8 years agoNLS: improve UTF8 -> UTF16 string conversion routine
Alan Stern [Thu, 17 Nov 2011 21:42:19 +0000 (16:42 -0500)]
NLS: improve UTF8 -> UTF16 string conversion routine

commit 0720a06a7518c9d0c0125bd5d1f3b6264c55c3dd upstream.

The utf8s_to_utf16s conversion routine needs to be improved.  Unlike
its utf16s_to_utf8s sibling, it doesn't accept arguments specifying
the maximum length of the output buffer or the endianness of its
16-bit output.

This patch (as1501) adds the two missing arguments, and adjusts the
only two places in the kernel where the function is called.  A
follow-on patch will add a third caller that does utilize the new

The two conversion routines are still annoyingly inconsistent in the
way they handle invalid byte combinations.  But that's a subject for a
different patch.

Signed-off-by: Alan Stern <>
CC: Clemens Ladisch <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agobtrfs: use rcu_barrier() to wait for bdev puts at unmount
Eric Sandeen [Sat, 9 Mar 2013 15:18:39 +0000 (15:18 +0000)]
btrfs: use rcu_barrier() to wait for bdev puts at unmount

commit bc178622d40d87e75abc131007342429c9b03351 upstream.

Doing this would reliably fail with -EBUSY for me:

# mount /dev/sdb2 /mnt/scratch; umount /mnt/scratch; mkfs.btrfs -f /dev/sdb2
unable to open /dev/sdb2: Device or resource busy

because mkfs.btrfs tries to open the device O_EXCL, and somebody still has it.

Using systemtap to track bdev gets & puts shows a kworker thread doing a
blkdev put after mkfs attempts a get; this is left over from the unmount

call_rcu(&device->rcu, free_device);
INIT_WORK(&device->rcu_work, __free_device);

so unmount might complete before __free_device fires & does its blkdev_put.

Adding an rcu_barrier() to btrfs_close_devices() causes unmount to wait
until all blkdev_put()s are done, and the device is truly free once
unmount completes.

Signed-off-by: Eric Sandeen <>
Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
Signed-off-by: Ben Hutchings <>
8 years agoloopdev: remove an user triggerable oops
Guo Chao [Thu, 21 Feb 2013 23:16:49 +0000 (15:16 -0800)]
loopdev: remove an user triggerable oops

commit b1a6650406875b9097a032eed89af50682fe1160 upstream.

When loopdev is built as module and we pass an invalid parameter,
loop_init() will return directly without deregister misc device, which
will cause an oops when insert loop module next time because we left some
garbage in the misc device list.

Test case:
sudo modprobe loop max_part=1024
(failed due to invalid parameter)
sudo modprobe loop

Clean up nicely to avoid such oops.

Signed-off-by: Guo Chao <>
Cc: Alexander Viro <>
Cc: Guo Chao <>
Cc: M. Hindess <>
Cc: Nikanth Karthikesan <>
Cc: Jens Axboe <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Jens Axboe <>
Signed-off-by: Ben Hutchings <>
8 years agoloopdev: fix a deadlock
Guo Chao [Thu, 21 Feb 2013 23:16:45 +0000 (15:16 -0800)]
loopdev: fix a deadlock

commit 5370019dc2d2c2ff90e95d181468071362934f3a upstream.

bd_mutex and lo_ctl_mutex can be held in different order.

Path #1:

  __blkdev_get (hold bd_mutex)
   lo_open (hold lo_ctl_mutex)

Path #2:

 lo_ioctl (hold lo_ctl_mutex)
  lo_set_capacity (hold bd_mutex)

Lockdep does not report it, because path #2 actually holds a subclass of
lo_ctl_mutex.  This subclass seems creep into the code by mistake.  The
patch author actually just mentioned it in the changelog, see commit
f028f3b2 ("loop: fix circular locking in loop_clr_fd()"), also see:

Path #2 hold bd_mutex to call bd_set_size(), I've protected it
with i_mutex in a previous patch, so drop bd_mutex at this site.

Signed-off-by: Guo Chao <>
Cc: Alexander Viro <>
Cc: Guo Chao <>
Cc: M. Hindess <>
Cc: Nikanth Karthikesan <>
Cc: Jens Axboe <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Jens Axboe <>
Signed-off-by: Ben Hutchings <>
8 years agoblock: use i_size_write() in bd_set_size()
Guo Chao [Thu, 21 Feb 2013 23:16:42 +0000 (15:16 -0800)]
block: use i_size_write() in bd_set_size()

commit d646a02a9d44d1421f273ae3923d97b47b918176 upstream.

blkdev_ioctl(GETBLKSIZE) uses i_size_read() to read size of block device.
If we update block size directly, reader may see intermediate result in
some machines and configurations.  Use i_size_write() instead.

Signed-off-by: Guo Chao <>
Cc: Alexander Viro <>
Cc: Guo Chao <>
Cc: M. Hindess <>
Cc: Nikanth Karthikesan <>
Cc: Jens Axboe <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Jens Axboe <>
Signed-off-by: Ben Hutchings <>
8 years agoxen-netfront: delay gARP until backend switches to Connected
Laszlo Ersek [Sun, 11 Dec 2011 01:48:59 +0000 (01:48 +0000)]
xen-netfront: delay gARP until backend switches to Connected

commit 08e34eb14fe4cfd934b5c169a7682a969457c4ea upstream.

After a guest is live migrated, the xen-netfront driver emits a gratuitous
ARP message, so that networking hardware on the target host's subnet can
take notice, and public routing to the guest is re-established. However,
if the packet appears on the backend interface before the backend is added
to the target host's bridge, the packet is lost, and the migrated guest's
peers become unable to talk to the guest.

A sufficient two-parts condition to prevent the above is:

(1) ensure that the backend only moves to Connected xenbus state after its
hotplug scripts completed, ie. the netback interface got added to the
bridge; and

(2) ensure the frontend only queues the gARP when it sees the backend move
to Connected.

These two together provide complete ordering. Sub-condition (1) is already
satisfied by commit f942dc2552b8 in Linus' tree, based on commit
6b0b80ca7165 from [1].

In general, the full condition is sufficient, not necessary, because,
according to [2], live migration has been working for a long time without
satisfying sub-condition (2). However, after 6b0b80ca7165 was backported
to the RHEL-5 host to ensure (1), (2) still proved necessary in the RHEL-6
guest. This patch intends to provide (2) for upstream.

The Reviewed-by line comes from [3].

[1] git://

Signed-off-by: Laszlo Ersek <>
Reviewed-by: Ian Campbell <>
Signed-off-by: David S. Miller <>
Signed-off-by: Ben Hutchings <>
8 years agomm/hotplug: correctly add new zone to all other nodes' zone lists
Jiang Liu [Tue, 31 Jul 2012 23:43:30 +0000 (16:43 -0700)]
mm/hotplug: correctly add new zone to all other nodes' zone lists

commit 08dff7b7d629807dbb1f398c68dd9cd58dd657a1 upstream.

When online_pages() is called to add new memory to an empty zone, it
rebuilds all zone lists by calling build_all_zonelists().  But there's a
bug which prevents the new zone to be added to other nodes' zone lists.

online_pages() {
node_set_state(zone_to_nid(zone), N_HIGH_MEMORY)

Here the node of the zone is put into N_HIGH_MEMORY state after calling
build_all_zonelists(), but build_all_zonelists() only adds zones from
nodes in N_HIGH_MEMORY state to the fallback zone lists.

->for_each_node_state(n, N_HIGH_MEMORY)

So memory in the new zone will never be used by other nodes, and it may
cause strange behavor when system is under memory pressure.  So put node
into N_HIGH_MEMORY state before calling build_all_zonelists().

Signed-off-by: Jianguo Wu <>
Signed-off-by: Jiang Liu <>
Cc: Mel Gorman <>
Cc: Michal Hocko <>
Cc: Minchan Kim <>
Cc: Rusty Russell <>
Cc: Yinghai Lu <>
Cc: Tony Luck <>
Cc: KAMEZAWA Hiroyuki <>
Cc: KOSAKI Motohiro <>
Cc: David Rientjes <>
Cc: Keping Chen <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agobatman-adv: Only write requested number of byte to user buffer
Sven Eckelmann [Sat, 10 Dec 2011 14:28:36 +0000 (15:28 +0100)]
batman-adv: Only write requested number of byte to user buffer

commit b5a1eeef04cc7859f34dec9b72ea1b28e4aba07c upstream.

Don't write more than the requested number of bytes of an batman-adv icmp
packet to the userspace buffer. Otherwise unrelated userspace memory might get
overridden by the kernel.

Signed-off-by: Sven Eckelmann <>
Signed-off-by: Marek Lindner <>
Signed-off-by: Ben Hutchings <>
8 years agobatman-adv: bat_socket_read missing checks
Paul Kot [Sat, 10 Dec 2011 14:28:34 +0000 (15:28 +0100)]
batman-adv: bat_socket_read missing checks

commit c00b6856fc642b234895cfabd15b289e76726430 upstream.

Writing a icmp_packet_rr and then reading icmp_packet can lead to kernel
memory corruption, if __user *buf is just below TASK_SIZE.

Signed-off-by: Paul Kot <>
[ made it checkpatch clean]
Signed-off-by: Sven Eckelmann <>
Signed-off-by: Marek Lindner <>
Signed-off-by: Ben Hutchings <>
8 years agoUSB: Fix connected device switch to Inactive state.
Sarah Sharp [Wed, 13 Mar 2013 18:10:50 +0000 (11:10 -0700)]
USB: Fix connected device switch to Inactive state.

commit d3b9d7a9051d7024a93c76a84b2f84b3b66ad6d5 upstream.

A USB 3.0 device can transition to the Inactive state if a U1 or U2 exit
transition fails.  The current code in hub_events simply issues a warm
reset, but does not call any pre-reset or post-reset driver methods (or
unbind/rebind drivers without them).  Therefore the drivers won't know
their device has just been reset.

hub_events should instead call usb_reset_device.  This means
hub_port_reset now needs to figure out whether it should issue a warm
reset or a hot reset.

Remove the FIXME note about needing disconnect() for a NOTATTACHED
device.  This patch fixes that.

Signed-off-by: Sarah Sharp <>
Acked-by: Alan Stern <>
Signed-off-by: Ben Hutchings <>
8 years agoUSB: Rip out recursive call on warm port reset.
Sarah Sharp [Wed, 13 Mar 2013 18:10:36 +0000 (11:10 -0700)]
USB: Rip out recursive call on warm port reset.

commit a24a6078754f28528bc91e7e7b3e6ae86bd936d8 upstream.

When a hot reset fails on a USB 3.0 port, the current port reset code
recursively calls hub_port_reset inside hub_port_wait_reset.  This isn't
ideal, since we should avoid recursive calls in the kernel, and it also
doesn't allow us to issue multiple warm resets on reset failures.

Rip out the recursive call.  Instead, add code to hub_port_reset to
issue a warm reset if the hot reset fails, and try multiple warm resets
before giving up on the port.

In hub_port_wait_reset, remove the recursive call and re-indent.  The
code is basically the same, except:

1. It bails out early if the port has transitioned to Inactive or
Compliance Mode after the reset completed.

2. It doesn't consider a connect status change to be a failed reset.  If
multiple warm resets needed to be issued, the connect status may have
changed, so we need to ignore that and look at the port link state
instead.  hub_port_reset will now do that.

3. It unconditionally sets udev->speed on all types of successful
resets.  The old recursive code would set the port speed when the second
hub_port_reset returned.

The old code did not handle connected devices needing a warm reset well.
There were only two situations that the old code handled correctly: an
empty port needing a warm reset, and a hot reset that migrated to a warm

When an empty port needed a warm reset, hub_port_reset was called with
the warm variable set.  The code in hub_port_finish_reset would skip
telling the USB core and the xHC host that the device was reset, because
otherwise that would result in a NULL pointer dereference.

When a USB 3.0 device reset migrated to a warm reset, the recursive call
made the call stack look like this:

hub_port_reset(warm = false)
        hub_wait_port_reset(warm = false)
                hub_port_reset(warm = true)
                        hub_wait_port_reset(warm = true)
                        hub_port_finish_reset(warm = true)
                        (return up the call stack to the first wait)

        hub_port_finish_reset(warm = false)

The old code didn't want to notify the USB core or the xHC host of device reset
twice, so it only did it in the second call to hub_port_finish_reset,
when warm was set to false.  This was necessary because
before patch two ("USB: Ignore xHCI Reset Device status."), the USB core
would pay attention to the xHC Reset Device command error status, and
the second call would always fail.

Now that we no longer have the recursive call, and warm can change from
false to true in hub_port_reset, we need to have hub_port_finish_reset
unconditionally notify the USB core and the xHC of the device reset.

In hub_port_finish_reset, unconditionally clear the connect status
change (CSC) bit for USB 3.0 hubs when the port reset is done.  If we
had to issue multiple warm resets for a device, that bit may have been
set if the device went into SS.Inactive and then was successfully warm

Signed-off-by: Sarah Sharp <>
Acked-by: Alan Stern <>
Signed-off-by: Ben Hutchings <>
8 years agoUSB: Prepare for refactoring by adding extra udev checks.
Sarah Sharp [Fri, 8 Mar 2013 00:24:47 +0000 (16:24 -0800)]
USB: Prepare for refactoring by adding extra udev checks.

commit 2d4fa940f99663c82ba55b2244638833b388e4e2 upstream.

The next patch will refactor the hub port code to rip out the recursive
call to hub_port_reset on a failed hot reset.  In preparation for that,
make sure all code paths can deal with being called with a NULL udev.
The usb_device will not be valid if warm reset was issued because a port
transitioned to the Inactive or Compliance Mode on a device connect.

This patch should have no effect on current behavior.

Signed-off-by: Sarah Sharp <>
Acked-by: Alan Stern <>
Signed-off-by: Ben Hutchings <>
8 years agoUSB: Don't use EHCI port sempahore for USB 3.0 hubs.
Sarah Sharp [Fri, 8 Mar 2013 00:24:45 +0000 (16:24 -0800)]
USB: Don't use EHCI port sempahore for USB 3.0 hubs.

commit 0fe51aa5eee51db7c7ecd201d42a977ad79c58b6 upstream.

The EHCI host controller needs to prevent EHCI initialization when the
UHCI or OHCI companion controller is in the middle of a port reset.  It
uses ehci_cf_port_reset_rwsem to do this.  USB 3.0 hubs can't be under
an EHCI host controller, so it makes no sense to down the semaphore for
USB 3.0 hubs.  It also makes the warm port reset code more complex.

Don't down ehci_cf_port_reset_rwsem for USB 3.0 hubs.

Signed-off-by: Sarah Sharp <>
Acked-by: Alan Stern <>
Signed-off-by: Ben Hutchings <>
8 years agoFix: compat_rw_copy_check_uvector() misuse in aio, readv, writev, and security keys
Mathieu Desnoyers [Mon, 25 Feb 2013 15:20:36 +0000 (10:20 -0500)]
Fix: compat_rw_copy_check_uvector() misuse in aio, readv, writev, and security keys

commit 8aec0f5d4137532de14e6554fd5dd201ff3a3c49 upstream.

Looking at mm/process_vm_access.c:process_vm_rw() and comparing it to
compat_process_vm_rw() shows that the compatibility code requires an
explicit "access_ok()" check before calling
compat_rw_copy_check_uvector(). The same difference seems to appear when
we compare fs/read_write.c:do_readv_writev() to

This subtle difference between the compat and non-compat requirements
should probably be debated, as it seems to be error-prone. In fact,
there are two others sites that use this function in the Linux kernel,
and they both seem to get it wrong:

Now shifting our attention to fs/aio.c, we see that aio_setup_iocb()
also ends up calling compat_rw_copy_check_uvector() through
aio_setup_vectored_rw(). Unfortunately, the access_ok() check appears to
be missing. Same situation for

I propose that we add the access_ok() check directly into
compat_rw_copy_check_uvector(), so callers don't have to worry about it,
and it therefore makes the compat call code similar to its non-compat
counterpart. Place the access_ok() check in the same location where
copy_from_user() can trigger a -EFAULT error in the non-compat code, so
the ABI behaviors are alike on both compat and non-compat.

While we are here, fix compat_do_readv_writev() so it checks for
compat_rw_copy_check_uvector() negative return values.

And also, fix a memory leak in compat_keyctl_instantiate_key_iov() error

Acked-by: Linus Torvalds <>
Acked-by: Al Viro <>
Signed-off-by: Mathieu Desnoyers <>
Signed-off-by: Linus Torvalds <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agocrypto: user - fix info leaks in report API
Mathias Krause [Tue, 5 Feb 2013 17:19:13 +0000 (18:19 +0100)]
crypto: user - fix info leaks in report API

commit 9a5467bf7b6e9e02ec9c3da4e23747c05faeaac6 upstream.

Three errors resulting in kernel memory disclosure:

1/ The structures used for the netlink based crypto algorithm report API
are located on the stack. As snprintf() does not fill the remainder of
the buffer with null bytes, those stack bytes will be disclosed to users
of the API. Switch to strncpy() to fix this.

2/ crypto_report_one() does not initialize all field of struct
crypto_user_alg. Fix this to fix the heap info leak.

3/ For the module name we should copy only as many bytes as
module_name() returns -- not as much as the destination buffer could
hold. But the current code does not and therefore copies random data
from behind the end of the module name, as the module name is always
shorter than CRYPTO_MAX_ALG_NAME.

Also switch to use strncpy() to copy the algorithm's name and
driver_name. They are strings, after all.

Signed-off-by: Mathias Krause <>
Cc: Steffen Klassert <>
Signed-off-by: Herbert Xu <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agohwmon: (pmbus/ltc2978) Fix temperature reporting
Guenter Roeck [Thu, 21 Feb 2013 18:27:54 +0000 (10:27 -0800)]
hwmon: (pmbus/ltc2978) Fix temperature reporting

commit 8c958c703ef8804093437959221951eaf0e1e664 upstream.

On LTC2978, only READ_TEMPERATURE is supported. It reports
the internal junction temperature. This register is unpaged.

READ_TEMPERATURE is paged and reports external temperatures.
READ_TEMPERATURE2 is unpaged and reports the internal junction

Signed-off-by: Guenter Roeck <>
Acked-by: Jean Delvare <>
Signed-off-by: Ben Hutchings <>
8 years agohwmon: (lineage-pem) Add missing terminating entry for pem_[input|fan]_attributes
Axel Lin [Thu, 14 Mar 2013 08:27:18 +0000 (16:27 +0800)]
hwmon: (lineage-pem) Add missing terminating entry for pem_[input|fan]_attributes

commit df069079c153d22adf6c28dcc0b1cf62bba75167 upstream.

Signed-off-by: Axel Lin <>
Acked-by: Jean Delvare <>
Signed-off-by: Guenter Roeck <>
Signed-off-by: Ben Hutchings <>
8 years agosignal: always clear sa_restorer on execve
Kees Cook [Wed, 13 Mar 2013 21:59:33 +0000 (14:59 -0700)]
signal: always clear sa_restorer on execve

commit 2ca39528c01a933f6689cd6505ce65bd6d68a530 upstream.

When the new signal handlers are set up, the location of sa_restorer is
not cleared, leaking a parent process's address space location to
children.  This allows for a potential bypass of the parent's ASLR by
examining the sa_restorer value returned when calling sigaction().

Based on what should be considered "secret" about addresses, it only
matters across the exec not the fork (since the VMAs haven't changed
until the exec).  But since exec sets SIG_DFL and keeps sa_restorer,
this is where it should be fixed.

Given the few uses of sa_restorer, a "set" function was not written
since this would be the only use.  Instead, we use
__ARCH_HAS_SA_RESTORER, as already done in other places.

Example of the leak before applying this patch:

  $ cat /proc/$$/maps
  7fb9f3083000-7fb9f3238000 r-xp 00000000 fd:01 404469 .../
  $ ./leak
  7f278bc74000-7f278be29000 r-xp 00000000 fd:01 404469 .../
  1 0 (nil) 0x7fb9f30b94a0
  2 4000000 (nil) 0x7f278bcaa4a0
  3 4000000 (nil) 0x7f278bcaa4a0
  4 0 (nil) 0x7fb9f30b94a0

[ use SA_RESTORER for backportability]
Signed-off-by: Kees Cook <>
Reported-by: Emese Revfy <>
Cc: Emese Revfy <>
Cc: PaX Team <>
Cc: Al Viro <>
Cc: Oleg Nesterov <>
Cc: "Eric W. Biederman" <>
Cc: Serge Hallyn <>
Cc: Julien Tinnes <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
8 years agoUSB: cdc-wdm: fix buffer overflow
Oliver Neukum [Tue, 12 Mar 2013 13:52:42 +0000 (14:52 +0100)]
USB: cdc-wdm: fix buffer overflow

commit c0f5ecee4e741667b2493c742b60b6218d40b3aa upstream.

The buffer for responses must not overflow.
If this would happen, set a flag, drop the data and return
an error after user space has read all remaining data.

Signed-off-by: Oliver Neukum <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agow1: fix oops when w1_search is called from netlink connector
Marcin Jurkowski [Sat, 2 Mar 2013 13:50:15 +0000 (14:50 +0100)]
w1: fix oops when w1_search is called from netlink connector

commit 9d1817cab2f030f6af360e961cc69bb1da8ad765 upstream.

On Sat, Mar 02, 2013 at 10:45:10AM +0100, Sven Geggus wrote:
> This is the bad commit I found doing git bisect:
04f482faf50535229a5a5c8d629cf963899f857c is the first bad commit
> commit 04f482faf50535229a5a5c8d629cf963899f857c
> Author: Patrick McHardy <>
> Date:   Mon Mar 28 08:39:36 2011 +0000

Good job. I was too lazy to bisect for bad commit;)

Reading the code I found problematic kthread_should_stop call from netlink
connector which causes the oops. After applying a patch, I've been testing
owfs+w1 setup for nearly two days and it seems to work very reliable (no
hangs, no memleaks etc).
More detailed description and possible fix is given below:

Function w1_search can be called from either kthread or netlink callback.
While the former works fine, the latter causes oops due to kthread_should_stop

This patch adds a check if w1_search is serving netlink command, skipping
kthread_should_stop invocation if so.

Signed-off-by: Marcin Jurkowski <>
Acked-by: Evgeniy Polyakov <>
Cc: Josh Boyer <>
Tested-by: Sven Geggus <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agousb: serial: Add Rigblaster Advantage to device table
Steve Conklin [Thu, 7 Mar 2013 23:19:33 +0000 (17:19 -0600)]
usb: serial: Add Rigblaster Advantage to device table

commit a57e82a18779ab8a5e5a1f5841cef937cf578913 upstream.

The Rigblaster Advantage is an amateur radio interface sold by West Mountain
Radio. It contains a cp210x serial interface but the device ID is not in
the driver.

Signed-off-by: Steve Conklin <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agotty: serial: fix typo "ARCH_S5P6450"
Paul Bolle [Sat, 9 Mar 2013 22:16:44 +0000 (23:16 +0100)]
tty: serial: fix typo "ARCH_S5P6450"

commit 827aa0d36d486f359808c8fb931cf7a71011a09d upstream.

This could have been either ARCH_S5P64X0 or CPU_S5P6450. Looking at
commit 2555e663b367b8d555e76023f4de3f6338c28d6c ("ARM: S5P64X0: Add UART
serial support for S5P6450") - which added this typo - makes clear this
should be CPU_S5P6450.

Signed-off-by: Paul Bolle <>
Acked-by: Kukjin Kim <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agoserial: 8250_pci: add support for another kind of NetMos Technology PCI 9835 Multi...
Wang YanQing [Fri, 1 Mar 2013 03:47:20 +0000 (11:47 +0800)]
serial: 8250_pci: add support for another kind of NetMos Technology PCI 9835 Multi-I/O Controller

commit 8d2f8cd424ca0b99001f3ff4f5db87c4e525f366 upstream.

01:08.0 Communication controller: NetMos Technology PCI 9835 Multi-I/O Controller (rev 01)
Subsystem: Device [1000:0012]
Control: I/O+ Mem+ BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
Status: Cap- 66MHz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Interrupt: pin A routed to IRQ 20
Region 0: I/O ports at e050 [size=8]
Region 1: I/O ports at e040 [size=8]
Region 2: I/O ports at e030 [size=8]
Region 3: I/O ports at e020 [size=8]
Region 4: I/O ports at e010 [size=8]
Region 5: I/O ports at e000 [size=16]

Signed-off-by: Wang YanQing <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <>
8 years agoFix 4 port and add support for 8 port 'Unknown' PCI serial port cards
Scott Ashcroft [Sun, 3 Mar 2013 21:35:06 +0000 (21:35 +0000)]
Fix 4 port and add support for 8 port 'Unknown' PCI serial port cards

commit d13402a4a944e72612a9ec5c9190e35717c02a9d upstream.

I've managed to find an 8 port version of the card 4 port card which was discussed here:

Looking back at that thread there were two issues in the original patch.

1) The I/O ports for the UARTs are within BAR2 not BAR0. This can been seen in the original post.
2) A serial quirk isn't needed as these cards have no memory in BAR0 which makes pci_plx9050_init just return.

This patch fixes the 4 port support to use BAR2, removes the bogus quirk and adds support for the 8 port card.

$ lspci -vvv -n -s 00:08.0
00:08.0 0780: 10b5:9050 (rev 01)
Subsystem: 10b5:1588
Control: I/O+ Mem- BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B+ ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
Interrupt: pin A routed to IRQ 17
Region 1: I/O ports at ff00 [size=128]
Region 2: I/O ports at fe00 [size=64]
Region 3: I/O ports at fd00 [size=8]
Capabilities: <access denied>
Kernel driver in use: serial

$ dmesg | grep 0000:00:08.0:
[    0.083320] pci 0000:00:08.0: [10b5:9050] type 0 class 0x000780
[    0.083355] pci 0000:00:08.0: reg 14: [io  0xff00-0xff7f]
[    0.083369] pci 0000:00:08.0: reg 18: [io  0xfe00-0xfe3f]
[    0.083382] pci 0000:00:08.0: reg 1c: [io  0xfd00-0xfd07]
[    0.083460] pci 0000:00:08.0: PME# supported from D0 D3hot
[    1.212867] 0000:00:08.0: ttyS4 at I/O 0xfe00 (irq = 17) is a 16550A
[    1.233073] 0000:00:08.0: ttyS5 at I/O 0xfe08 (irq = 17) is a 16550A
[    1.253270] 0000:00:08.0: ttyS6 at I/O 0xfe10 (irq = 17) is a 16550A
[    1.273468] 0000:00:08.0: ttyS7 at I/O 0xfe18 (irq = 17) is a 16550A
[    1.293666] 0000:00:08.0: ttyS8 at I/O 0xfe20 (irq = 17) is a 16550A
[    1.313863] 0000:00:08.0: ttyS9 at I/O 0xfe28 (irq = 17) is a 16550A
[    1.334061] 0000:00:08.0: ttyS10 at I/O 0xfe30 (irq = 17) is a 16550A
[    1.354258] 0000:00:08.0: ttyS11 at I/O 0xfe38 (irq = 17) is a 16550A

Signed-off-by: Scott Ashcroft <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: adjust filename, context]
Signed-off-by: Ben Hutchings <>
8 years agotty/serial: Add support for Altera serial port
Ley Foon Tan [Thu, 7 Mar 2013 02:28:37 +0000 (10:28 +0800)]
tty/serial: Add support for Altera serial port

commit e06c93cacb82dd147266fd1bdb2d0a0bd45ff2c1 upstream.

Add support for Altera 8250/16550 compatible serial port.

Signed-off-by: Ley Foon Tan <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Backported to 3.2: adjust filenames, context]
Signed-off-by: Ben Hutchings <>
8 years agokeys: fix race with concurrent install_user_keyrings()
David Howells [Tue, 12 Mar 2013 05:44:31 +0000 (16:44 +1100)]
keys: fix race with concurrent install_user_keyrings()

commit 0da9dfdd2cd9889201bc6f6f43580c99165cd087 upstream.

This fixes CVE-2013-1792.

There is a race in install_user_keyrings() that can cause a NULL pointer
dereference when called concurrently for the same user if the uid and
uid-session keyrings are not yet created.  It might be possible for an
unprivileged user to trigger this by calling keyctl() from userspace in
parallel immediately after logging in.

Assume that we have two threads both executing lookup_user_key(), both

=============================== ===============================
==>call install_user_keyrings();
if (!cred->user->session_keyring)
==>call install_user_keyrings()
user->uid_keyring = uid_keyring;
if (user->uid_keyring)
return 0;
key = cred->user->session_keyring [== NULL]
user->session_keyring = session_keyring;
atomic_inc(&key->usage); [oops]

At the point thread A dereferences cred->user->session_keyring, thread B
hasn't updated user->session_keyring yet, but thread A assumes it is
populated because install_user_keyrings() returned ok.

The race window is really small but can be exploited if, for example,
thread B is interrupted or preempted after initializing uid_keyring, but
before doing setting session_keyring.

This couldn't be reproduced on a stock kernel.  However, after placing
systemtap probe on 'user->session_keyring = session_keyring;' that
introduced some delay, the kernel could be crashed reliably.

Fix this by checking both pointers before deciding whether to return.
Alternatively, the test could be done away with entirely as it is checked
inside the mutex - but since the mutex is global, that may not be the best

Signed-off-by: David Howells <>
Reported-by: Mateusz Guzik <>
Signed-off-by: Andrew Morton <>
Signed-off-by: James Morris <>
Signed-off-by: Ben Hutchings <>
8 years agoext3: Fix format string issues
Lars-Peter Clausen [Sat, 9 Mar 2013 14:28:44 +0000 (15:28 +0100)]
ext3: Fix format string issues

commit 8d0c2d10dd72c5292eda7a06231056a4c972e4cc upstream.

ext3_msg() takes the printk prefix as the second parameter and the
format string as the third parameter. Two callers of ext3_msg omit the
prefix and pass the format string as the second parameter and the first
parameter to the format string as the third parameter. In both cases
this string comes from an arbitrary source. Which means the string may
contain format string characters, which will
lead to undefined and potentially harmful behavior.

The issue was introduced in commit 4cf46b67eb("ext3: Unify log messages
in ext3") and is fixed by this patch.

Signed-off-by: Lars-Peter Clausen <>
Signed-off-by: Jan Kara <>
Signed-off-by: Ben Hutchings <>
8 years agoqcaux: add Franklin U600
Dan Williams [Tue, 19 Feb 2013 15:47:09 +0000 (09:47 -0600)]
qcaux: add Franklin U600

commit 2d90e63603ac235aecd7d20e234616e0682c8b1f upstream.

4 ports; AT/PPP is standard CDC-ACM.  The other three (added by this
patch) are QCDM/DIAG, possibly GPS, and unknown.

Signed-off-by: Dan Williams <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agostaging: vt6656: Fix oops on resume from suspend.
Malcolm Priestley [Mon, 18 Feb 2013 19:54:18 +0000 (19:54 +0000)]
staging: vt6656: Fix oops on resume from suspend.

commit 6987a6dabfc40222ef767f67b57212fe3a0225fb upstream.

Remove usb_put_dev from vt6656_suspend and usb_get_dev
from vt6566_resume.

These are not normally in suspend/resume functions.

Signed-off-by: Malcolm Priestley <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agousb: cp210x new Vendor/Device IDs
Matwey V. Kornilov [Sat, 9 Mar 2013 09:57:32 +0000 (13:57 +0400)]
usb: cp210x new Vendor/Device IDs

commit be3101c23394af59694c8a2aae6d07f5da62fea5 upstream.

This patch adds support for the Lake Shore Cryotronics devices to
the CP210x driver.

These lines are ported from cp210x driver distributed by Lake Shore web site:
and licensed under the terms of GPLv2.

Moreover, I've tested this changes with Lake Shore 335 in my labs.

Signed-off-by: Matwey V. Kornilov <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agoALSA: seq: Fix missing error handling in snd_seq_timer_open()
Takashi Iwai [Fri, 8 Mar 2013 17:11:17 +0000 (18:11 +0100)]
ALSA: seq: Fix missing error handling in snd_seq_timer_open()

commit 66efdc71d95887b652a742a5dae51fa834d71465 upstream.

snd_seq_timer_open() didn't catch the whole error path but let through
if the timer id is a slave.  This may lead to Oops by accessing the
uninitialized pointer.

 BUG: unable to handle kernel NULL pointer dereference at 00000000000002ae
 IP: [<ffffffff819b3477>] snd_seq_timer_open+0xe7/0x130
 PGD 785cd067 PUD 76964067 PMD 0
 Oops: 0002 [#4] SMP
 CPU 0
 Pid: 4288, comm: trinity-child7 Tainted: G      D W 3.9.0-rc1+ #100 Bochs Bochs
 RIP: 0010:[<ffffffff819b3477>]  [<ffffffff819b3477>] snd_seq_timer_open+0xe7/0x130
 RSP: 0018:ffff88006ece7d38  EFLAGS: 00010246
 RAX: 0000000000000286 RBX: ffff88007851b400 RCX: 0000000000000000
 RDX: 000000000000ffff RSI: ffff88006ece7d58 RDI: ffff88006ece7d38
 RBP: ffff88006ece7d98 R08: 000000000000000a R09: 000000000000fffe
 R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
 R13: ffff8800792c5400 R14: 0000000000e8f000 R15: 0000000000000007
 FS:  00007f7aaa650700(0000) GS:ffff88007f800000(0000) GS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 00000000000002ae CR3: 000000006efec000 CR4: 00000000000006f0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
 Process trinity-child7 (pid: 4288, threadinfo ffff88006ece6000, task ffff880076a8a290)
  0000000000000286 ffffffff828f2be0 ffff88006ece7d58 ffffffff810f354d
  65636e6575716573 2065756575712072 ffff8800792c0030 0000000000000000
  ffff88006ece7d98 ffff8800792c5400 ffff88007851b400 ffff8800792c5520
 Call Trace:
  [<ffffffff810f354d>] ? trace_hardirqs_on+0xd/0x10
  [<ffffffff819b17e9>] snd_seq_queue_timer_open+0x29/0x70
  [<ffffffff819ae01a>] snd_seq_ioctl_set_queue_timer+0xda/0x120
  [<ffffffff819acb9b>] snd_seq_do_ioctl+0x9b/0xd0
  [<ffffffff819acbe0>] snd_seq_ioctl+0x10/0x20
  [<ffffffff811b9542>] do_vfs_ioctl+0x522/0x570
  [<ffffffff8130a4b3>] ? file_has_perm+0x83/0xa0
  [<ffffffff810f354d>] ? trace_hardirqs_on+0xd/0x10
  [<ffffffff811b95ed>] sys_ioctl+0x5d/0xa0
  [<ffffffff813663fe>] ? trace_hardirqs_on_thunk+0x3a/0x3f
  [<ffffffff81faed69>] system_call_fastpath+0x16/0x1b

Reported-and-tested-by: Tommi Rantala <>
Signed-off-by: Takashi Iwai <>
Signed-off-by: Ben Hutchings <>
8 years agovirtio: rng: disallow multiple device registrations, fixes crashes
Amit Shah [Fri, 8 Mar 2013 00:30:18 +0000 (11:30 +1100)]
virtio: rng: disallow multiple device registrations, fixes crashes

commit e84e7a56a3aa2963db506299e29a5f3f09377f9b upstream.

The code currently only supports one virtio-rng device at a time.
Invoking guests with multiple devices causes the guest to blow up.

Check if we've already registered and initialised the driver.  Also
cleanup in case of registration errors or hot-unplug so that a new
device can be used.

Reported-by: Peter Krempa <>
Reported-by: <>
Signed-off-by: Amit Shah <>
Signed-off-by: Rusty Russell <>
Signed-off-by: Ben Hutchings <>
8 years agoe1000e: fix pci-device enable-counter balance
Konstantin Khlebnikov [Tue, 5 Mar 2013 09:42:59 +0000 (09:42 +0000)]
e1000e: fix pci-device enable-counter balance

commit 4e0855dff094b0d56d6b5b271e0ce7851cc1e063 upstream.

This patch removes redundant and unbalanced pci_disable_device() from
__e1000_shutdown(). pci_clear_master() is enough, device can go into
suspended state with elevated enable_cnt.

Bug was introduced in commit 23606cf5d1192c2b17912cb2ef6e62f9b11de133
("e1000e / PCI / PM: Add basic runtime PM support (rev. 4)") in v2.6.35

Cc: Bruce Allan <>
Signed-off-by: Konstantin Khlebnikov <>
Acked-by: Rafael J. Wysocki <>
Tested-by: Borislav Petkov <>
Tested-by: Aaron Brown <>
Signed-off-by: Jeff Kirsher <>
Signed-off-by: Ben Hutchings <>
8 years agoUSB: added support for Cinterion's products AH6 and PLS8
Christian Schmiedl [Wed, 6 Mar 2013 16:08:50 +0000 (17:08 +0100)]
USB: added support for Cinterion's products AH6 and PLS8

commit 1941138e1c024ecb5bd797d414928d3eb94d8662 upstream.

add support for Cinterion's products AH6 and PLS8 by adding Product IDs
and USB_DEVICE tuples.

Signed-off-by: Christian Schmiedl <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agoUSB: storage: fix Huawei mode switching regression
Bjørn Mork [Mon, 4 Mar 2013 13:19:21 +0000 (14:19 +0100)]
USB: storage: fix Huawei mode switching regression

commit ab4b71644a26d1ab92b987b2fd30e17c25e89f85 upstream.

This reverts commit 200e0d99 ("USB: storage: optimize to match the
Huawei USB storage devices and support new switch command" and the
followup bugfix commit cd060956 ("USB: storage: properly handle
the endian issues of idProduct").

The commit effectively added a large number of Huawei devices to
the deprecated usb-storage mode switching logic.  Many of these
devices have been in use and supported by the userspace
usb_modeswitch utility for years.  Forcing the switching inside
the kernel causes a number of regressions as a result of ignoring
existing onfigurations, and also completely takes away the ability
to configure mode switching per device/system/user.

Known regressions caused by this:
 - Some of the devices support multiple modes, using different
  switching commands.  There are existing configurations taking
  advantage of this.

 - There is a real use case for disabling mode switching and
  instead mounting the exposed storage device. This becomes
  impossible with switching logic inside the usb-storage driver.

 - At least on device fail as a result of the usb-storage switching
  command, becoming completely unswitchable. This is possibly a
  firmware bug, but still a regression because the device work as
  expected using usb_modeswitch defaults.

In-kernel mode switching was deprecated years ago with the
development of the more user friendly userspace alternatives. The
existing list of devices in usb-storage was only kept to prevent
breaking already working systems.  The long term plan is to remove
the list, not to add to it. Ref:

Cc: <>
Signed-off-by: Bjørn Mork <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agoUSB: option: add Huawei E5331
Bjørn Mork [Wed, 27 Feb 2013 14:52:56 +0000 (15:52 +0100)]
USB: option: add Huawei E5331

commit daec90e7382cbd0e73eb6861109b3da91e5ab1f3 upstream.

Another device using CDC ACM with vendor specific protocol to mark
serial functions.

Signed-off-by: Bjørn Mork <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agoxen/pciback: Don't disable a PCI device that is already disabled.
Konrad Rzeszutek Wilk [Tue, 5 Mar 2013 18:14:19 +0000 (13:14 -0500)]
xen/pciback: Don't disable a PCI device that is already disabled.

commit bdc5c1812cea6efe1aaefb3131fcba28cd0b2b68 upstream.

While shuting down a HVM guest with pci devices passed through we
get this:

pciback 0000:04:00.0: restoring config space at offset 0x4 (was 0x100000, writing 0x100002)
------------[ cut here ]------------
WARNING: at drivers/pci/pci.c:1397 pci_disable_device+0x88/0xa0()
Hardware name: MS-7640
Device pciback
disabling already-disabled device
Modules linked in:
Pid: 53, comm: xenwatch Not tainted 3.9.0-rc1-20130304a+ #1
Call Trace:
 [<ffffffff8106994a>] warn_slowpath_common+0x7a/0xc0
 [<ffffffff81069a31>] warn_slowpath_fmt+0x41/0x50
 [<ffffffff813cf288>] pci_disable_device+0x88/0xa0
 [<ffffffff814554a7>] xen_pcibk_reset_device+0x37/0xd0
 [<ffffffff81454b6f>] ? pcistub_put_pci_dev+0x6f/0x120
 [<ffffffff81454b8d>] pcistub_put_pci_dev+0x8d/0x120
 [<ffffffff814582a9>] __xen_pcibk_release_devices+0x59/0xa0

This fixes the bug.

Reported-and-Tested-by: Sander Eikelenboom <>
Signed-off-by: Konrad Rzeszutek Wilk <>
Signed-off-by: Ben Hutchings <>
8 years agovfs: fix pipe counter breakage
Al Viro [Tue, 12 Mar 2013 02:59:49 +0000 (02:59 +0000)]
vfs: fix pipe counter breakage

commit a930d8790552658140d7d0d2e316af4f0d76a512 upstream.

If you open a pipe for neither read nor write, the pipe code will not
add any usage counters to the pipe, causing the 'struct pipe_inode_info"
to be potentially released early.

That doesn't normally matter, since you cannot actually use the pipe,
but the pipe release code - particularly fasync handling - still expects
the actual pipe infrastructure to all be there.  And rather than adding
NULL pointer checks, let's just disallow this case, the same way we
already do for the named pipe ("fifo") case.

This is ancient going back to pre-2.4 days, and until trinity, nobody
naver noticed.

Reported-by: Dave Jones <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
8 years agoFix memory leak in cpufreq stats.
Tu, Xiaobing [Mon, 22 Oct 2012 23:03:00 +0000 (01:03 +0200)]
Fix memory leak in cpufreq stats.

commit e37736777254ce1abc85493a5cacbefe5983b896 upstream.

When system enters sleep, non-boot CPUs will be disabled.
Cpufreq stats sysfs is created when the CPU is up, but it is not
freed when the CPU is going down. This will cause memory leak.

Signed-off-by: xiaobing tu <>
Signed-off-by: guifang tang <>
Signed-off-by: Rafael J. Wysocki <>
Signed-off-by: Ben Hutchings <>
8 years agodecnet: Fix disappearing sysctl entries
Eric W. Biederman [Tue, 12 Mar 2013 00:41:28 +0000 (17:41 -0700)]
decnet: Fix disappearing sysctl entries

When decnet is built as a module a simple:
echo 0.0 >/proc/sys/net/decnet/node_address

results in most of the sysctl entries under /proc/sys/net/decnet and
/proc/sys/net/decnet/conf disappearing.

For more details see

This change applies the same workaround used in
net/core/sysctl_net_core.c and net/ipv6/sysctl_net_ipv6.c of creating
a skeleton of decnet sysctl entries before doing anything else.

The problem first appeared in kernel 2.6.27.  The later rewrite of
sysctl in kernel 3.4 restored the previous behavior and eliminated the
need for this workaround.

This patch was heavily inspired by a similar but more complex patch by
Larry Baker.

Reported-by: Larry Baker <>
Signed-off-by: "Eric W. Biederman" <>
Signed-off-by: Ben Hutchings <>
8 years agoftrace: Update the kconfig for DYNAMIC_FTRACE
Steven Rostedt [Thu, 28 Feb 2013 02:48:09 +0000 (21:48 -0500)]
ftrace: Update the kconfig for DYNAMIC_FTRACE

commit db05021d49a994ee40a9735d9c3cb0060c9babb8 upstream.

The prompt to enable DYNAMIC_FTRACE (the ability to nop and
enable function tracing at run time) had a confusing statement:

 "enable/disable ftrace tracepoints dynamically"

This was written before tracepoints were added to the kernel,
but now that tracepoints have been added, this is very confusing
and has confused people enough to give wrong information during

Not only that, I looked at the help text, and it still references
that dreaded daemon that use to wake up once a second to update
the nop locations and brick NICs, that hasn't been around for over
five years.

Time to bring the text up to the current decade.

Reported-by: Ezequiel Garcia <>
Signed-off-by: Steven Rostedt <>
Signed-off-by: Ben Hutchings <>
8 years agoefi: be more paranoid about available space when creating variables
Josh Boyer [Mon, 11 Mar 2013 21:48:53 +0000 (17:48 -0400)]
efi: be more paranoid about available space when creating variables

commit 68d929862e29a8b52a7f2f2f86a0600423b093cd upstream.

UEFI variables are typically stored in flash. For various reasons, avaiable
space is typically not reclaimed immediately upon the deletion of a
variable - instead, the system will garbage collect during initialisation
after a reboot.

Some systems appear to handle this garbage collection extremely poorly,
failing if more than 50% of the system flash is in use. This can result in
the machine refusing to boot. The safest thing to do for the moment is to
forbid writes if they'd end up using more than half of the storage space.
We can make this more finegrained later if we come up with a method for
identifying the broken machines.

Signed-off-by: Matthew Garrett <>
Signed-off-by: Matt Fleming <>
[bwh: Backported to 3.2:
 - Drop efivarfs changes and unused check_var_size()
 - Add error codes to include/linux/efi.h, added upstream by
   commit 5d9db883761a ('efi: Add support for a UEFI variable filesystem')
 - Add efi_status_to_err(), added upstream by commit 7253eaba7b17
   ('efivarfs: Return an error if we fail to read a variable')]
Signed-off-by: Ben Hutchings <>
8 years agoefivars: Disable external interrupt while holding efivars->lock
Josh Boyer [Mon, 11 Mar 2013 21:47:42 +0000 (17:47 -0400)]
efivars: Disable external interrupt while holding efivars->lock

commit 81fa4e581d9283f7992a0d8c534bb141eb840a14 upstream.

There is a scenario which efi_pstore fails to log messages in a panic case.

 - CPUA holds an efi_var->lock in either efivarfs parts
   or efi_pstore with interrupt enabled.
 - CPUB panics and sends IPI to CPUA in smp_send_stop().
 - CPUA stops with holding the lock.
 - CPUB kicks efi_pstore_write() via kmsg_dump(KSMG_DUMP_PANIC)
   but it returns without logging messages.

[Patch Description]
This patch disables an external interruption while holding efivars->lock
as follows.

In efi_pstore_write() and get_var_data(), spin_lock/spin_unlock is
replaced by spin_lock_irqsave/spin_unlock_irqrestore because they may
be called in an interrupt context.

In other functions, they are replaced by spin_lock_irq/spin_unlock_irq.
because they are all called from a process context.

By applying this patch, we can avoid the problem above with
a following senario.

 - CPUA holds an efi_var->lock with interrupt disabled.
 - CPUB panics and sends IPI to CPUA in smp_send_stop().
 - CPUA receives the IPI after releasing the lock because it is
   disabling interrupt while holding the lock.
 - CPUB waits for one sec until CPUA releases the lock.
 - CPUB kicks efi_pstore_write() via kmsg_dump(KSMG_DUMP_PANIC)
   And it can hold the lock successfully.

Signed-off-by: Seiji Aguchi <>
Acked-by: Mike Waychison <>
Acked-by: Matt Fleming <>
Signed-off-by: Tony Luck <>
[bwh: Backported to 3.2:
 - Drop efivarfs changes
 - Adjust context
 - Drop change to efi_pstore_erase(), which is implemented using
   efi_pstore_write() here]
Signed-off-by: Ben Hutchings <>
8 years agoefi_pstore: Check remaining space with QueryVariableInfo() before writing data
Seiji Aguchi [Wed, 14 Nov 2012 20:25:37 +0000 (20:25 +0000)]
efi_pstore: Check remaining space with QueryVariableInfo() before writing data

commit d80a361d779a9f19498943d1ca84243209cd5647 upstream.


As discussed in a thread below, Running out of space in EFI isn't a well-tested scenario.
And we wouldn't expect all firmware to handle it gracefully.

On the other hand, current efi_pstore doesn't check a remaining space of storage at writing time.
Therefore, efi_pstore may not work if it tries to write a large amount of data.

[Patch Description]

To avoid handling the situation above, this patch checks if there is a space enough to log with
QueryVariableInfo() before writing data.

Signed-off-by: Seiji Aguchi <>
Acked-by: Mike Waychison <>
Signed-off-by: Tony Luck <>
Signed-off-by: Ben Hutchings <>
8 years agotty: Correct tty buffer flush.
Ilya Zykov [Mon, 4 Mar 2013 19:19:41 +0000 (23:19 +0400)]
tty: Correct tty buffer flush.

commit 64325a3be08d364a62ee8f84b2cf86934bc2544a upstream.

  The root of problem is carelessly zeroing pointer(in function __tty_buffer_flush()),
when another thread can use it. It can be cause of "NULL pointer dereference".
  Main idea of the patch, this is never free last (struct tty_buffer) in the active buffer.
Only flush the data for ldisc(buf->head->read = buf->head->commit).
At that moment driver can collect(write) data in buffer without conflict.
It is repeat behavior of flush_to_ldisc(), only without feeding data to ldisc.

Signed-off-by: Ilya Zykov <>
Signed-off-by: Ben Hutchings <>
8 years agoiommu/amd: Initialize device table after dma_ops
Shuah Khan [Thu, 28 Feb 2013 21:39:50 +0000 (14:39 -0700)]
iommu/amd: Initialize device table after dma_ops

commit f528d980c17b8714aedc918ba86e058af914d66b upstream.

When dma_ops are initialized the unity mappings are created. The
init_device_table_dma() function makes sure DMA from all devices is
blocked by default. This opens a short window in time where DMA to
unity mapped regions is blocked by the IOMMU. Make sure this does not
happen by initializing the device table after dma_ops.

Tested on 3.2.38

Signed-off-by: Joerg Roedel <>
Signed-off-by: Shuah Khan <>
Signed-off-by: Ben Hutchings <>
8 years agoahci: Add Device IDs for Intel Wellsburg PCH
James Ralston [Sat, 9 Feb 2013 01:34:47 +0000 (17:34 -0800)]
ahci: Add Device IDs for Intel Wellsburg PCH

commit 151743fd8dfb02956c5184b5f4f0f42677eb75bc upstream.

This patch adds the AHCI-mode SATA Device IDs for the Intel Wellsburg PCH

Signed-off-by: James Ralston <>
Signed-off-by: Jeff Garzik <>
Signed-off-by: Ben Hutchings <>
8 years agoahci: AHCI-mode SATA patch for Intel Avoton DeviceIDs
Seth Heasley [Fri, 25 Jan 2013 20:01:05 +0000 (12:01 -0800)]
ahci: AHCI-mode SATA patch for Intel Avoton DeviceIDs

commit 29e674dd5c8e781589f09c3ee139c80f6da274e4 upstream.

This patch adds the AHCI and RAID-mode SATA DeviceIDs for the Intel Avoton SOC.

Signed-off-by: Seth Heasley <>
Signed-off-by: Jeff Garzik <>
Signed-off-by: Ben Hutchings <>
8 years agoahci: Add Device IDs for Intel Lynx Point-LP PCH
James Ralston [Thu, 9 Aug 2012 16:02:31 +0000 (09:02 -0700)]
ahci: Add Device IDs for Intel Lynx Point-LP PCH

commit 77b12bc9cf7b10c7c1a04ca45272fbb4287902d0 upstream.

This patch adds the AHCI-mode SATA Device IDs for the Intel Lynx Point-LP PCH

Signed-off-by: James Ralston <>
Signed-off-by: Jeff Garzik <>
Signed-off-by: Ben Hutchings <>
8 years agoahci: AHCI-mode SATA patch for Intel Lynx Point DeviceIDs
Seth Heasley [Tue, 24 Jan 2012 00:27:30 +0000 (16:27 -0800)]
ahci: AHCI-mode SATA patch for Intel Lynx Point DeviceIDs

commit ea4ace66782fc35245133d2766f38d62827761cc upstream.

This patch adds the AHCI-mode SATA DeviceIDs for the Intel Lynx Point PCH.

Signed-off-by: Seth Heasley <>
Signed-off-by: Jeff Garzik <>
Signed-off-by: Ben Hutchings <>
8 years agoHID: clean up quirk for Sony RF receivers
Fernando Luis Vázquez Cao [Tue, 22 Jan 2013 06:20:38 +0000 (15:20 +0900)]
HID: clean up quirk for Sony RF receivers

commit 99d249021abd4341771523ed8dd7946276103432 upstream.

Document what the fix-up is does and make it more robust by ensuring
that it is only applied to the USB interface that corresponds to the
mouse (sony_report_fixup() is called once per interface during probing).

Signed-off-by: Fernando Luis Vazquez Cao <>
Signed-off-by: Jiri Kosina <>
Signed-off-by: Ben Hutchings <>
8 years agoHID: add support for Sony RF receiver with USB product id 0x0374
Fernando Luis Vázquez Cao [Tue, 15 Jan 2013 10:40:48 +0000 (19:40 +0900)]
HID: add support for Sony RF receiver with USB product id 0x0374

commit a464918419f94a0043d2f549d6defb4c3f69f68a upstream.

Some Vaio desktop computers, among them the VGC-LN51JGB multimedia PC, have
a RF receiver, multi-interface USB device 054c:0374, that is used to connect
a wireless keyboard and a wireless mouse.

The keyboard works flawlessly, but the mouse (VGP-WMS3 in my case) does not
seem to be generating any pointer events. The problem is that the mouse pointer
is wrongly declared as a constant non-data variable in the report descriptor
(see lsusb and usbhid-dump output below), with the consequence that it is
ignored by the HID code.

Add this device to the have-special-driver list and fix up the report
descriptor in the Sony-specific driver which happens to already have a fixup
for a similar firmware bug.

# lsusb -vd 054C:0374
Bus 003 Device 002: ID 054c:0374 Sony Corp.
Device Descriptor:
  bLength                18
  bDescriptorType         1
  bcdUSB               2.00
  bDeviceClass            0 (Defined at Interface level)
  bDeviceSubClass         0
  bDeviceProtocol         0
  bMaxPacketSize0         8
  idVendor           0x054c Sony Corp.
  idProduct          0x0374
  iSerial                 0
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        1
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         3 Human Interface Device
      bInterfaceSubClass      1 Boot Interface Subclass
      bInterfaceProtocol      2 Mouse
      iInterface              2 RF Receiver
          Report Descriptor: (length is 100)
            Item(Global): Usage Page, data= [ 0x01 ] 1
                            Generic Desktop Controls
            Item(Local ): Usage, data= [ 0x30 ] 48
            Item(Local ): Usage, data= [ 0x31 ] 49
            Item(Global): Report Count, data= [ 0x02 ] 2
            Item(Global): Report Size, data= [ 0x08 ] 8
            Item(Global): Logical Minimum, data= [ 0x81 ] 129
            Item(Global): Logical Maximum, data= [ 0x7f ] 127
            Item(Main  ): Input, data= [ 0x07 ] 7
                            Constant Variable Relative No_Wrap Linear
                            Preferred_State No_Null_Position Non_Volatile Bitfield

# usbhid-dump
003:002:001:DESCRIPTOR         1357910009.758544
 05 01 09 02 A1 01 05 01 09 02 A1 02 85 01 09 01
 A1 00 05 09 19 01 29 05 95 05 75 01 15 00 25 01
 81 02 75 03 95 01 81 01 05 01 09 30 09 31 95 02
 75 08 15 81 25 7F 81 07 A1 02 85 01 09 38 35 00
 45 00 15 81 25 7F 95 01 75 08 81 06 C0 A1 02 85
 01 05 0C 15 81 25 7F 95 01 75 08 0A 38 02 81 06
 C0 C0 C0 C0

Signed-off-by: Fernando Luis Vazquez Cao <>
Signed-off-by: Jiri Kosina <>
Signed-off-by: Ben Hutchings <>
8 years agoiwlwifi: always copy first 16 bytes of commands
Johannes Berg [Mon, 25 Feb 2013 15:01:34 +0000 (16:01 +0100)]
iwlwifi: always copy first 16 bytes of commands

commit 8a964f44e01ad3bbc208c3e80d931ba91b9ea786 upstream.

The FH hardware will always write back to the scratch field
in commands, even host commands not just TX commands, which
can overwrite parts of the command. This is problematic if
the command is re-used (with IWL_HCMD_DFL_NOCOPY) and can
cause calibration issues.

Address this problem by always putting at least the first
16 bytes into the buffer we also use for the command header
and therefore make the DMA engine write back into this.

For commands that are smaller than 16 bytes also always map
enough memory for the DMA engine to write back to.

Reviewed-by: Emmanuel Grumbach <>
Signed-off-by: Johannes Berg <>
[bwh: Backported to 3.2:
 - Adjust context
 - Drop the IWL_HCMD_DFL_DUP handling
 - Fix descriptor addresses and lengths for tracepoint, but otherwise
   leave it unchanged]
Signed-off-by: Ben Hutchings <>
8 years agodmi_scan: fix missing check for _DMI_ signature in smbios_present()
Ben Hutchings [Fri, 8 Mar 2013 20:43:32 +0000 (12:43 -0800)]
dmi_scan: fix missing check for _DMI_ signature in smbios_present()

commit a40e7cf8f06b4e322ba902e4e9f6a6b0c2daa907 upstream.

Commit 9f9c9cbb6057 ("drivers/firmware/dmi_scan.c: fetch dmi version
from SMBIOS if it exists") hoisted the check for "_DMI_" into
dmi_scan_machine(), which means that we don't bother to check for
"_DMI_" at offset 16 in an SMBIOS entry.  smbios_present() may also call
dmi_present() for an address where we found "_SM_", if it failed further

Check for "_DMI_" in smbios_present() before calling dmi_present().

[ fix build]
Signed-off-by: Ben Hutchings <>
Reported-by: Tim McGrath <>
Tested-by: Tim Mcgrath <>
Cc: Zhenzhong Duan <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
8 years agodrm/radeon: add primary dac adj quirk for R200 board
Alex Deucher [Wed, 27 Feb 2013 17:01:58 +0000 (12:01 -0500)]
drm/radeon: add primary dac adj quirk for R200 board

commit e8fc41377f5037ff7a661ea06adc05f1daec1548 upstream.

vbios values are wrong leading to colors that are
too bright.  Use the default values instead.

Signed-off-by: Alex Deucher <>
Signed-off-by: Ben Hutchings <>
8 years agoALSA: vmaster: Fix slave change notification
Takashi Iwai [Tue, 5 Mar 2013 14:43:39 +0000 (15:43 +0100)]
ALSA: vmaster: Fix slave change notification

commit 2069d483b39a603a5f3428a19d3b4ac89aa97f48 upstream.

When a value of a vmaster slave control is changed, the ctl change
notification is sometimes ignored.  This happens when the master
control overrides, e.g. when the corresponding master control is
muted.  The reason is that slave_put() returns the value of the actual
slave put callback, and it doesn't reflect the virtual slave value

This patch fixes the function just to return 1 whenever a slave value
is changed.

Signed-off-by: Takashi Iwai <>
Signed-off-by: Ben Hutchings <>
8 years agohw_random: make buffer usable in scatterlist.
Rusty Russell [Mon, 4 Mar 2013 23:37:08 +0000 (10:07 +1030)]
hw_random: make buffer usable in scatterlist.

commit f7f154f1246ccc5a0a7e9ce50932627d60a0c878 upstream.

virtio_rng feeds the randomness buffer handed by the core directly
into the scatterlist, since commit bb347d98079a547e80bd4722dee1de61e4dca0e8.

However, if CONFIG_HW_RANDOM=m, the static buffer isn't a linear address
(at least on most archs).  We could fix this in virtio_rng, but it's actually
far easier to just do it in the core as virtio_rng would have to allocate
a buffer every time (it doesn't know how much the core will want to read).

Reported-by: Aurelien Jarno <>
Tested-by: Aurelien Jarno <>
Signed-off-by: Rusty Russell <>
Signed-off-by: Ben Hutchings <>
8 years agohwmon: (sht15) Check return value of regulator_enable()
Mark Brown [Sat, 2 Mar 2013 07:33:30 +0000 (15:33 +0800)]
hwmon: (sht15) Check return value of regulator_enable()

commit 3e78080f81481aa8340374d5a37ae033c1cf4272 upstream.

Not having power is a pretty serious error so check that we are able to
enable the supply and error out if we can't.

Signed-off-by: Mark Brown <>
Signed-off-by: Guenter Roeck <>
[bwh: Backported to 3.2: driver does not use the devm API to manage
 memory, so goto err_free_data rather than returning on error]
Signed-off-by: Ben Hutchings <>
8 years agohwmon: (pmbus/ltc2978) Use detected chip ID to select supported functionality
Guenter Roeck [Thu, 21 Feb 2013 18:49:40 +0000 (10:49 -0800)]
hwmon: (pmbus/ltc2978) Use detected chip ID to select supported functionality

commit f366fccd0809f13ba20d64cae3c83f7338c88af7 upstream.

We read the chip ID from the chip, use it to determine if the chip ID provided
to the driver is correct, and report it if wrong. We should also use the
correct chip ID to select supported functionality.

Signed-off-by: Guenter Roeck <>
Acked-by: Jean Delvare <>
Signed-off-by: Ben Hutchings <>
8 years agohwmon: (pmbus/ltc2978) Fix peak attribute handling
Guenter Roeck [Thu, 21 Feb 2013 17:33:25 +0000 (09:33 -0800)]
hwmon: (pmbus/ltc2978) Fix peak attribute handling

commit dbd712c2272764a536e29ad6841dba74989a39d9 upstream.

Peak attributes were not initialized and cleared correctly.
Also, temp2_max is only supported on page 0 and thus does not need to be
an array.

Signed-off-by: Guenter Roeck <>
Acked-by: Jean Delvare <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agoARM: 7663/1: perf: fix ARMv7 EVTYPE_MASK to include NSH bit
Will Deacon [Thu, 28 Feb 2013 16:49:11 +0000 (17:49 +0100)]
ARM: 7663/1: perf: fix ARMv7 EVTYPE_MASK to include NSH bit

commit f2fe09b055e2549de41fb107b34c60bac4a1b0cf upstream.

Masked out PMXEVTYPER.NSH means that we can't enable profiling at PL2,
regardless of the settings in the HDCR.

This patch fixes the broken mask.

Reported-by: Christoffer Dall <>
Signed-off-by: Will Deacon <>
Signed-off-by: Russell King <>
Signed-off-by: Ben Hutchings <>
8 years agodrm/i915: Don't clobber crtc->fb when queue_flip fails
Ville Syrjälä [Fri, 22 Feb 2013 14:53:38 +0000 (16:53 +0200)]
drm/i915: Don't clobber crtc->fb when queue_flip fails

commit 4a35f83b2b7c6aae3fc0d1c4554fdc99dc33ad07 upstream.

Restore crtc->fb to the old framebuffer if queue_flip fails.

While at it, kill the pointless intel_fb temp variable.

v2: Update crtc->fb before queue_flip and restore it back
    after a failure.

Signed-off-by: Ville Syrjälä <>
Reviewed-by: Chris Wilson <>
Reported-and-Tested-by: Mika Kuoppala <>
Signed-off-by: Daniel Vetter <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agodm snapshot: add missing module aliases
Mikulas Patocka [Fri, 1 Mar 2013 22:45:47 +0000 (22:45 +0000)]
dm snapshot: add missing module aliases

commit 23cb21092eb9dcec9d3604b68d95192b79915890 upstream.

Add module aliases so that autoloading works correctly if the user
tries to activate "snapshot-origin" or "snapshot-merge" targets.


Reported-by: Chao Yang <>
Signed-off-by: Mikulas Patocka <>
Signed-off-by: Alasdair G Kergon <>
Signed-off-by: Ben Hutchings <>
8 years agodm: fix truncated status strings
Mikulas Patocka [Fri, 1 Mar 2013 22:45:44 +0000 (22:45 +0000)]
dm: fix truncated status strings

commit fd7c092e711ebab55b2688d3859d95dfd0301f73 upstream.

Avoid returning a truncated table or status string instead of setting
the DM_BUFFER_FULL_FLAG when the last target of a table fills the

When processing a table or status request, the function retrieve_status
calls ti->type->status. If ti->type->status returns non-zero,
retrieve_status assumes that the buffer overflowed and sets

However, targets don't return non-zero values from their status method
on overflow. Most targets returns always zero.

If a buffer overflow happens in a target that is not the last in the
table, it gets noticed during the next iteration of the loop in
retrieve_status; but if a buffer overflow happens in the last target, it
goes unnoticed and erroneously truncated data is returned.

In the current code, the targets behave in the following way:
* dm-crypt returns -ENOMEM if there is not enough space to store the
  key, but it returns 0 on all other overflows.
* dm-thin returns errors from the status method if a disk error happened.
  This is incorrect because retrieve_status doesn't check the error
  code, it assumes that all non-zero values mean buffer overflow.
* all the other targets always return 0.

This patch changes the ti->type->status function to return void (because
most targets don't use the return code). Overflow is detected in
retrieve_status: if the status method fills up the remaining space
completely, it is assumed that buffer overflow happened.

Signed-off-by: Mikulas Patocka <>
Signed-off-by: Alasdair G Kergon <>
[bwh: Backported to 3.2:
 - Adjust context
 - dm_status_fn doesn't take a status_flags parameter
 - Bump the last component of each current version (verified not to
   match any version used in mainline)
 - Drop changes to dm-verity]
Signed-off-by: Ben Hutchings <>
8 years agoxen/pci: We don't do multiple MSI's.
Konrad Rzeszutek Wilk [Thu, 28 Feb 2013 14:05:41 +0000 (09:05 -0500)]
xen/pci: We don't do multiple MSI's.

commit 884ac2978a295b7df3c4a686d3bff6932bbbb460 upstream.

There is no hypercall to setup multiple MSI per PCI device.
As such with these two new commits:
-  08261d87f7d1b6253ab3223756625a5c74532293
   PCI/MSI: Enable multiple MSIs with pci_enable_msi_block_auto()
   AHCI: Support multiple MSIs

we would call the PHYSDEVOP_map_pirq 'nvec' times with the same
contents of the PCI device. Sander discovered that we would get
the same PIRQ value 'nvec' times and return said values to the
caller. That of course meant that the device was configured only
with one MSI and AHCI would fail with:

ahci 0000:00:11.0: version 3.0
xen: registering gsi 19 triggering 0 polarity 1
xen: --> pirq=19 -> irq=19 (gsi=19)
(XEN) [2013-02-27 19:43:07] IOAPIC[0]: Set PCI routing entry (6-19 -> 0x99 -> IRQ 19 Mode:1 Active:1)
ahci 0000:00:11.0: AHCI 0001.0200 32 slots 4 ports 6 Gbps 0xf impl SATA mode
ahci 0000:00:11.0: flags: 64bit ncq sntf ilck pm led clo pmp pio slum part
ahci: probe of 0000:00:11.0 failed with error -22

That is b/c in ahci_host_activate the second call to
devm_request_threaded_irq  would return -EINVAL as we passed in
(on the second run) an IRQ that was never initialized.

Reported-and-Tested-by: Sander Eikelenboom <>
Signed-off-by: Konrad Rzeszutek Wilk <>
Signed-off-by: Ben Hutchings <>
8 years agocifs: ensure that cifs_get_root() only traverses directories
Jeff Layton [Fri, 1 Feb 2013 20:11:01 +0000 (15:11 -0500)]
cifs: ensure that cifs_get_root() only traverses directories

commit ce2ac52105aa663056dfc17966ebed1bf93e6e64 upstream.

Kjell Braden reported this oops:

[  833.211970] BUG: unable to handle kernel NULL pointer dereference at           (null)
[  833.212816] IP: [<          (null)>]           (null)
[  833.213280] PGD 1b9b2067 PUD e9f7067 PMD 0
[  833.213874] Oops: 0010 [#1] SMP
[  833.214344] CPU 0
[  833.214458] Modules linked in: des_generic md4 nls_utf8 cifs vboxvideo drm snd_intel8x0 snd_ac97_codec ac97_bus snd_pcm snd_seq_midi snd_rawmidi snd_seq_midi_event snd_seq bnep rfcomm snd_timer bluetooth snd_seq_device ppdev snd vboxguest parport_pc joydev mac_hid soundcore snd_page_alloc psmouse i2c_piix4 serio_raw lp parport usbhid hid e1000
[  833.215629]
[  833.215629] Pid: 1752, comm: mount.cifs Not tainted 3.0.0-rc7-bisectcifs-fec11dd9a0+ #18 innotek GmbH VirtualBox/VirtualBox
[  833.215629] RIP: 0010:[<0000000000000000>]  [<          (null)>]           (null)
[  833.215629] RSP: 0018:ffff8800119c9c50  EFLAGS: 00010282
[  833.215629] RAX: ffffffffa02186c0 RBX: ffff88000c427780 RCX: 0000000000000000
[  833.215629] RDX: 0000000000000000 RSI: ffff88000c427780 RDI: ffff88000c4362e8
[  833.215629] RBP: ffff8800119c9c88 R08: ffff88001fc15e30 R09: 00000000d69515c7
[  833.215629] R10: ffffffffa0201972 R11: ffff88000e8f6a28 R12: ffff88000c4362e8
[  833.215629] R13: 0000000000000000 R14: 0000000000000000 R15: ffff88001181aaa6
[  833.215629] FS:  00007f2986171700(0000) GS:ffff88001fc00000(0000) knlGS:0000000000000000
[  833.215629] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[  833.215629] CR2: 0000000000000000 CR3: 000000001b982000 CR4: 00000000000006f0
[  833.215629] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  833.215629] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[  833.215629] Process mount.cifs (pid: 1752, threadinfo ffff8800119c8000, task ffff88001c1c16f0)
[  833.215629] Stack:
[  833.215629]  ffffffff8116a9b5 ffff8800119c9c88 ffffffff81178075 0000000000000286
[  833.215629]  0000000000000000 ffff88000c4276c0 ffff8800119c9ce8 ffff8800119c9cc8
[  833.215629]  ffffffff8116b06e ffff88001bc6fc00 ffff88000c4276c0 ffff88000c4276c0
[  833.215629] Call Trace:
[  833.215629]  [<ffffffff8116a9b5>] ? d_alloc_and_lookup+0x45/0x90
[  833.215629]  [<ffffffff81178075>] ? d_lookup+0x35/0x60
[  833.215629]  [<ffffffff8116b06e>] __lookup_hash.part.14+0x9e/0xc0
[  833.215629]  [<ffffffff8116b1d6>] lookup_one_len+0x146/0x1e0
[  833.215629]  [<ffffffff815e4f7e>] ? _raw_spin_lock+0xe/0x20
[  833.215629]  [<ffffffffa01eef0d>] cifs_do_mount+0x26d/0x500 [cifs]
[  833.215629]  [<ffffffff81163bd3>] mount_fs+0x43/0x1b0
[  833.215629]  [<ffffffff8117d41a>] vfs_kern_mount+0x6a/0xd0
[  833.215629]  [<ffffffff8117e584>] do_kern_mount+0x54/0x110
[  833.215629]  [<ffffffff8117fdc2>] do_mount+0x262/0x840
[  833.215629]  [<ffffffff81108a0e>] ? __get_free_pages+0xe/0x50
[  833.215629]  [<ffffffff8117f9ca>] ? copy_mount_options+0x3a/0x180
[  833.215629]  [<ffffffff8118075d>] sys_mount+0x8d/0xe0
[  833.215629]  [<ffffffff815ece82>] system_call_fastpath+0x16/0x1b
[  833.215629] Code:  Bad RIP value.
[  833.215629] RIP  [<          (null)>]           (null)
[  833.215629]  RSP <ffff8800119c9c50>
[  833.215629] CR2: 0000000000000000
[  833.238525] ---[ end trace ec00758b8d44f529 ]---

When walking down the path on the server, it's possible to hit a
symlink. The path walking code assumes that the caller will handle that
situation properly, but cifs_get_root() isn't set up for it. This patch
prevents the oops by simply returning an error.

A better solution would be to try and chase the symlinks here, but that's
fairly complicated to handle.


Reported-and-tested-by: Kjell Braden <>
Signed-off-by: Jeff Layton <>
Signed-off-by: Steve French <>
Signed-off-by: Ben Hutchings <>
8 years agomwifiex: correct sleep delay counter
Avinash Patil [Tue, 26 Feb 2013 00:01:34 +0000 (16:01 -0800)]
mwifiex: correct sleep delay counter

commit 3e7a4ff7c5b6423ddb644df9c41b8b6d2fb79d30 upstream.

Maximum delay for waking up card is 50 ms. Because of typo in
counter, this delay goes to 500ms. This patch fixes the bug.

Signed-off-by: Avinash Patil <>
Signed-off-by: Amitkumar Karwar <>
Signed-off-by: Yogesh Ashok Powar <>
Signed-off-by: Bing Zhao <>
Signed-off-by: John W. Linville <>
Signed-off-by: Ben Hutchings <>
8 years agoath9k_htc: fix signal strength handling issues
Felix Fietkau [Fri, 22 Feb 2013 20:37:25 +0000 (21:37 +0100)]
ath9k_htc: fix signal strength handling issues

commit 838f427955dcfd16858b0108ce29029da0d56a4e upstream.

The ath9k commit 2ef167557c0a26c88162ecffb017bfcc51eb7b29
(ath9k: fix signal strength reporting issues) fixed an issue where the
reported per-frame signal strength reported to mac80211 was being
overwritten with an internal average. The same issue is also present
in ath9k_htc.
In addition to preventing the driver from overwriting the value, this
commit also ensures that the internal average (which is used for ANI)
only tracks beacons of the AP that we're connected to.

Signed-off-by: Felix Fietkau <>
Signed-off-by: John W. Linville <>
[bwh: Backported to 3.2: use compare_ether_addr() instead of
 ether_addr_equal(), with opposite sense]
Signed-off-by: Ben Hutchings <>
8 years agoath9k: fix RSSI dummy marker value
Felix Fietkau [Fri, 22 Feb 2013 20:09:17 +0000 (21:09 +0100)]
ath9k: fix RSSI dummy marker value

commit a3d63cadbad97671d740a9698acc2c95d1ca6e79 upstream.

RSSI is being stored internally as s8 in several places. The indication
of an unset RSSI value, ATH_RSSI_DUMMY_MARKER, was supposed to have been
set to 127, but ended up being set to 0x127 because of a code cleanup
mistake. This could lead to invalid signal strength values in a few

Signed-off-by: Felix Fietkau <>
Signed-off-by: John W. Linville <>
Signed-off-by: Ben Hutchings <>
8 years agoproc connector: reject unprivileged listener bumps
Kees Cook [Mon, 25 Feb 2013 21:32:25 +0000 (21:32 +0000)]
proc connector: reject unprivileged listener bumps

commit e70ab977991964a5a7ad1182799451d067e62669 upstream.

While PROC_CN_MCAST_LISTEN/IGNORE is entirely advisory, it was possible
for an unprivileged user to turn off notifications for all listeners by
sending PROC_CN_MCAST_IGNORE. Instead, require the same privileges as
required for a multicast bind.

Signed-off-by: Kees Cook <>
Cc: Evgeniy Polyakov <>
Cc: Matt Helsley <>
Acked-by: Evgeniy Polyakov <>
Acked-by: Matt Helsley <>
Signed-off-by: David S. Miller <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agomd: raid0: fix error return from create_stripe_zones.
NeilBrown [Thu, 21 Feb 2013 04:36:38 +0000 (15:36 +1100)]
md: raid0: fix error return from create_stripe_zones.

commit 58ebb34c49fcfcaa029e4b1c1453d92583900f9a upstream.

Create_stripe_zones returns an error slightly differently to
raid0_run and to raid0_takeover_*.

The error returned used by the second was wrong and an error would
result in mddev->private being set to NULL and sooner or later a

So never return NULL, return ERR_PTR(err), not NULL from

This bug has been present since 2.6.35 so the fix is suitable
for any kernel since then.

Signed-off-by: NeilBrown <>
Signed-off-by: Ben Hutchings <>
8 years agomd: fix two bugs when attempting to resize RAID0 array.
NeilBrown [Thu, 21 Feb 2013 03:33:17 +0000 (14:33 +1100)]
md: fix two bugs when attempting to resize RAID0 array.

commit a64685399181780998281fe07309a94b25dd24c3 upstream.

You cannot resize a RAID0 array (in terms of making the devices
bigger), but the code doesn't entirely stop you.

 disable setting of the available size on each device for
 RAID0 and Linear devices.  This must not change as doing so
 can change the effective layout of data.

 Make sure that the size that raid0_size() reports is accurate,
 but rounding devices sizes to chunk sizes.  As the device sizes
 cannot change now, this isn't so important, but it is best to be

Without this change:
  mdadm --grow /dev/md0 -z max
  mdadm --grow /dev/md0 -Z max
  then read to the end of the array

can cause a BUG in a RAID0 array.

These bugs have been present ever since it became possible
to resize any device, which is a long time.  So the fix is
suitable for any -stable kerenl.

Signed-off-by: NeilBrown <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agoARM: fix scheduling while atomic warning in alignment handling code
Russell King [Mon, 25 Feb 2013 16:10:42 +0000 (16:10 +0000)]
ARM: fix scheduling while atomic warning in alignment handling code

commit b255188f90e2bade1bd11a986dd1ca4861869f4d upstream.

Paolo Pisati reports that IPv6 triggers this warning:

BUG: scheduling while atomic: swapper/0/0/0x40000100
Modules linked in:
[<c001b1c4>] (unwind_backtrace+0x0/0xf0) from [<c0503c5c>] (__schedule_bug+0x48/0x5c)
[<c0503c5c>] (__schedule_bug+0x48/0x5c) from [<c0508608>] (__schedule+0x700/0x740)
[<c0508608>] (__schedule+0x700/0x740) from [<c007007c>] (__cond_resched+0x24/0x34)
[<c007007c>] (__cond_resched+0x24/0x34) from [<c05086dc>] (_cond_resched+0x3c/0x44)
[<c05086dc>] (_cond_resched+0x3c/0x44) from [<c0021f6c>] (do_alignment+0x178/0x78c)
[<c0021f6c>] (do_alignment+0x178/0x78c) from [<c00083e0>] (do_DataAbort+0x34/0x98)
[<c00083e0>] (do_DataAbort+0x34/0x98) from [<c0509a60>] (__dabt_svc+0x40/0x60)
Exception stack(0xc0763d70 to 0xc0763db8)
3d60:                                     e97e805e e97e806e 2c000000 11000000
3d80: ea86bb00 0000002c 00000011 e97e807e c076d2a8 e97e805e e97e806e 0000002c
3da0: 3d000000 c0763dbc c04b98fc c02a8490 00000113 ffffffff
[<c0509a60>] (__dabt_svc+0x40/0x60) from [<c02a8490>] (__csum_ipv6_magic+0x8/0xc8)

Fix this by using probe_kernel_address() stead of __get_user().

Reported-by: Paolo Pisati <>
Tested-by: Paolo Pisati <>
Signed-off-by: Russell King <>
Signed-off-by: Ben Hutchings <>
8 years agoARM: VFP: fix emulation of second VFP instruction
Russell King [Mon, 25 Feb 2013 16:09:12 +0000 (16:09 +0000)]
ARM: VFP: fix emulation of second VFP instruction

commit 5e4ba617c1b584b2e376f31a63bd4e734109318a upstream.

Martin Storsjö reports that the sequence:

        ee312ac1        vsub.f32        s4, s3, s2
        ee702ac0        vsub.f32        s5, s1, s0
        e59f0028        ldr             r0, [pc, #40]
        ee111a90        vmov            r1, s3

on Raspberry Pi (implementor 41 architecture 1 part 20 variant b rev 5)
where s3 is a denormal and s2 is zero results in incorrect behaviour -
the instruction "vsub.f32 s5, s1, s0" is not executed:

        VFP: bounce: trigger ee111a90 fpexc d0000780
        VFP: emulate: INST=0xee312ac1 SCR=0x00000000

As we can see, the instruction triggering the exception is the "vmov"
instruction, and we emulate the "vsub.f32 s4, s3, s2" but fail to
properly take account of the FPEXC_FP2V flag in FPEXC.  This is because
the test for the second instruction register being valid is bogus, and
will always skip emulation of the second instruction.

Reported-by: Martin Storsjö <>
Tested-by: Martin Storsjö <>
Signed-off-by: Russell King <>
Signed-off-by: Ben Hutchings <>
8 years agodc395x: uninitialized variable in device_alloc()
Dan Carpenter [Mon, 11 Feb 2013 19:03:18 +0000 (22:03 +0300)]
dc395x: uninitialized variable in device_alloc()

commit 208afec4f3be8c51ad6eebe6611dd6d2ad2fa298 upstream.

This bug was introduced back in bitkeeper days in 2003.  We use
"dcb->dev_mode" before it has been initialized.

Signed-off-by: Dan Carpenter <>
Acked-by: Oliver Neukum <>
Signed-off-by: James Bottomley <>
Signed-off-by: Ben Hutchings <>
8 years agostorvsc: Initialize the sglist
K. Y. Srinivasan [Wed, 6 Feb 2013 13:15:28 +0000 (05:15 -0800)]
storvsc: Initialize the sglist

commit 9d2696e658ef4f209955ddaa987d43f1a1bd81a1 upstream.

Properly initialize scatterlist before using it.

Signed-off-by: K. Y. Srinivasan <>
Signed-off-by: James Bottomley <>
[bwh: Backported to 3.2: adjust filename]
Signed-off-by: Ben Hutchings <>
8 years agoSUNRPC: Don't start the retransmission timer when out of socket space
Trond Myklebust [Fri, 22 Feb 2013 19:57:57 +0000 (14:57 -0500)]
SUNRPC: Don't start the retransmission timer when out of socket space

commit a9a6b52ee1baa865283a91eb8d443ee91adfca56 upstream.

If the socket is full, we're better off just waiting until it empties,
or until the connection is broken. The reason why we generally don't
want to time out is that the call to xprt->ops->release_xprt() will
trigger a connection reset, which isn't helpful...

Let's make an exception for soft RPC calls, since they have to provide
timeout guarantees.

Signed-off-by: Trond Myklebust <>
Signed-off-by: Ben Hutchings <>
8 years agoNFS: Don't allow NFS silly-renamed files to be deleted, no signal
Trond Myklebust [Fri, 22 Feb 2013 17:53:43 +0000 (12:53 -0500)]
NFS: Don't allow NFS silly-renamed files to be deleted, no signal

commit 5a7a613a47a715711b3f2d3322a0eac21d459166 upstream.

Commit 73ca100 broke the code that prevents the client from deleting
a silly renamed dentry.  This affected "delete on last close"
semantics as after that commit, nothing prevented removal of
silly-renamed files.  As a result, a process holding a file open
could easily get an ESTALE on the file in a directory where some
other process issued 'rm -rf some_dir_containing_the_file' twice.
Before the commit, any attempt at unlinking silly renamed files would
fail inside may_delete() with -EBUSY because of the
DCACHE_NFSFS_RENAMED flag.  The following testcase demonstrates
the problem:
  tail -f /nfsmnt/dir/file &
  rm -rf /nfsmnt/dir
  rm -rf /nfsmnt/dir
  # second removal does not fail, 'tail' process receives ESTALE

The problem with the above commit is that it unhashes the old and
new dentries from the lookup path, even in the normal case when
a signal is not encountered and it would have been safe to call
d_move.  Unfortunately the old dentry has the special
DCACHE_NFSFS_RENAMED flag set on it.  Unhashing has the
side-effect that future lookups call d_alloc(), allocating a new
dentry without the special flag for any silly-renamed files.  As a
result, subsequent calls to unlink silly renamed files do not fail
but allow the removal to go through.  This will result in ESTALE
errors for any other process doing operations on the file.

To fix this, go back to using d_move on success.
For the signal case, it's unclear what we may safely do beyond d_drop.

Reported-by: Dave Wysochanski <>
Signed-off-by: Trond Myklebust <>
Acked-by: Jeff Layton <>
Signed-off-by: Ben Hutchings <>
8 years agomd: protect against crash upon fsync on ro array
Sebastian Riemer [Thu, 21 Feb 2013 02:28:09 +0000 (13:28 +1100)]
md: protect against crash upon fsync on ro array

commit bbfa57c0f2243a7c31fd248d22e9861a2802cad5 upstream.

If an fsync occurs on a read-only array, we need to send a
completion for the IO and may not increment the active IO count.
Otherwise, we hit a bug trace and can't stop the MD array anymore.

By advice of Christoph Hellwig we return success upon a flush
request but we return -EROFS for other writes.
We detect flush requests by checking if the bio has zero sectors.

This patch is suitable to any -stable kernel to which it applies.

Cc: Christoph Hellwig <>
Cc: Ben Hutchings <>
Cc: NeilBrown <>
Signed-off-by: Sebastian Riemer <>
Reported-by: Ben Hutchings <>
Acked-by: Paul Menzel <>
Signed-off-by: NeilBrown <>
Signed-off-by: Ben Hutchings <>
8 years agobtrfs: Init io_lock after cloning btrfs device struct
Thomas Gleixner [Wed, 20 Feb 2013 19:06:20 +0000 (14:06 -0500)]
btrfs: Init io_lock after cloning btrfs device struct

commit 1cba0cdf5e4dbcd9e5fa5b54d7a028e55e2ca057 upstream.

__btrfs_close_devices() clones btrfs device structs with
memcpy(). Some of the fields in the clone are reinitialized, but it's
missing to init io_lock. In mainline this goes unnoticed, but on RT it
leaves the plist pointing to the original about to be freed lock

Initialize io_lock after cloning, so no references to the original
struct are left.

Reported-and-tested-by: Mike Galbraith <>
Signed-off-by: Thomas Gleixner <>
Signed-off-by: Chris Mason <>
Signed-off-by: Ben Hutchings <>
8 years agoRevert "powerpc/eeh: Fix crash when adding a device in a slot with DDW"
Ben Hutchings [Thu, 14 Mar 2013 00:15:05 +0000 (00:15 +0000)]
Revert "powerpc/eeh: Fix crash when adding a device in a slot with DDW"

This reverts commit 066f289835f09a3f744d6bac96f25e25d20b3ded which was
6a040ce72598159a74969a2d01ab0ba5ee6536b3 upstream.

This was not needed and is not suitable for 3.2.y.

Reported-by: Michael Neuling <>
Signed-off-by: Ben Hutchings <>
8 years agoLinux 3.2.40 v3.2.40
Ben Hutchings [Wed, 6 Mar 2013 03:24:30 +0000 (03:24 +0000)]
Linux 3.2.40

8 years agoext4: fix kernel BUG on large-scale rm -rf commands
Theodore Ts'o [Fri, 17 Aug 2012 12:54:52 +0000 (08:54 -0400)]
ext4: fix kernel BUG on large-scale rm -rf commands

commit 89a4e48f8479f8145eca9698f39fe188c982212f upstream.

Commit 968dee7722: "ext4: fix hole punch failure when depth is greater
than 0" introduced a regression in v3.5.1/v3.6-rc1 which caused kernel
crashes when users ran run "rm -rf" on large directory hierarchy on
ext4 filesystems on RAID devices:

    BUG: unable to handle kernel NULL pointer dereference at 0000000000000028

    Process rm (pid: 18229, threadinfo ffff8801276bc000, task ffff880123631710)
    Call Trace:
     [<ffffffff81236483>] ? __ext4_handle_dirty_metadata+0x83/0x110
     [<ffffffff812353d3>] ext4_ext_truncate+0x193/0x1d0
     [<ffffffff8120a8cf>] ? ext4_mark_inode_dirty+0x7f/0x1f0
     [<ffffffff81207e05>] ext4_truncate+0xf5/0x100
     [<ffffffff8120cd51>] ext4_evict_inode+0x461/0x490
     [<ffffffff811a1312>] evict+0xa2/0x1a0
     [<ffffffff811a1513>] iput+0x103/0x1f0
     [<ffffffff81196d84>] do_unlinkat+0x154/0x1c0
     [<ffffffff8118cc3a>] ? sys_newfstatat+0x2a/0x40
     [<ffffffff81197b0b>] sys_unlinkat+0x1b/0x50
     [<ffffffff816135e9>] system_call_fastpath+0x16/0x1b
    Code: 8b 4d 20 0f b7 41 02 48 8d 04 40 48 8d 04 81 49 89 45 18 0f b7 49 02 48 83 c1 01 49 89 4d 00 e9 ae f8 ff ff 0f 1f 00 49 8b 45 28 <48> 8b 40 28 49 89 45 20 e9 85 f8 ff ff 0f 1f 80 00 00 00

    RIP  [<ffffffff81233164>] ext4_ext_remove_space+0xa34/0xdf0

This could be reproduced as follows:

The problem in commit 968dee7722 was that caused the variable 'i' to
be left uninitialized if the truncate required more space than was
available in the journal.  This resulted in the function
ext4_ext_truncate_extend_restart() returning -EAGAIN, which caused
ext4_ext_remove_space() to restart the truncate operation after
starting a new jbd2 handle.

Reported-by: Maciej Żenczykowski <>
Reported-by: Marti Raudsepp <>
Tested-by: Fengguang Wu <>
Signed-off-by: "Theodore Ts'o" <>
Signed-off-by: Ben Hutchings <>
8 years agoext4: fix hole punch failure when depth is greater than 0
Ashish Sangwan [Mon, 23 Jul 2012 02:49:08 +0000 (22:49 -0400)]
ext4: fix hole punch failure when depth is greater than 0

commit 968dee77220768a5f52cf8b21d0bdb73486febef upstream.

Whether to continue removing extents or not is decided by the return
value of function ext4_ext_more_to_rm() which checks 2 conditions:
a) if there are no more indexes to process.
b) if the number of entries are decreased in the header of "depth -1".

In case of hole punch, if the last block to be removed is not part of
the last extent index than this index will not be deleted, hence the
number of valid entries in the extent header of "depth - 1" will
remain as it is and ext4_ext_more_to_rm will return 0 although the
required blocks are not yet removed.

This patch fixes the above mentioned problem as instead of removing
the extents from the end of file, it starts removing the blocks from
the particular extent from which removing blocks is actually required
and continue backward until done.

Signed-off-by: Ashish Sangwan <>
Signed-off-by: Namjae Jeon <>
Reviewed-by: Lukas Czerner <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agoext4: rewrite punch hole to use ext4_ext_remove_space()
Lukas Czerner [Tue, 20 Mar 2012 03:03:19 +0000 (23:03 -0400)]
ext4: rewrite punch hole to use ext4_ext_remove_space()

commit 5f95d21fb6f2aaa52830e5b7fb405f6c71d3ab85 upstream.

This commit rewrites ext4 punch hole implementation to use
ext4_ext_remove_space() instead of its home gown way of doing this via
ext4_ext_map_blocks(). There are several reasons for changing this.

Firstly it is quite non obvious that punching hole needs to
ext4_ext_map_blocks() to punch a hole, especially given that this
function should map blocks, not unmap it. It also required a lot of new
code in ext4_ext_map_blocks().

Secondly the design of it is not very effective. The reason is that we
are trying to punch out blocks in ext4_ext_punch_hole() in opposite
direction than in ext4_ext_rm_leaf() which causes the ext4_ext_rm_leaf()
to iterate through the whole tree from the end to the start to find the
requested extent for every extent we are going to punch out.

And finally the current implementation does not use the existing code,
but bring a lot of new code, which is IMO unnecessary since there
already is some infrastructure we can use. Specifically

This commit changes ext4_ext_remove_space() to accept 'end' parameter so
we can not only truncate to the end of file, but also remove the space
in the middle of the file (punch a hole). Moreover, because the last
block to punch out, might be in the middle of the extent, we have to
split the extent at 'end + 1' so ext4_ext_rm_leaf() can easily either
remove the whole fist part of split extent, or change its size.

ext4_ext_remove_space() is then used to actually remove the space
(extents) from within the hole, instead of ext4_ext_map_blocks().

Note that this also fix the issue with punch hole, where we would forget
to remove empty index blocks from the extent tree, resulting in double
free block error and file system corruption. This is simply because we
now use different code path, where this problem does not exist.

This has been tested with fsx running for several days and xfstests,
plus xfstest #251 with '-o discard' run on the loop image (which
converts discard requestes into punch hole to the backing file). All of
it on 1K and 4K file system block size.

Signed-off-by: Lukas Czerner <>
Signed-off-by: "Theodore Ts'o" <>
[bwh: Backported to 3.2.y: move EXT4_EXT_DATA_VALID{1,2} along with the
 other extent splitting flags]
Signed-off-by: Ben Hutchings <>
8 years agofs: cachefiles: add support for large files in filesystem caching
Justin Lecher [Mon, 30 Jul 2012 21:42:53 +0000 (14:42 -0700)]
fs: cachefiles: add support for large files in filesystem caching

commit 98c350cda2c14a343d34ea01a3d9c24fea5ec66d upstream.

Support the caching of large files.


Signed-off-by: Justin Lecher <>
Signed-off-by: Suresh Jayaraman <>
Tested-by: Suresh Jayaraman <>
Acked-by: David Howells <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
[bwh: Backported to 3.2:
 - Adjust context
 - dentry_open() takes dentry and vfsmount pointers, not a path pointer]
Signed-off-by: Ben Hutchings <>
8 years agoexec: use -ELOOP for max recursion depth
Kees Cook [Thu, 6 Dec 2012 06:00:21 +0000 (17:00 +1100)]
exec: use -ELOOP for max recursion depth

commit d740269867021faf4ce38a449353d2b986c34a67 upstream.

To avoid an explosion of request_module calls on a chain of abusive
scripts, fail maximum recursion with -ELOOP instead of -ENOEXEC. As soon
as maximum recursion depth is hit, the error will fail all the way back
up the chain, aborting immediately.

This also has the side-effect of stopping the user's shell from attempting
to reexecute the top-level file as a shell script. As seen in the
dash source:

        if (cmd != path_bshell && errno == ENOEXEC) {
                *argv-- = cmd;
                *argv = cmd = path_bshell;
                goto repeat;

The above logic was designed for running scripts automatically that lacked
the "#!" header, not to re-try failed recursion. On a legitimate -ENOEXEC,
things continue to behave as the shell expects.

Additionally, when tracking recursion, the binfmt handlers should not be
involved. The recursion being tracked is the depth of calls through
search_binary_handler(), so that function should be exclusively responsible
for tracking the depth.

Signed-off-by: Kees Cook <>
Cc: halfdog <>
Cc: P J P <>
Cc: Alexander Viro <>
Signed-off-by: Andrew Morton <>
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings <>
8 years agokmod: make __request_module() killable
Oleg Nesterov [Fri, 23 Mar 2012 22:02:50 +0000 (15:02 -0700)]
kmod: make __request_module() killable

commit 1cc684ab75123efe7ff446eb821d44375ba8fa30 upstream.

As Tetsuo Handa pointed out, request_module() can stress the system
while the oom-killed caller sleeps in TASK_UNINTERRUPTIBLE.

The task T uses "almost all" memory, then it does something which
triggers request_module().  Say, it can simply call sys_socket().  This
in turn needs more memory and leads to OOM.  oom-killer correctly
chooses T and kills it, but this can't help because it sleeps in
TASK_UNINTERRUPTIBLE and after that oom-killer becomes "disabled" by the

Make __request_module() killable.  The only necessary change is that
call_modprobe() should kmalloc argv and module_name, they can't live in
the stack if we use UMH_KILLABLE.  This memory is freed via

Reported-by: Tetsuo Handa <>
Signed-off-by: Oleg Nesterov <>
Cc: Rusty Russell <>
Cc: Tejun Heo <>
Cc: David Rientjes <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
8 years agokmod: introduce call_modprobe() helper
Oleg Nesterov [Fri, 23 Mar 2012 22:02:49 +0000 (15:02 -0700)]
kmod: introduce call_modprobe() helper

commit 3e63a93b987685f02421e18b2aa452d20553a88b upstream.

No functional changes.  Move the call_usermodehelper code from
__request_module() into the new simple helper, call_modprobe().

Signed-off-by: Oleg Nesterov <>
Cc: Tetsuo Handa <>
Cc: Rusty Russell <>
Cc: Tejun Heo <>
Cc: David Rientjes <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
8 years agousermodehelper: ____call_usermodehelper() doesn't need do_exit()
Oleg Nesterov [Fri, 23 Mar 2012 22:02:49 +0000 (15:02 -0700)]
usermodehelper: ____call_usermodehelper() doesn't need do_exit()

commit 5b9bd473e3b8a8c6c4ae99be475e6e9b27568555 upstream.

Minor cleanup.  ____call_usermodehelper() can simply return, no need to
call do_exit() explicitely.

Signed-off-by: Oleg Nesterov <>
Cc: Tetsuo Handa <>
Cc: Rusty Russell <>
Cc: Tejun Heo <>
Cc: David Rientjes <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
8 years agousermodehelper: implement UMH_KILLABLE
Oleg Nesterov [Fri, 23 Mar 2012 22:02:47 +0000 (15:02 -0700)]
usermodehelper: implement UMH_KILLABLE

commit d0bd587a80960d7ba7e0c8396e154028c9045c54 upstream.

Implement UMH_KILLABLE, should be used along with UMH_WAIT_EXEC/PROC.
The caller must ensure that subprocess_info->path/etc can not go away
until call_usermodehelper_freeinfo().

call_usermodehelper_exec(UMH_KILLABLE) does
wait_for_completion_killable.  If it fails, it uses
xchg(&sub_info->complete, NULL) to serialize with umh_complete() which
does the same xhcg() to access sub_info->complete.

If call_usermodehelper_exec wins, it can safely return.  umh_complete()
should get NULL and call call_usermodehelper_freeinfo().

Otherwise we know that umh_complete() was already called, in this case
call_usermodehelper_exec() falls back to wait_for_completion() which
should succeed "very soon".

Note: UMH_NO_WAIT == -1 but it obviously should not be used with
UMH_KILLABLE.  We delay the neccessary cleanup to simplify the back

Signed-off-by: Oleg Nesterov <>
Cc: Tetsuo Handa <>
Cc: Rusty Russell <>
Cc: Tejun Heo <>
Cc: David Rientjes <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
8 years agousermodehelper: introduce umh_complete(sub_info)
Oleg Nesterov [Fri, 23 Mar 2012 22:02:47 +0000 (15:02 -0700)]
usermodehelper: introduce umh_complete(sub_info)

commit b3449922502f5a161ee2b5022a33aec8472fbf18 upstream.

Preparation.  Add the new trivial helper, umh_complete().  Currently it
simply does complete(sub_info->complete).

Signed-off-by: Oleg Nesterov <>
Cc: Tetsuo Handa <>
Cc: Rusty Russell <>
Cc: Tejun Heo <>
Cc: David Rientjes <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
Signed-off-by: Ben Hutchings <>
8 years agoasus-laptop: Do not call HWRS on init
Ben Hutchings [Thu, 29 Nov 2012 08:12:37 +0000 (09:12 +0100)]
asus-laptop: Do not call HWRS on init

commit cb7da022450cdaaebd33078b6b32fb7dd2aaf6db upstream.

Since commit 8871e99f89b7 ('asus-laptop: HRWS/HWRS typo'), module
initialisation is very slow on the Asus UL30A.  The HWRS method takes
about 12 seconds to run, and subsequent initialisation also seems to
be delayed.  Since we don't really need the result, don't bother
calling it on init.  Those who are curious can still get the result
through the 'infos' device attribute.

Update the comment about HWRS in show_infos().

Reported-by: ryan <>
Signed-off-by: Ben Hutchings <>
Signed-off-by: Corentin Chary <>
Signed-off-by: Matthew Garrett <>
8 years agospeakup: lower default software speech rate
Samuel Thibault [Sun, 26 Aug 2012 21:35:17 +0000 (23:35 +0200)]
speakup: lower default software speech rate

commit cfd757010691eae4e17acc246f74e7622c3a2f05 upstream.

Speech synthesis beginners need a low speech rate, and trained people
want a high speech rate.  A medium speech rate is thus actually not a
good default for neither.  Since trained people will typically know how
to change the rate, better default for a low speech rate, which
beginners can grasp and learn how to increase it afterwards

This was agreed with users on the speakup mailing list.

Signed-off-by: Samuel Thibault <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years agousb: Add USB_QUIRK_RESET_RESUME for all Logitech UVC webcams
Laurent Pinchart [Thu, 19 Jul 2012 10:39:14 +0000 (12:39 +0200)]
usb: Add USB_QUIRK_RESET_RESUME for all Logitech UVC webcams

commit e387ef5c47ddeaeaa3cbdc54424cdb7a28dae2c0 upstream.

Most Logitech UVC webcams (both early models that don't advertise UVC
compatibility and newer UVC-advertised devices) require the RESET_RESUME
quirk. Instead of listing each and every model, match the devices based
on the UVC interface information.

Signed-off-by: Laurent Pinchart <>
Acked-by: Alan Stern <>
Signed-off-by: Greg Kroah-Hartman <>
[bwh: Adjust context to apply after 3.2.38]
Signed-off-by: Ben Hutchings <>
8 years agousb: Add quirk detection based on interface information
Laurent Pinchart [Thu, 19 Jul 2012 10:39:13 +0000 (12:39 +0200)]
usb: Add quirk detection based on interface information

commit 80da2e0df5af700518611b7d1cc4fc9945bcaf95 upstream.

When a whole class of devices (possibly from a specific vendor, or
across multiple vendors) require a quirk, explictly listing all devices
in the class make the quirks table unnecessarily large. Fix this by
allowing matching devices based on interface information.

Signed-off-by: Laurent Pinchart <>
Acked-by: Alan Stern <>
Signed-off-by: Greg Kroah-Hartman <>
Signed-off-by: Ben Hutchings <>
8 years ago8250: use correct value for PORT_BRCM_TRUMANAGE
Ben Hutchings [Sun, 3 Mar 2013 03:24:34 +0000 (03:24 +0000)]
8250: use correct value for PORT_BRCM_TRUMANAGE

When backporting commit ebebd49a8eab ('8250/16?50: Add support for
Broadcom TruManage redirected serial port') I took the next
available port type number for PORT_BRCM_TRUMANAGE (22).

However, the 8250 port type numbers are exposed to userland through
the TIOC{G,S}SERIAL ioctls and so must remain stable.  Redefine
PORT_BRCM_TRUMANAGE as 25, matching mainline as of commit

This leaves port types 22-24 within the valid range for 8250 but not
implemented there.  Change serial8250_verify_port() to specifically
reject these and change serial8250_type() to return "unknown" for them
(though I'm not sure why it would ever see them).

Signed-off-by: Ben Hutchings <>
8 years agovhost: fix length for cross region descriptor
Michael S. Tsirkin [Mon, 26 Nov 2012 05:57:27 +0000 (05:57 +0000)]
vhost: fix length for cross region descriptor

commit bd97120fc3d1a11f3124c7c9ba1d91f51829eb85 upstream.

If a single descriptor crosses a region, the
second chunk length should be decremented
by size translated so far, instead it includes
the full descriptor length.

Signed-off-by: Michael S. Tsirkin <>
Acked-by: Jason Wang <>
Signed-off-by: David S. Miller <>
Signed-off-by: Ben Hutchings <>
8 years agorc: unlock on error in show_protocols()
Dan Carpenter [Tue, 27 Nov 2012 16:35:09 +0000 (13:35 -0300)]
rc: unlock on error in show_protocols()

commit 30ebc5e44d057a1619ad63fe32c8c1670c37c4b8 upstream.

We recently introduced a new return -ENODEV in this function but we need
to unlock before returning.

[ found two patches with the same fix. Merged SOB's/acks into one patch]
Acked-by: Herton R. Krzesinski <>
Signed-off-by: Dan Carpenter <>
Signed-off-by: Douglas Bagnall <>
Signed-off-by: Mauro Carvalho Chehab <>
Signed-off-by: Ben Hutchings <>
8 years agoAvoid sysfs oops when an rc_dev's raw device is absent
Douglas Bagnall [Sat, 7 Jul 2012 02:27:57 +0000 (23:27 -0300)]
Avoid sysfs oops when an rc_dev's raw device is absent

commit 720bb6436ff30fccad05cf5bdf961ea5b1f5686d upstream.

For some reason, when the lirc daemon learns that a usb remote control
has been unplugged, it wants to read the sysfs attributes of the
disappearing device. This is useful for uncovering transient
inconsistencies, but less so for keeping the system running when such
inconsistencies exist.

Under some circumstances (like every time I unplug my dvb stick from
my laptop), lirc catches an rc_dev whose raw event handler has been
removed (presumably by ir_raw_event_unregister), and proceeds to
interrogate the raw protocols supported by the NULL pointer.

This patch avoids the NULL dereference, and ignores the issue of how
this state of affairs came about in the first place.

Version 2 incorporates changes recommended by Mauro Carvalho Chehab
(-ENODEV instead of -EINVAL, and a signed-off-by).

Signed-off-by: Douglas Bagnall <>
Signed-off-by: Mauro Carvalho Chehab <>
Signed-off-by: Ben Hutchings <>