6>[   35.250822] PM: late suspend of devices complete after 0.799 msecs
<7>[   35.252343] ehci_hcd 0000:00:1d.0: wakeup: 1
<7>[   35.262923] ehci_hcd 0000:00:1d.0: --> PCI D3hot
<7>[   35.263302] ehci_hcd 0000:00:1a.0: wakeup: 1
<7>[   35.273912] ehci_hcd 0000:00:1a.0: --> PCI D3hot
<6>[   35.274254] PM: noirq suspend of devices complete after 23.442 msecs
<6>[   35.274975] ACPI: Preparing to enter system sleep state S3
<6>[   35.292666] PM: Saving platform NVS memory
<7>[   35.295030] Disabling non-boot CPUs ...
<6>[   35.297351] CPU 1 is now offline
<6>[   35.300345] CPU 2 is now offline
<6>[   35.303929] CPU 3 is now offline
<7>[   35.303931] lockdep: fixing up alternatives.
<6>[   35.304825] Extended CMOS year: 2000

When the device will resume the EHCI driver will get stuck in
ehci_endpoint_disable waiting for the tt_clearing flag to reset:

<0>[   47.610967] usb 2-1.3: **** DPM device timeout ****
<7>[   47.610972]  f2f11c60 00000092 f2f11c0c c10624a5 00000003 f4c6e880 c1c8a4c0 c1c8a4c0
<7>[   47.610983]  15c55698 0000000b f56b34c0 f2a45b70 f4c6e880 00000082 f2a4602c f2f11c30
<7>[   47.610993]  c10787f8 f4cac000 f2a45b70 00000000 f4cac010 f2f11c58 00000046 00000001
<7>[   47.611004] Call Trace:
<7>[   47.611006]  [<c10624a5>] ? sched_clock_cpu+0xf5/0x160
<7>[   47.611019]  [<c10787f8>] ? lock_release_holdtime.part.22+0x88/0xf0
<7>[   47.611026]  [<c103ed46>] ? lock_timer_base.isra.35+0x26/0x50
<7>[   47.611034]  [<c17592d3>] ? schedule_timeout+0x133/0x290
<7>[   47.611044]  [<c175b43e>] schedule+0x1e/0x50
<7>[   47.611051]  [<c17592d8>] schedule_timeout+0x138/0x290
<7>[   47.611057]  [<c10624a5>] ? sched_clock_cpu+0xf5/0x160
<7>[   47.611063]  [<c103e560>] ? usleep_range+0x40/0x40
<7>[   47.611070]  [<c1759445>] schedule_timeout_uninterruptible+0x15/0x20
<7>[   47.611077]  [<c14935f4>] ehci_endpoint_disable+0x64/0x160
<7>[   47.611084]  [<c147d1ee>] ? usb_hcd_flush_endpoint+0x10e/0x1d0
<7>[   47.611092]  [<c1165663>] ? sysfs_add_file+0x13/0x20
<7>[   47.611100]  [<c147d5a9>] usb_hcd_disable_endpoint+0x29/0x40
<7>[   47.611107]  [<c147fafc>] usb_disable_endpoint+0x5c/0x80
<7>[   47.611111]  [<c147fb57>] usb_disable_interface+0x37/0x50
<7>[   47.611116]  [<c1477650>] usb_reset_and_verify_device+0x4b0/0x640
<7>[   47.611122]  [<c1474665>] ? hub_port_status+0xb5/0x100
<7>[   47.611129]  [<c147a975>] usb_port_resume+0xd5/0x220
<7>[   47.611136]  [<c148877f>] generic_resume+0xf/0x30
<7>[   47.611142]  [<c14821a3>] usb_resume+0x133/0x180
<7>[   47.611147]  [<c1473b10>] ? usb_dev_thaw+0x10/0x10
<7>[   47.611152]  [<c1473b1d>] usb_dev_resume+0xd/0x10
<7>[   47.611157]  [<c13baa60>] dpm_run_callback+0x40/0xb0
<7>[   47.611164]  [<c13bdb03>] ? pm_runtime_enable+0x43/0x70
<7>[   47.611171]  [<c13bafc6>] device_resume+0x1a6/0x2c0
<7>[   47.611177]  [<c13ba940>] ? dpm_show_time+0xe0/0xe0
<7>[   47.611183]  [<c13bb0f9>] async_resume+0x19/0x40
<7>[   47.611189]  [<c10580c4>] async_run_entry_fn+0x64/0x160
<7>[   47.611196]  [<c104a244>] ? process_one_work+0x104/0x480
<7>[   47.611203]  [<c104a24c>] ? process_one_work+0x10c/0x480
<7>[   47.611209]  [<c104a2c0>] process_one_work+0x180/0x480
<7>[   47.611215]  [<c104a244>] ? process_one_work+0x104/0x480
<7>[   47.611220]  [<c1058060>] ? async_schedule+0x10/0x10
<7>[   47.611226]  [<c104c15c>] worker_thread+0x11c/0x2f0
<7>[   47.611233]  [<c104c040>] ? manage_workers.isra.27+0x1f0/0x1f0
<7>[   47.611239]  [<c10507f8>] kthread+0x78/0x80
<7>[   47.611244]  [<c1750000>] ? timer_cpu_notify+0xd6/0x20d
<7>[   47.611253]  [<c1050780>] ? __init_kthread_worker+0x60/0x60
<7>[   47.611258]  [<c176357e>] kernel_thread_helper+0x6/0xd
<7>[   47.611283] ------------[ cut here ]------------

This patch changes hub_quiesce behavior to flush the TT clear work
instead of canceling it, to make sure that no TT clear request remains
uncompleted before suspend.

Signed-off-by: Octavian Purdila <octavian.purdila@intel.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
11 years agoARM: 7559/1: smp: switch away from the idmap before updating init_mm.mm_count
Will Deacon [Fri, 19 Oct 2012 16:53:01 +0000 (17:53 +0100)]
ARM: 7559/1: smp: switch away from the idmap before updating init_mm.mm_count

commit 5f40b909728ad784eb43aa309d3c4e9bdf050781 upstream.

When booting a secondary CPU, the primary CPU hands two sets of page
tables via the secondary_data struct:

(1) swapper_pg_dir: a normal, cacheable, shared (if SMP) mapping
    of the kernel image (i.e. the tables used by init_mm).

(2) idmap_pgd: an uncached mapping of the .idmap.text ELF
    section.

The idmap is generally used when enabling and disabling the MMU, which
includes early CPU boot. In this case, the secondary CPU switches to
swapper as soon as it enters C code:

struct mm_struct *mm = &init_mm;
unsigned int cpu = smp_processor_id();

/*
 * All kernel threads share the same mm context; grab a
 * reference and switch to it.
 */
atomic_inc(&mm->mm_count);
current->active_mm = mm;
cpumask_set_cpu(cpu, mm_cpumask(mm));
cpu_switch_mm(mm->pgd, mm);

This causes a problem on ARMv7, where the identity mapping is treated as
strongly-ordered leading to architecturally UNPREDICTABLE behaviour of
exclusive accesses, such as those used by atomic_inc.

This patch re-orders the secondary_start_kernel function so that we
switch to swapper before performing any exclusive accesses.

Cc: David McKay <david.mckay@st.com>
Reported-by: Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>