pandora-kernel.git
13 years agoALSA: hda - fix digital mic selection in mixer on 92HD8X codecs
Vitaliy Kulikov [Thu, 10 Mar 2011 01:47:43 +0000 (19:47 -0600)]
ALSA: hda - fix digital mic selection in mixer on 92HD8X codecs

commit 094a42452abd5564429045e210281c6d22e67fca upstream.

When the mux for digital mic is different from the mux for other mics,
the current auto-parser doesn't handle them in a right way but provides
only one mic.  This patch fixes the issue.

Signed-off-by: Vitaliy Kulikov <Vitaliy.Kulikov@idt.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoALSA: HDA: Realtek ALC88x: Do not over-initialize speakers and hp that are primary...
David Henningsson [Fri, 4 Mar 2011 15:54:52 +0000 (16:54 +0100)]
ALSA: HDA: Realtek ALC88x: Do not over-initialize speakers and hp that are primary outputs

commit 0a3fabe30e1a3b2037a12b863b8c45fffce38ee9 upstream.

Do not initialize again the what has already been initialized as
multi outs, as this breaks surround speakers.

Tested-by: Bartłomiej Żogała <nusch88@gmail.com>
Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoNFS: Fix a decoding problem in nfs3_decode_dirent
Trond Myklebust [Fri, 18 Mar 2011 01:54:39 +0000 (21:54 -0400)]
NFS: Fix a decoding problem in nfs3_decode_dirent

[This needs to be applied to 2.6.37 only. The bug in question was
inadvertently fixed by a series of cleanups in 2.6.38, but the patches
in question are too large to be backported. This patch is a minimal fix
that serves the same purpose.]

When we decode a filename followed by an 8-byte cookie, we need to
consider the fact that the filename and cookie are 32-bit word aligned.
Presently, we may end up copying insufficient amounts of data when
xdr_inline_decode() needs to invoke xdr_copy_to_scratch to deal
with a page boundary.

The following patch fixes the issue by first decoding the filename, and
then decoding the cookie.

Reported-by: Neil Brown <neilb@suse.de>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Reviewed-by: NeilBrown <neilb@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoUSB: Add support for SuperSpeed isoc endpoints
Paul Zimmerman [Tue, 1 Mar 2011 02:11:27 +0000 (18:11 -0800)]
USB: Add support for SuperSpeed isoc endpoints

commit 500132a0f26ad7d9916102193cbc6c1b1becb373 upstream.

Use the Mult and bMaxBurst values from the endpoint companion
descriptor to calculate the max length of an isoc transfer.

Add USB_SS_MULT macro to access Mult field of bmAttributes, at
Sarah's suggestion.

This patch should be queued for the 2.6.36 and 2.6.37 stable trees, since
those were the first kernels to have isochronous support for SuperSpeed
devices.

Signed-off-by: Paul Zimmerman <paulz@synopsys.com>
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoxhci: Fix cycle bit calculation during stall handling.
Sarah Sharp [Thu, 24 Feb 2011 02:12:29 +0000 (18:12 -0800)]
xhci: Fix cycle bit calculation during stall handling.

commit 01a1fdb9a7afa5e3c14c9316d6f380732750b4e4 upstream.

When an endpoint stalls, we need to update the xHCI host's internal
dequeue pointer to move it past the stalled transfer.  This includes
updating the cycle bit (TRB ownership bit) if we have moved the dequeue
pointer past a link TRB with the toggle cycle bit set.

When we're trying to find the new dequeue segment, find_trb_seg() is
supposed to keep track of whether we've passed any link TRBs with the
toggle cycle bit set.  However, this while loop's body

while (cur_seg->trbs > trb ||
&cur_seg->trbs[TRBS_PER_SEGMENT - 1] < trb) {

Will never get executed if the ring only contains one segment.
find_trb_seg() will return immediately, without updating the new cycle
bit.  Since find_trb_seg() has no idea where in the segment the TD that
stalled was, make the caller, xhci_find_new_dequeue_state(), check for
this special case and update the cycle bit accordingly.

This patch should be queued to kernels all the way back to 2.6.31.

Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Tested-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoxhci: Update internal dequeue pointers after stalls.
Sarah Sharp [Wed, 23 Feb 2011 23:46:42 +0000 (15:46 -0800)]
xhci: Update internal dequeue pointers after stalls.

commit bf161e85fb153c0dd5a95faca73fd6a9d237c389 upstream.

When an endpoint stalls, the xHCI driver must move the endpoint ring's
dequeue pointer past the stalled transfer.  To do that, the driver issues
a Set TR Dequeue Pointer command, which will complete some time later.

Takashi was having issues with USB 1.1 audio devices that stalled, and his
analysis of the code was that the old code would not update the xHCI
driver's ring dequeue pointer after the command completes.  However, the
dequeue pointer is set in xhci_find_new_dequeue_state(), just before the
set command is issued to the hardware.

Setting the dequeue pointer before the Set TR Dequeue Pointer command
completes is a dangerous thing to do, since the xHCI hardware can fail the
command.  Instead, store the new dequeue pointer in the xhci_virt_ep
structure, and update the ring's dequeue pointer when the Set TR dequeue
pointer command completes.

While we're at it, make sure we can't queue another Set TR Dequeue Command
while the first one is still being processed.  This just won't work with
the internal xHCI state code.  I'm still not sure if this is the right
thing to do, since we might have a case where a driver queues multiple
URBs to a control ring, one of the URBs Stalls, and then the driver tries
to cancel the second URB.  There may be a race condition there where the
xHCI driver might try to issue multiple Set TR Dequeue Pointer commands,
but I would have to think very hard about how the Stop Endpoint and
cancellation code works.  Keep the fix simple until when/if we run into
that case.

This patch should be queued to kernels all the way back to 2.6.31.

Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Tested-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoUSB: move usbcore away from hcd->state
Alan Stern [Mon, 7 Mar 2011 16:11:52 +0000 (11:11 -0500)]
USB: move usbcore away from hcd->state

commit 9b37596a2e860404503a3f2a6513db60c296bfdc upstream.

The hcd->state variable is a disaster.  It's not clearly owned by
either usbcore or the host controller drivers, and they both change it
from time to time, potentially stepping on each other's toes.  It's
not protected by any locks.  And there's no mechanism to prevent it
from going through an invalid transition.

This patch (as1451) takes a first step toward fixing these problems.
As it turns out, usbcore uses hcd->state for essentially only two
things: checking whether the controller's root hub is running and
checking whether the controller has died.  Therefore the patch adds
two new atomic bitflags to the hcd structure, to store these pieces of
information.  The new flags are used only by usbcore, and a private
spinlock prevents invalid combinations (a dead controller's root hub
cannot be running).

The patch does not change the places where usbcore sets hcd->state,
since HCDs may depend on them.  Furthermore, there is one place in
usb_hcd_irq() where usbcore still must use hcd->state: An HCD's
interrupt handler can implicitly indicate that the controller died by
setting hcd->state to HC_STATE_HALT.  Nevertheless, the new code is a
big improvement over the current code.

The patch makes one other change.  The hcd_bus_suspend() and
hcd_bus_resume() routines now check first whether the host controller
has died; if it has then they return immediately without calling the
HCD's bus_suspend or bus_resume methods.

This fixes the major problem reported in Bugzilla #29902: The system
fails to suspend after a host controller dies during system resume.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Alex Terekhov <a.terekhov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoehci: Check individual port status registers on resume
Matthew Garrett [Tue, 11 Jan 2011 17:26:48 +0000 (12:26 -0500)]
ehci: Check individual port status registers on resume

commit 294d95f2cbc2aef5346258f216cd9df570e271a5 upstream.

If a device plug/unplug is detected on an ATI SB700 USB controller in D3,
it appears to set the port status register but not the controller status
register. As a result we'll fail to detect the plug event. Check the port
status register on resume as well in order to catch this case.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoUSB: isp1760: Implement solution for erratum 2
Sebastian Andrzej Siewior [Tue, 8 Feb 2011 20:07:40 +0000 (21:07 +0100)]
USB: isp1760: Implement solution for erratum 2

commit b14e840d04dba211fbdc930247e379085623eacd upstream.

The document says:
|2.1 Problem description
|    When at least two USB devices are simultaneously running, it is observed that
|    sometimes the INT corresponding to one of the USB devices stops occurring. This may
|    be observed sometimes with USB-to-serial or USB-to-network devices.
|    The problem is not noticed when only USB mass storage devices are running.
|2.2 Implication
|    This issue is because of the clearing of the respective Done Map bit on reading the ATL
|    PTD Done Map register when an INT is generated by another PTD completion, but is not
|    found set on that read access. In this situation, the respective Done Map bit will remain
|    reset and no further INT will be asserted so the data transfer corresponding to that USB
|    device will stop.
|2.3 Workaround
|    An SOF INT can be used instead of an ATL INT with polling on Done bits. A time-out can
|    be implemented and if a certain Done bit is never set, verification of the PTD completion
|    can be done by reading PTD contents (valid bit).
|    This is a proven workaround implemented in software.

Russell King run into this with an USB-to-serial converter. This patch
implements his suggestion to enable the high frequent SOF interrupt only
at the time we have ATL packages queued. It goes even one step further
and enables the SOF interrupt only if we have more than one ATL packet
queued at the same time.

Tested-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agostaging: tidspbridge: protect dmm_map properly
Felipe Contreras [Sat, 12 Mar 2011 00:29:06 +0000 (18:29 -0600)]
staging: tidspbridge: protect dmm_map properly

commit ab42abf33a3efdf754710a0a513c00c40854cd61 upstream.

We need to protect not only the dmm_map list, but the individual
map_obj's, otherwise, we might be building the scatter-gather list with
garbage. So, use the existing proc_lock for that.

I observed race conditions which caused kernel panics while running
stress tests, also, Tuomas Kulve found it happening quite often in
Gumstix Over. This patch fixes those.

Cc: Tuomas Kulve <tuomas@kulve.fi>
Signed-off-by: Felipe Contreras <felipe.contreras@nokia.com>
Signed-off-by: Omar Ramirez Luna <omar.ramirez@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agostaging: winbond: needs <linux/delay.h> for msleep and friends
Jeff Mahoney [Thu, 24 Feb 2011 19:49:00 +0000 (14:49 -0500)]
staging: winbond: needs <linux/delay.h> for msleep and friends

commit cecf826df8648c843ea8db63b1f82c154a74db36 upstream.

linux/delay.h is pulled in somehow on x86 but not on ia64 or powerpc.

This fixes a build failure on those arches since they use [mu]delay.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoUSB: serial: ch341: add new id
wangyanqing [Fri, 11 Mar 2011 14:24:38 +0000 (06:24 -0800)]
USB: serial: ch341: add new id

commit d0781383038e983a63843a9a6a067ed781db89c1 upstream.

I picked up a new DAK-780EX(professional digitl reverb/mix system),
which use CH341T chipset to communication with computer on 3/2011
and the CH341T's vendor code is 1a86

Looking up the CH341T's vendor and product id's I see:

1a86  QinHeng Electronics
  5523  CH341 in serial mode, usb to serial port converter

CH341T,CH341 are the products of the same company, maybe
have some common hardware, and I test the ch341.c works
well with CH341T

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoUSB: serial: option: Apply OPTION_BLACKLIST_SENDSETUP also for ZTE MF626
Herton Ronaldo Krzesinski [Wed, 9 Mar 2011 09:19:48 +0000 (09:19 +0000)]
USB: serial: option: Apply OPTION_BLACKLIST_SENDSETUP also for ZTE MF626

commit 7a89e4cb9cdaba92f5fbc509945cf4e3c48db4e2 upstream.

On https://bugs.launchpad.net/ubuntu/+source/linux/+bug/636091, one of
the cases reported is a big timeout on option_send_setup, which causes
some side effects as tty_lock is held. Looks like some of ZTE MF626
devices also don't like the RTS/DTR setting in option_send_setup, like
with 4G XS Stick W14. The reporter confirms which this it solves the
long freezes in his system.

Signed-off-by: Herton Ronaldo Krzesinski <herton.krzesinski@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoUSB: serial/kobil_sct, fix potential tty NULL dereference
Jiri Slaby [Mon, 28 Feb 2011 09:34:06 +0000 (10:34 +0100)]
USB: serial/kobil_sct, fix potential tty NULL dereference

commit 6960f40a954619857e7095a6179eef896f297077 upstream.

Make sure that we check the return value of tty_port_tty_get.
Sometimes it may return NULL and we later dereference that.

The only place here is in kobil_read_int_callback, so fix it.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Alan Cox <alan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoUSB: serial drivers need to use larger bulk-in buffers
Alan Stern [Wed, 23 Feb 2011 20:28:18 +0000 (15:28 -0500)]
USB: serial drivers need to use larger bulk-in buffers

commit 969e3033ae7733a0af8f7742ca74cd16c0857e71 upstream.

When a driver doesn't know how much data a device is going to send,
the buffer size should be at least as big as the endpoint's maxpacket
value.  The serial drivers don't follow this rule; many of them
request only 256-byte bulk-in buffers.  As a result, they suffer
overflow errors if a high-speed device wants to send a lot of data,
because high-speed bulk endpoints are required to have a maxpacket
size of 512.

This patch (as1450) fixes the problem by using the driver's
bulk_in_size value as a minimum, always allocating buffers no smaller
than the endpoint's maxpacket size.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Flynn Marquardt <flynn@flynnux.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoath9k: remove support for the FIF_PROMISC_IN_BSS filter flag
Felix Fietkau [Wed, 9 Mar 2011 00:48:12 +0000 (01:48 +0100)]
ath9k: remove support for the FIF_PROMISC_IN_BSS filter flag

commit 2e286947f1294239527c11f9f466ddce6466455b upstream.

The hardware rx filter flag triggered by FIF_PROMISC_IN_BSS is overly broad
and covers even frames with PHY errors. When this flag is enabled, this message
shows up frequently during scanning or hardware resets:

ath: Could not stop RX, we could be confusing the DMA engine when we start RX up

Since promiscuous mode is usually not particularly useful, yet enabled by
default by bridging (either used normally in 4-addr mode, or with hacks
for various virtualization software), we should sacrifice it for better
reliability during normal operation.

This patch leaves it enabled if there are active monitor mode interfaces, since
it's very useful for debugging.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoath9k_hw: Fix incorrect macversion and macrev checks
Senthil Balasubramanian [Wed, 22 Dec 2010 15:44:20 +0000 (21:14 +0530)]
ath9k_hw: Fix incorrect macversion and macrev checks

commit ac45c12dfb3f727a5a7a3332ed9c11b4a5ab287e upstream.

There are few places where we are checking for macversion and revsions
before RTC is powered ON. However we are reading the macversion and
revisions only after RTC is powered ON and so both macversion and
revisions are actully zero and this leads to incorrect srev checks

Incorrect srev checks can cause registers to be configured wrongly and can
cause unexpected behavior. Fixing this seems to address the ASPM issue that
we have observed. The laptop becomes very slow and hangs mostly with ASPM L1
enabled without this fix.

fix this by reading the macversion and revisisons even before we start
using them. There is no reason why should we delay reading this info
until RTC is powered on as this is just a register information.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoath9k_hw: read and backup AR_WA register value even before chip reset on.
Senthil Balasubramanian [Wed, 22 Dec 2010 13:47:18 +0000 (19:17 +0530)]
ath9k_hw: read and backup AR_WA register value even before chip reset on.

commit 0a8d7cb0c8182df7a28ad719780071178c386f0f upstream.

We need to read and backup AR_WA register value permanently and reading
this after the chip is awakened results in this register being zeroed out.

This seems to fix the ASPM with L1 enabled issue that we have observed.
The laptop becomes very slow and hangs mostly with ASPM L1 enabled without
this fix.

Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agodrm: Hold the mode mutex whilst probing for sysfs status
Chris Wilson [Tue, 15 Mar 2011 11:40:00 +0000 (11:40 +0000)]
drm: Hold the mode mutex whilst probing for sysfs status

commit 007c80a5497a3f9c8393960ec6e6efd30955dcb1 upstream.

As detect will use hw registers and may modify structures, it needs to be
serialised by use of the dev->mode_config.mutex. Make it so.

Otherwise, we may cause random crashes as the sysfs file is queried
whilst a concurrent hotplug poll is being run. For example:

[ 1189.189626] BUG: unable to handle kernel NULL pointer dereference at 00000100
[ 1189.189821] IP: [<e0c22019>] intel_tv_detect_type+0xa2/0x203 [i915]
[ 1189.190020] *pde = 00000000
[ 1189.190104] Oops: 0000 [#1] SMP
[ 1189.190209] last sysfs file: /sys/devices/pci0000:00/0000:00:02.0/drm/card0/card0-SVIDEO-1/status
[ 1189.190412] Modules linked in: mperf cpufreq_conservative cpufreq_userspace cpufreq_powersave cpufreq_stats decnet uinput fuse loop joydev snd_hd a_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm i915 snd_seq_midi snd_rawmidi snd_seq_midi_event snd_seq drm_kms_helper snd_timer uvcvideo d rm snd_seq_device eeepc_laptop tpm_tis usbhid videodev i2c_algo_bit v4l1_compat snd sparse_keymap i2c_core hid serio_raw tpm psmouse evdev tpm_bios rfkill shpchp ac processor rng_c ore battery video power_supply soundcore pci_hotplug button output snd_page_alloc usb_storage uas ext3 jbd mbcache sd_mod crc_t10dif ata_generic ahci libahci ata_piix libata uhci_h cd ehci_hcd scsi_mod usbcore thermal atl2 thermal_sys nls_base [last unloaded: scsi_wait_scan]
[ 1189.192007]
[ 1189.192007] Pid: 1464, comm: upowerd Not tainted 2.6.37-2-686 #1 ASUSTeK Computer INC. 701/701
[ 1189.192007] EIP: 0060:[<e0c22019>] EFLAGS: 00010246 CPU: 0
[ 1189.192007] EIP is at intel_tv_detect_type+0xa2/0x203 [i915]
[ 1189.192007] EAX: 00000000 EBX: dca74000 ECX: e0f68004 EDX: 00068004
[ 1189.192007] ESI: dd110c00 EDI: 400c0c37 EBP: dca7429c ESP: de365e2c
[ 1189.192007]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
[ 1189.192007] Process upowerd (pid: 1464, ti=de364000 task=dcc8acb0 task.ti=de364000)
[ 1189.192007] Stack: Mar 15 03:43:23 hostname kernel: [ 1189.192007]  e0c2cda4 70000000 400c0c30 00000000 dd111000 de365e54 de365f24 dd110c00
[ 1189.192007]  e0c22203 01000000 00000003 00000000 00000000 00000000 00000000 4353544e
[ 1189.192007]  30383420 00000069 00000000 00000000 00000000 00000000 00000000 00000000
[ 1189.192007] Call Trace: Mar 15 03:43:23 hostname kernel: [ 1189.192007]  [<e0c22203>] ?  intel_tv_detect+0x89/0x12d [i915]
[ 1189.192007]  [<e0a9dcef>] ?  status_show+0x0/0x2f [drm]
[ 1189.192007]  [<e0a9dd03>] ?  status_show+0x14/0x2f [drm]

[Digression: what is upowerd doing reading those power hungry files?]

Reported-by: Paul Menzel <paulepanter@users.sourceforge.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoTOMOYO: Fix memory leak upon file open.
Tetsuo Handa [Wed, 2 Mar 2011 07:54:24 +0000 (16:54 +0900)]
TOMOYO: Fix memory leak upon file open.

commit eae61f3c829439f8f9121b5cd48a14be04df451f upstream.

In tomoyo_check_open_permission() since 2.6.36, TOMOYO was by error
recalculating already calculated pathname when checking allow_rewrite
permission. As a result, memory will leak whenever a file is opened for writing
without O_APPEND flag. Also, performance will degrade because TOMOYO is
calculating pathname regardless of profile configuration.
This patch fixes the leak and performance degrade.

Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: James Morris <jmorris@namei.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agox86: stop_machine_text_poke() should issue sync_core()
Mathieu Desnoyers [Thu, 3 Mar 2011 16:01:37 +0000 (11:01 -0500)]
x86: stop_machine_text_poke() should issue sync_core()

commit 0e00f7aed6af21fc09b2a94d28bc34e449bd3a53 upstream.

Intel Archiecture Software Developer's Manual section 7.1.3 specifies that a
core serializing instruction such as "cpuid" should be executed on _each_ core
before the new instruction is made visible.

Failure to do so can lead to unspecified behavior (Intel XMC erratas include
General Protection Fault in the list), so we should avoid this at all cost.

This problem can affect modified code executed by interrupt handlers after
interrupt are re-enabled at the end of stop_machine, because no core serializing
instruction is executed between the code modification and the moment interrupts
are reenabled.

Because stop_machine_text_poke performs the text modification from the first CPU
decrementing stop_machine_first, modified code executed in thread context is
also affected by this problem. To explain why, we have to split the CPUs in two
categories: the CPU that initiates the text modification (calls text_poke_smp)
and all the others. The scheduler, executed on all other CPUs after
stop_machine, issues an "iret" core serializing instruction, and therefore
handles core serialization for all these CPUs. However, the text modification
initiator can continue its execution on the same thread and access the modified
text without any scheduler call. Given that the CPU that initiates the code
modification is not guaranteed to be the one actually performing the code
modification, it falls into the XMC errata.

Q: Isn't this executed from an IPI handler, which will return with IRET (a
   serializing instruction) anyway?
A: No, now stop_machine uses per-cpu workqueue, so that handler will be
   executed from worker threads. There is no iret anymore.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
LKML-Reference: <20110303160137.GB1590@Krystal>
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Arjan van de Ven <arjan@infradead.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agomicroblaze: Fix /dev/zero corruption from __clear_user()
Steven J. Magnani [Thu, 10 Feb 2011 18:12:13 +0000 (12:12 -0600)]
microblaze: Fix /dev/zero corruption from __clear_user()

commit 6f3946b421395ff853bc0bcdab9c26b50ebbba8f upstream.

A userland read of more than PAGE_SIZE bytes from /dev/zero results in
(a) not all of the bytes returned being zero, and
(b) memory corruption due to zeroing of bytes beyond the user buffer.

This is caused by improper constraints on the assembly __clear_user function.
The constrints don't indicate to the compiler that the pointer argument is
modified. Since the function is inline, this results in double-incrementing
of the pointer when __clear_user() is invoked through a multi-page read() of
/dev/zero.

Signed-off-by: Steven J. Magnani <steve@digidescorp.com>
Acked-by: Michal Simek <monstr@monstr.eu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agox86, quirk: Fix SB600 revision check
Andreas Herrmann [Tue, 15 Mar 2011 14:31:37 +0000 (15:31 +0100)]
x86, quirk: Fix SB600 revision check

commit 1d3e09a304e6c4e004ca06356578b171e8735d3c upstream.

Commit 7f74f8f28a2bd9db9404f7d364e2097a0c42cc12
(x86 quirk: Fix polarity for IRQ0 pin2 override on SB800
systems) introduced a regression. It removed some SB600 specific
code to determine the revision ID without adapting a
corresponding revision ID check for SB600.

See this mail thread:

  http://marc.info/?l=linux-kernel&m=129980296006380&w=2

This patch adapts the corresponding check to cover all SB600
revisions.

Tested-by: Wang Lei <f3d27b@gmail.com>
Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
LKML-Reference: <20110315143137.GD29499@alberich.amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoIB/cm: Bump reference count on cm_id before invoking callback
Sean Hefty [Wed, 23 Feb 2011 16:17:40 +0000 (08:17 -0800)]
IB/cm: Bump reference count on cm_id before invoking callback

commit 29963437a48475036353b95ab142bf199adb909e upstream.

When processing a SIDR REQ, the ib_cm allocates a new cm_id.  The
refcount of the cm_id is initialized to 1.  However, cm_process_work
will decrement the refcount after invoking all callbacks.  The result
is that the cm_id will end up with refcount set to 0 by the end of the
sidr req handler.

If a user tries to destroy the cm_id, the destruction will proceed,
under the incorrect assumption that no other threads are referencing
the cm_id.  This can lead to a crash when the cm callback thread tries
to access the cm_id.

This problem was noticed as part of a larger investigation with kernel
crashes in the rdma_cm when running on a real time OS.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Acked-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoIncrease OSF partition limit from 8 to 18
Linus Torvalds [Wed, 16 Mar 2011 15:04:07 +0000 (08:04 -0700)]
Increase OSF partition limit from 8 to 18

commit 34d211a2d5df4984a35b18d8ccacbe1d10abb067 upstream.

It turns out that while a maximum of 8 partitions may be what people
"should" have had, you can actually fit up to 18 entries(*) in a sector.

And some people clearly were taking advantage of that, like Michael
Cree, who had ten partitions on one of his OSF disks.

(*) The OSF partition data starts at byte offset 64 in the first sector,
    and the array of 16-byte partition entries start at offset 148 in
    the on-disk partition structure.

Reported-by: Michael Cree <mcree@orcon.net.nz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoRDMA/cma: Fix crash in request handlers
Sean Hefty [Wed, 23 Feb 2011 16:11:32 +0000 (08:11 -0800)]
RDMA/cma: Fix crash in request handlers

commit 25ae21a10112875763c18b385624df713a288a05 upstream.

Doug Ledford and Red Hat reported a crash when running the rdma_cm on
a real-time OS.  The crash has the following call trace:

    cm_process_work
       cma_req_handler
          cma_disable_callback
          rdma_create_id
             kzalloc
             init_completion
          cma_get_net_info
          cma_save_net_info
          cma_any_addr
             cma_zero_addr
          rdma_translate_ip
             rdma_copy_addr
          cma_acquire_dev
             rdma_addr_get_sgid
             ib_find_cached_gid
             cma_attach_to_dev
          ucma_event_handler
             kzalloc
             ib_copy_ah_attr_to_user
          cma_comp

[ preempted ]

    cma_write
        copy_from_user
        ucma_destroy_id
           copy_from_user
           _ucma_find_context
           ucma_put_ctx
           ucma_free_ctx
              rdma_destroy_id
                 cma_exch
                 cma_cancel_operation
                 rdma_node_get_transport

        rt_mutex_slowunlock
        bad_area_nosemaphore
        oops_enter

They were able to reproduce the crash multiple times with the
following details:

    Crash seems to always happen on the:
            mutex_unlock(&conn_id->handler_mutex);
    as conn_id looks to have been freed during this code path.

An examination of the code shows that a race exists in the request
handlers.  When a new connection request is received, the rdma_cm
allocates a new connection identifier.  This identifier has a single
reference count on it.  If a user calls rdma_destroy_id() from another
thread after receiving a callback, rdma_destroy_id will proceed to
destroy the id and free the associated memory.  However, the request
handlers may still be in the process of running.  When control returns
to the request handlers, they can attempt to access the newly created
identifiers.

Fix this by holding a reference on the newly created rdma_cm_id until
the request handler is through accessing it.

Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Acked-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agolibata: fix hotplug for drivers which don't implement LPM
Tejun Heo [Thu, 24 Feb 2011 18:30:37 +0000 (19:30 +0100)]
libata: fix hotplug for drivers which don't implement LPM

commit eb0e85e36b971ec31610eda7e3ff5c11c1c44785 upstream.

ata_eh_analyze_serror() suppresses hotplug notifications if LPM is
being used because LPM generates spurious hotplug events.  It compared
whether link->lpm_policy was different from ATA_LPM_MAX_POWER to
determine whether LPM is enabled; however, this is incorrect as for
drivers which don't implement LPM, lpm_policy is always
ATA_LPM_UNKNOWN.  This disabled hotplug detection for all drivers
which don't implement LPM.

Fix it by comparing whether lpm_policy is greater than
ATA_LPM_MAX_POWER.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoahci: AHCI mode SATA patch for Intel Patsburg SATA RAID controller
Seth Heasley [Fri, 11 Mar 2011 19:57:42 +0000 (11:57 -0800)]
ahci: AHCI mode SATA patch for Intel Patsburg SATA RAID controller

commit 64a3903d0885879ba8706a8bcf71c5e3e7664db2 upstream.

This patch adds an updated SATA RAID DeviceID for the Intel Patsburg PCH.

Signed-off-by: Seth Heasley <seth.heasley@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoahci: AHCI mode SATA patch for Intel DH89xxCC DeviceIDs
Seth Heasley [Mon, 10 Jan 2011 20:57:17 +0000 (12:57 -0800)]
ahci: AHCI mode SATA patch for Intel DH89xxCC DeviceIDs

commit a4a461a6df6c0481d5a3d61660ed97f5b539cf16 upstream.

This patch adds the AHCI-mode SATA DeviceID for the Intel DH89xxCC PCH.

Signed-off-by: Seth Heasley <seth.heasley@intel.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoahci: recognize Marvell 88se9125 PCIe SATA 6.0 Gb/s controller
Per Jessen [Tue, 8 Feb 2011 12:54:32 +0000 (13:54 +0100)]
ahci: recognize Marvell 88se9125 PCIe SATA 6.0 Gb/s controller

commit 467b41c688c79d1b5e076fbdf082f9cd5d6a000c upstream.

Recognize Marvell 88SE9125 PCIe SATA 6.0 Gb/s controller.

Signed-off-by: Per Jessen <per@computer.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agox86: Emit "mem=nopentium ignored" warning when not supported
Kamal Mostafa [Fri, 4 Feb 2011 01:38:05 +0000 (17:38 -0800)]
x86: Emit "mem=nopentium ignored" warning when not supported

commit 9a6d44b9adb777ca9549e88cd55bd8f2673c52a2 upstream.

Emit warning when "mem=nopentium" is specified on any arch other
than x86_32 (the only that arch supports it).

Signed-off-by: Kamal Mostafa <kamal@canonical.com>
BugLink: http://bugs.launchpad.net/bugs/553464
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Len Brown <len.brown@intel.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
LKML-Reference: <1296783486-23033-2-git-send-email-kamal@canonical.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agox86: Fix panic when handling "mem={invalid}" param
Kamal Mostafa [Fri, 4 Feb 2011 01:38:04 +0000 (17:38 -0800)]
x86: Fix panic when handling "mem={invalid}" param

commit 77eed821accf5dd962b1f13bed0680e217e49112 upstream.

Avoid removing all of memory and panicing when "mem={invalid}"
is specified, e.g. mem=blahblah, mem=0, or mem=nopentium (on
platforms other than x86_32).

Signed-off-by: Kamal Mostafa <kamal@canonical.com>
BugLink: http://bugs.launchpad.net/bugs/553464
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Len Brown <len.brown@intel.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
LKML-Reference: <1296783486-23033-1-git-send-email-kamal@canonical.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoftrace: Fix memory leak with function graph and cpu hotplug
Steven Rostedt [Fri, 11 Feb 2011 02:26:13 +0000 (21:26 -0500)]
ftrace: Fix memory leak with function graph and cpu hotplug

commit 868baf07b1a259f5f3803c1dc2777b6c358f83cf upstream.

When the fuction graph tracer starts, it needs to make a special
stack for each task to save the real return values of the tasks.
All running tasks have this stack created, as well as any new
tasks.

On CPU hot plug, the new idle task will allocate a stack as well
when init_idle() is called. The problem is that cpu hotplug does
not create a new idle_task. Instead it uses the idle task that
existed when the cpu went down.

ftrace_graph_init_task() will add a new ret_stack to the task
that is given to it. Because a clone will make the task
have a stack of its parent it does not check if the task's
ret_stack is already NULL or not. When the CPU hotplug code
starts a CPU up again, it will allocate a new stack even
though one already existed for it.

The solution is to treat the idle_task specially. In fact, the
function_graph code already does, just not at init_idle().
Instead of using the ftrace_graph_init_task() for the idle task,
which that function expects the task to be a clone, have a
separate ftrace_graph_init_idle_task(). Also, we will create a
per_cpu ret_stack that is used by the idle task. When we call
ftrace_graph_init_idle_task() it will check if the idle task's
ret_stack is NULL, if it is, then it will assign it the per_cpu
ret_stack.

Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Suggested-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agox86/mm: Handle mm_fault_error() in kernel space
Andrey Vagin [Wed, 9 Mar 2011 23:22:23 +0000 (15:22 -0800)]
x86/mm: Handle mm_fault_error() in kernel space

commit f86268549f424f83b9eb0963989270e14fbfc3de upstream.

mm_fault_error() should not execute oom-killer, if page fault
occurs in kernel space.  E.g. in copy_from_user()/copy_to_user().

This would happen if we find ourselves in OOM on a
copy_to_user(), or a copy_from_user() which faults.

Without this patch, the kernels hangs up in copy_from_user(),
because OOM killer sends SIG_KILL to current process, but it
can't handle a signal while in syscall, then the kernel returns
to copy_from_user(), reexcute current command and provokes
page_fault again.

With this patch the kernel return -EFAULT from copy_from_user().

The code, which checks that page fault occurred in kernel space,
has been copied from do_sigbus().

This situation is handled by the same way on powerpc, xtensa,
tile, ...

Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
LKML-Reference: <201103092322.p29NMNPH001682@imap1.linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agox86: Don't check for BIOS corruption in first 64K when there's no need to
Naga Chumbalkar [Fri, 25 Feb 2011 20:31:55 +0000 (20:31 +0000)]
x86: Don't check for BIOS corruption in first 64K when there's no need to

commit a7bd1dafdcc13ec7add4aafc927eb5e3a8d597e6 upstream.

Due to commit 781c5a67f152c17c3e4a9ed9647f8c0be6ea5ae9 it is
likely that the number of areas to scan for BIOS corruption is 0
 -- especially when the first 64K is already reserved
(X86_RESERVE_LOW is 64K by default).

If that's the case then don't set up the scan.

Signed-off-by: Naga Chumbalkar <nagananda.chumbalkar@hp.com>
LKML-Reference: <20110225202838.2229.71011.sendpatchset@nchumbalkar.americas.hpqcorp.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agox86/mm: Fix pgd_lock deadlock
Andrea Arcangeli [Wed, 16 Feb 2011 23:45:22 +0000 (15:45 -0800)]
x86/mm: Fix pgd_lock deadlock

commit a79e53d85683c6dd9f99c90511028adc2043031f upstream.

It's forbidden to take the page_table_lock with the irq disabled
or if there's contention the IPIs (for tlb flushes) sent with
the page_table_lock held will never run leading to a deadlock.

Nobody takes the pgd_lock from irq context so the _irqsave can be
removed.

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Tested-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
LKML-Reference: <201102162345.p1GNjMjm021738@imap1.linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoMIPS: MTX-1: Make au1000_eth probe all PHY addresses
Florian Fainelli [Sun, 27 Feb 2011 18:53:53 +0000 (19:53 +0100)]
MIPS: MTX-1: Make au1000_eth probe all PHY addresses

commit bf3a1eb85967dcbaae42f4fcb53c2392cec32677 upstream.

When au1000_eth probes the MII bus for PHY address, if we do not set
au1000_eth platform data's phy_search_highest_address, the MII probing
logic will exit early and will assume a valid PHY is found at address 0.
For MTX-1, the PHY is at address 31, and without this patch, the link
detection/speed/duplex would not work correctly.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
To: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/2111/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agonfs: fix compilation warning
Jovi Zhang [Wed, 2 Mar 2011 23:19:37 +0000 (23:19 +0000)]
nfs: fix compilation warning

commit 43b7c3f051dea504afccc39bcb56d8e26c2e0b77 upstream.

this commit fix compilation warning as following:
linux-2.6/fs/nfs/nfs4proc.c:3265: warning: comparison of distinct pointer types lacks a cast

Signed-off-by: Jovi Zhang <bookjovi@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoblock: fix mis-synchronisation in blkdev_issue_zeroout()
Lukas Czerner [Fri, 11 Mar 2011 09:23:53 +0000 (10:23 +0100)]
block: fix mis-synchronisation in blkdev_issue_zeroout()

commit 0aeea18964173715a1037034ef6838198f319319 upstream.

BZ29402
https://bugzilla.kernel.org/show_bug.cgi?id=29402

We can hit serious mis-synchronization in bio completion path of
blkdev_issue_zeroout() leading to a panic.

The problem is that when we are going to wait_for_completion() in
blkdev_issue_zeroout() we check if the bb.done equals issued (number of
submitted bios). If it does, we can skip the wait_for_completition()
and just out of the function since there is nothing to wait for.
However, there is a ordering problem because bio_batch_end_io() is
calling atomic_inc(&bb->done) before complete(), hence it might seem to
blkdev_issue_zeroout() that all bios has been completed and exit. At
this point when bio_batch_end_io() is going to call complete(bb->wait),
bb and wait does not longer exist since it was allocated on stack in
blkdev_issue_zeroout() ==> panic!

(thread 1)                      (thread 2)
bio_batch_end_io()              blkdev_issue_zeroout()
  if(bb) {                      ...
    if (bb->end_io)             ...
      bb->end_io(bio, err);     ...
    atomic_inc(&bb->done);      ...
    ...                         while (issued != atomic_read(&bb.done))
    ...                         (let issued == bb.done)
    ...                         (do the rest of the function)
    ...                         return ret;
    complete(bb->wait);
    ^^^^^^^^
    panic

We can fix this easily by simplifying bio_batch and completion counting.

Also remove bio_end_io_t *end_io since it is not used.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Reported-by: Eric Whitney <eric.whitney@hp.com>
Tested-by: Eric Whitney <eric.whitney@hp.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
CC: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agodrm/i915: Fix calculation of backlight value in combined mode
Takashi Iwai [Fri, 11 Mar 2011 11:34:41 +0000 (12:34 +0100)]
drm/i915: Fix calculation of backlight value in combined mode

commit ba3820ade317ee36e496b9b40d2ec3987dd4aef0 upstream.

[The upstream commit above is a combination of revert + one-liner fix.
 This patch contains only the latter that is needed for 2.6.37.x.]

This patch fixes the backlight level calculation for combination mode
used in some models like GM45.  It's due to a wrong bit shift
introduced by the commit a95735569312f2ab0c80425e2cd1e5cb0b4e1870
    drm/i915: Refactor panel backlight controls

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=23472
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=34524
Bugzilla: https://bugzilla.novell.com/show_bug.cgi?id=672946

Signed-off-by: Takashi Iwai <tiwai@suse.de>
13 years agort2x00: fix hang when unplugging USB device in use
Johannes Stezenbach [Mon, 13 Dec 2010 11:32:49 +0000 (12:32 +0100)]
rt2x00: fix hang when unplugging USB device in use

commit d7bb5f845f437662296adbfeaab8fbfce1c32289 upstream.

When an rt2x00 USB device is unplugged while in use, it reliably
hangs the whole system.  After some time the watchdog prints:

BUG: soft lockup - CPU#0 stuck for 64s! [kworker/u:0:5]
...
[<c01a88d8>] (usb_submit_urb+0x0/0x2ac) from [<bf0e752c>] (rt2x00usb_kick_rx_entry+0xb4/0xe8 [rt2x00usb])
[<bf0e7478>] (rt2x00usb_kick_rx_entry+0x0/0xe8 [rt2x00usb]) from [<bf0e7588>] (rt2x00usb_clear_entry+x28/0x2c [rt2x00usb])
[<bf0e7560>] (rt2x00usb_clear_entry+0x0/0x2c [rt2x00usb]) from [<bf0d5bc4>] (rt2x00lib_rxdone+0x2e0/0x2f8 [rt2x00lib])
[<bf0d58e4>] (rt2x00lib_rxdone+0x0/0x2f8 [rt2x00lib]) from [<bf0e7e00>] (rt2x00usb_work_rxdone+0x54/0x74 [rt2x00usb])
[<bf0e7dac>] (rt2x00usb_work_rxdone+0x0/0x74 [rt2x00usb]) from [<c00542b4>] (process_one_work+0x20c/0x35c)

Clear the DEVICE_STATE_PRESENT flag when usb_submit_urb()
returns -ENODEV to fix this.

Signed-off-by: Johannes Stezenbach <js@sig21.net>
Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Cc: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agort2x00: Fix crash on USB unplug
Ivo van Doorn [Thu, 4 Nov 2010 19:41:05 +0000 (20:41 +0100)]
rt2x00: Fix crash on USB unplug

commit 070192dd2975c0e97bbdeac7623b755235c6db7d upstream.

By not scheduling the TX/RX completion worker threads
when Radio is disabled, or hardware has been unplugged,
the queues cannot be completely cleaned.

This causes crashes when the hardware has been unplugged while
the radio is still enabled.

Signed-off-by: Ivo van Doorn <IvDoorn@gmail.com>
Acked-by: Gertjan van Wingerde <gwingerde@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Cc: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agonetfilter: fix export secctx error handling
Pablo Neira Ayuso [Thu, 6 Jan 2011 19:25:00 +0000 (11:25 -0800)]
netfilter: fix export secctx error handling

commit cba85b532e4aabdb97f44c18987d45141fd93faa upstream.

In 1ae4de0cdf855305765592647025bde55e85e451, the secctx was exported
via the /proc/net/netfilter/nf_conntrack and ctnetlink interfaces
instead of the secmark.

That patch introduced the use of security_secid_to_secctx() which may
return a non-zero value on error.

In one of my setups, I have NF_CONNTRACK_SECMARK enabled but no
security modules. Thus, security_secid_to_secctx() returns a negative
value that results in the breakage of the /proc and `conntrack -L'
outputs. To fix this, we skip the inclusion of secctx if the
aforementioned function fails.

This patch also fixes the dynamic netlink message size calculation
if security_secid_to_secctx() returns an error, since its logic is
also wrong.

This problem exists in Linux kernel >= 2.6.37.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoBtrfs: deal with short returns from copy_from_user
Chris Mason [Mon, 7 Mar 2011 16:10:24 +0000 (11:10 -0500)]
Btrfs: deal with short returns from copy_from_user

commit 31339acd07b4ba687906702085127895a56eb920 upstream.

When copy_from_user is only able to copy some of the bytes we requested,
we may end up creating a partially up to date page.  To avoid garbage in
the page, we need to treat a partial copy as a zero length copy.

This makes the rest of the file_write code drop the page and
retry the whole copy instead of marking the partially up to
date page as dirty.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoFix corrupted OSF partition table parsing
Timo Warns [Mon, 14 Mar 2011 13:59:33 +0000 (14:59 +0100)]
Fix corrupted OSF partition table parsing

commit 1eafbfeb7bdf59cfe173304c76188f3fd5f1fd05 upstream.

The kernel automatically evaluates partition tables of storage devices.
The code for evaluating OSF partitions contains a bug that leaks data
from kernel heap memory to userspace for certain corrupted OSF
partitions.

In more detail:

  for (i = 0 ; i < le16_to_cpu(label->d_npartitions); i++, partition++) {

iterates from 0 to d_npartitions - 1, where d_npartitions is read from
the partition table without validation and partition is a pointer to an
array of at most 8 d_partitions.

Add the proper and obvious validation.

Signed-off-by: Timo Warns <warns@pre-sense.de>
[ Changed the patch trivially to not repeat the whole le16_to_cpu()
  thing, and to use an explicit constant for the magic value '8' ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoNFS: NFSROOT should default to "proto=udp"
Chuck Lever [Fri, 11 Mar 2011 20:31:06 +0000 (15:31 -0500)]
NFS: NFSROOT should default to "proto=udp"

commit 53d4737580535e073963b91ce87d4216e434fab5 upstream.

There have been a number of recent reports that NFSROOT is no longer
working with default mount options, but fails only with certain NICs.

Brian Downing <bdowning@lavos.net> bisected to commit 56463e50 "NFS:
Use super.c for NFSROOT mount option parsing".  Among other things,
this commit changes the default mount options for NFSROOT to use TCP
instead of UDP as the underlying transport.

TCP seems less able to deal with NICs that are slow to initialize.
The system logs that have accompanied reports of problems all show
that NFSROOT attempts to establish a TCP connection before the NIC is
fully initialized, and thus the TCP connection attempt fails.

When a TCP connection attempt fails during a mount operation, the
NFS stack needs to fail the operation.  Usually user space knows how
and when to retry it.  The network layer does not report a distinct
error code for this particular failure mode.  Thus, there isn't a
clean way for the RPC client to see that it needs to retry in this
case, but not in others.

Because NFSROOT is used in some environments where it is not possible
to update the kernel command line to specify "udp", the proper thing
to do is change NFSROOT to use UDP by default, as it did before commit
56463e50.

To make it easier to see how to change default mount options for
NFSROOT and to distinguish default settings from mandatory settings,
I've adjusted a couple of areas to document the specifics.

root_nfs_cat() is also modified to deal with commas properly when
concatenating strings containing mount option lists.  This keeps
root_nfs_cat() call sites simpler, now that we may be concatenating
multiple mount option strings.

Tested-by: Brian Downing <bdowning@lavos.net>
Tested-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agosunrpc: Propagate errors from xs_bind() through xs_create_sock()
Ben Hutchings [Tue, 22 Feb 2011 21:54:34 +0000 (21:54 +0000)]
sunrpc: Propagate errors from xs_bind() through xs_create_sock()

commit 4cea288aaf0e11647880cc487350b1dc45d9febc upstream.

xs_create_sock() is supposed to return a pointer or an ERR_PTR-encoded
error, but it currently returns 0 if xs_bind() fails.

Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoASoC: Fix broken bitfield definitions in WM8978
Mark Brown [Wed, 2 Mar 2011 11:01:18 +0000 (11:01 +0000)]
ASoC: Fix broken bitfield definitions in WM8978

commit 823dba5191220fc94b83dc0b3f2178ff0842e294 upstream.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agomtd: fix race in cfi_cmdset_0001 driver
Joakim Tjernlund [Mon, 7 Feb 2011 16:07:11 +0000 (17:07 +0100)]
mtd: fix race in cfi_cmdset_0001 driver

commit ecf3fde07c8dcb92a1bf3fbdfe70905d85cd00e1 upstream.

As inval_cache_and_wait_for_operation() drop and reclaim the lock
to invalidate the cache, some other thread may suspend the operation
before reaching the for(;;) loop. Therefore the loop must start with
checking the chip->state before reading status from the chip.

Signed-off-by: Joakim Tjernlund <Joakim.Tjernlund@transmode.se>
Acked-by: Michael Cashwell <mboards@prograde.net>
Acked-by: Stefan Bigler <stefan.bigler@keymile.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agomtd: mtd_blkdevs: fix double free on error path
Maxim Levitsky [Sat, 8 Jan 2011 23:25:06 +0000 (01:25 +0200)]
mtd: mtd_blkdevs: fix double free on error path

commit bd637f6f22235b4613f9ab6555e8088a455c1ed4 upstream.

This one liner patch fixes double free that will occur if add_mtd_blktrans_dev
fails. On failure it frees the input argument, but all its users also free it
on error which is natural thing to do. Thus don't free it.

All credit for finding that bug belongs to reporters of the bug in the android bugzilla
http://code.google.com/p/android/issues/detail?id=13761

Commit message tweaked by Artem.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agomtd: add "platform:" prefix for platform modalias
Axel Lin [Mon, 7 Mar 2011 03:04:24 +0000 (11:04 +0800)]
mtd: add "platform:" prefix for platform modalias

commit c804c733846572ca85c2bba60c7fe6fa024dff18 upstream.

Since 43cc71eed1250755986da4c0f9898f9a635cb3bf (platform: prefix MODALIAS
with "platform:"), the platform modalias is prefixed with "platform:".

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@nokia.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agomtd: jedec_probe: initialise make sector erase command variable
Antony Pavlov [Fri, 11 Feb 2011 10:00:37 +0000 (13:00 +0300)]
mtd: jedec_probe: initialise make sector erase command variable

commit ceabebb2bd2672f709e4454e16bc6042732e2dfe upstream.

In the commit 08968041bef437ec363623cd3218c2b083537ada
 (mtd: cfi_cmdset_0002: make sector erase command variable)
introdused a field sector_erase_cmd. In the same commit initialisation
of cfi->sector_erase_cmd made in cfi_chip_setup()
(file drivers/mtd/chips/cfi_probe.c), so the CFI chip has no problem:

...
        cfi->cfi_mode = CFI_MODE_CFI;
        cfi->sector_erase_cmd = CMD(0x30);
...

But for the JEDEC chips this initialisation is not carried out,
so the JEDEC chips have sector_erase_cmd == 0.

This patch adds the missing initialisation.

Signed-off-by: Antony Pavlov <antony@niisi.msk.ru>
Acked-by: Guillaume LECERF <glecerf@gmail.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agomtd: jedec_probe: Change variable name from cfi_p to cfi
Antony Pavlov [Fri, 11 Feb 2011 10:00:37 +0000 (13:00 +0300)]
mtd: jedec_probe: Change variable name from cfi_p to cfi

commit efba2e313ea1b1bd69a7c4659263becf43bb1adc upstream.

In the following commit, we'll need to use the CMD() macro in order to
fix the initialisation of the sector_erase_cmd field. That requires the
local variable to be called 'cfi', so change it first in a simple patch.

Signed-off-by: Antony Pavlov <antony@niisi.msk.ru>
Acked-by: Guillaume LECERF <glecerf@gmail.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agohwmon/f71882fg: Set platform drvdata to NULL later
Hans de Goede [Sun, 13 Mar 2011 12:50:33 +0000 (13:50 +0100)]
hwmon/f71882fg: Set platform drvdata to NULL later

commit d9ebaa45472c92704f4814682eec21455edcfa1f upstream.

This avoids a possible race leading to trying to dereference NULL.

Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoLinux 2.6.37.4 v2.6.37.4
Greg Kroah-Hartman [Mon, 14 Mar 2011 21:19:26 +0000 (14:19 -0700)]
Linux 2.6.37.4

13 years agoip6ip6: autoload ip6 tunnel
stephen hemminger [Thu, 10 Mar 2011 11:43:19 +0000 (11:43 +0000)]
ip6ip6: autoload ip6 tunnel

commit 6dfbd87a20a737641ef228230c77f4262434fa24 upstream

ip6ip6: autoload ip6 tunnel

Add necessary alias to autoload ip6ip6 tunnel module.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoDiB7000M: add pid filtering
Olivier Grenie [Fri, 14 Jan 2011 16:58:59 +0000 (13:58 -0300)]
DiB7000M: add pid filtering

commit e192a7cf0effe7680264a5bc35c0ad1bdcdc921c upstream.

This patch adds the pid filtering for the dib7000M demod. It also
corrects the pid filtering for the dib7700 based board. It should
prevent an oops, when using dib7700p based board.

References: https://bugzilla.novell.com/show_bug.cgi?id=644807

Signed-off-by: Olivier Grenie <olivier.grenie@dibcom.fr>
Signed-off-by: Patrick Boettcher <patrick.boettcher@dibcom.fr>
Tested-by: Pavel SKARKA <paul.sp@seznam.cz>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agocx25840: fix probing of cx2583x chips
Sven Barth [Mon, 14 Feb 2011 01:09:43 +0000 (22:09 -0300)]
cx25840: fix probing of cx2583x chips

commit 1e6406b8f0dc1ae7d7c39c9e1ac6ca78e016ebfb upstream.

Fix the probing of cx2583x chips, because two controls were clustered
that are not created for these chips.

This regression was introduced in 2.6.36.

Signed-off-by: Sven Barth <pascaldragon@googlemail.com>
Signed-off-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agocx23885: Revert "Check for slave nack on all transactions"
Andy Walls [Mon, 14 Feb 2011 00:52:50 +0000 (21:52 -0300)]
cx23885: Revert "Check for slave nack on all transactions"

commit 67914b5c400d6c213f9e56d7547a2038ab5c06f4 upstream.

This reverts commit 44835f197bf1e3f57464f23dfb239fef06cf89be.

With the CX23885 hardware I2C master, checking for I2C slave ACK/NAK
is not valid when the I2C_EXTEND or I2C_NOSTOP bits are set.
Revert the commit that checks for I2C slave ACK/NAK on all transactions,
so that XC5000 tuners work with the CX23885 again.

Thanks go to Mark Zimmerman for reporting and bisecting this problem.

Bisected-by: Mark Zimmerman <markzimm@frii.com>
Reported-by: Mark Zimmerman <markzimm@frii.com>
Signed-off-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang
Michael [Sat, 26 Feb 2011 04:56:34 +0000 (01:56 -0300)]
ivtv: Fix corrective action taken upon DMA ERR interrupt to avoid hang

commit d213ad08362909ab50fbd6568fcc9fd568268d29 upstream.

After upgrading the kernel from stock Ubuntu 7.10 to
10.04, with no hardware changes, I started getting the dreaded DMA
TIMEOUT errors, followed by inability to encode until the machine was
rebooted.

I came across a post from Andy in March
(http://www.gossamer-threads.com/lists/ivtv/users/40943#40943) where he
speculates that perhaps the corrective actions being taken after a DMA
ERROR are not sufficient to recover the situation.  After some testing
I suspect that this is indeed the case, and that in fact the corrective
action may be what hangs the card's DMA engine, rather than the
original error.

Specifically these DMA ERROR IRQs seem to present with two different
values in the IVTV_REG_DMASTATUS register: 0x11 and 0x13.  The current
corrective action is to clear that status register back to 0x01 or
0x03, and then issue the next DMA request.  In the case of a 0x13 this
seems to result in a minor glitch in the encoded stream due to the
failed transfer that was not retried, but otherwise things continue OK.
In the case of a 0x11 the card's DMA write engine is never heard from
again, and a DMA TIMEOUT follows shortly after.  0x11 is the killer.

I suspect that the two cases need to be handled differently.  The
difference is in bit 1 (0x02), which is set when the error is about to
be successfully recovered, and clear when things are about to go bad.

Bit 1 of DMASTATUS is described differently in different places either
as a positive "write finished", or an inverted "write busy".  If we
take the first definition, then when an error arises with state 0x11,
it means that the write did not complete.   It makes sense to start a
new transfer, as in the current code.  But if we take the second
definition, then 0x11 means "an error but the write engine is still
busy".  Trying to feed it a new transfer in this situation might not be
a good idea.

As an experiment, I added code to ignore the DMA ERROR IRQ if DMASTATUS
is 0x11.  I.e., don't start a new transfer, don't clear our flags, etc.
The hope was that the card would complete the transfer and issue a ENC
DMA COMPLETE, either successfully or with an error condition there.
However the card still hung.

The only remaining corrective action being taken with a 0x11 status was
then the write back to the status register to clear the error, i.e.
DMASTATUS = DMASTATUS & ~3.  This would have the effect of clearing the
error bit 4, while leaving the lower bits indicating DMA write busy.

Strangely enough, removing this write to the status register solved the
problem!  If the DMA ERROR IRQ with DMASTATUS=0x11 is completely
ignored, with no corrective action at all, then the card will complete
the transfer and issue a new IRQ.  If the status register is written to
when it has the value 0x11, then the DMA engine hangs.  Perhaps it's
illegal to write to
DMASTATUS while the read or write busy bit is set?  At any rate, it
appears that the current corrective action is indeed making things
worse rather than better.

I put together a patch that modifies ivtv_irq_dma_err to do the
following:

- Don't write back to IVTV_REG_DMASTATUS.
- If write-busy is asserted, leave the card alone.  Just extend the
timeout slightly.
- If write-busy is de-asserted, retry the current transfer.

This has completely fixed my DMA TIMEOUT woes.  DMA ERR events still
occur, but now they seem to be correctly handled.  0x11 events no
longer hang the card, and 0x13 events no longer result in a glitch in
the stream, as the failed transfer is retried.  I'm happy.

I've inlined the patch below in case it is of interest.  As described
above, I have a theory about why it works (based on a different
interpretation of bit 1 of DMASTATUS), but I can't guarantee that my
theory is correct.  There may be another explanation, or it may be a
fluke.  Maybe ignoring that IRQ entirely would be equally effective?
Maybe the status register read/writeback sequence is race condition if
the card changes it in the mean time?  Also as I am using a PVR-150
only, I have not been able to test it on other cards, which may be
especially relevant for 350s that support concurrent decoding.
Hopefully the patch does not break the DMA READ path.

Mike

[awalls@md.metrocast.net: Modified patch to add a verbose comment, make minor
brace reformats, and clear the error flags in the IVTV_REG_DMASTATUS iff both
read and write DMA were not in progress.  Mike's conjecture about a race
condition with the writeback is correct; it can confuse the DMA engine.]

[Comment and analysis from the ML post by Michael <mike@rsy.com>]
Signed-off-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agosched: Fix sched rt group scheduling when hierachy is enabled
Balbir Singh [Thu, 3 Mar 2011 11:34:35 +0000 (17:04 +0530)]
sched: Fix sched rt group scheduling when hierachy is enabled

commit 0c3b9168017cbad2c4af3dd65ec93fe646eeaa62 upstream.

The current sched rt code is broken when it comes to hierarchical
scheduling, this patch fixes two problems

1. It adds redundant enqueuing (harmless) when it finds a queue
   has tasks enqueued, but it has no run time and it is not
   throttled.

2. The most important change is in sched_rt_rq_enqueue/dequeue.
   The code just picks the rt_rq belonging to the current cpu
   on which the period timer runs, the patch fixes it, so that
   the correct rt_se is enqueued/dequeued.

Tested with a simple hierarchy

/c/d, c and d assigned similar runtimes of 50,000 and a while
1 loop runs within "d". Both c and d get throttled, without
the patch, the task just stops running and never runs (depends
on where the sched_rt b/w timer runs). With the patch, the
task is throttled and runs as expected.

[ bharata, suggestions on how to pick the rt_se belong to the
  rt_rq and correct cpu ]

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: Bharata B Rao <bharata@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20110303113435.GA2868@balbir.in.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agodrivers/net: Call netif_carrier_off at the end of the probe
Ivan Vecera [Tue, 15 Feb 2011 02:08:39 +0000 (02:08 +0000)]
drivers/net: Call netif_carrier_off at the end of the probe

commit 0d672e9f8ac320c6d1ea9103db6df7f99ea20361 upstream.

Without calling of netif_carrier_off at the end of the probe the operstate
is unknown when the device is initially opened. By default the carrier is
on so when the device is opened and netif_carrier_on is called the link
watch event is not fired and operstate remains zero (unknown).

This patch fixes this behavior in forcedeth and r8169.

Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Acked-by: Francois Romieu <romieu@fr.zoreil.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agor8169: prevent RxFIFO induced loops in the irq handler.
Francois Romieu [Thu, 3 Feb 2011 16:27:52 +0000 (17:27 +0100)]
r8169: prevent RxFIFO induced loops in the irq handler.

commit f60ac8e7ab7cbb413a0131d5665b053f9f386526 upstream.

While the RxFIFO interruption is masked for most 8168, nothing prevents
it to appear in the irq status word. This is no excuse to crash.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Ivan Vecera <ivecera@redhat.com>
Cc: Hayes <hayeswang@realtek.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agor8169: RxFIFO overflow oddities with 8168 chipsets.
Francois Romieu [Thu, 3 Feb 2011 11:02:36 +0000 (12:02 +0100)]
r8169: RxFIFO overflow oddities with 8168 chipsets.

commit 1519e57fe81c14bb8fa4855579f19264d1ef63b4 upstream.

Some experiment-based action to prevent my 8168 chipsets locking-up hard
in the irq handler under load (pktgen ~1Mpps). Apparently a reset is not
always mandatory (is it at all ?).

- RTL_GIGA_MAC_VER_12
- RTL_GIGA_MAC_VER_25
  Missed ~55% packets. Note:
  - this is an old SiS 965L motherboard
  - the 8168 chipset emits (lots of) control frames towards the sender

- RTL_GIGA_MAC_VER_26
  The chipset does not go into a frenzy of mac control pause when it
  crashes yet but it can still be crashed. It needs more work.

Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Ivan Vecera <ivecera@redhat.com>
Cc: Hayes <hayeswang@realtek.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agor8169: use RxFIFO overflow workaround for 8168c chipset.
Ivan Vecera [Thu, 27 Jan 2011 11:24:11 +0000 (12:24 +0100)]
r8169: use RxFIFO overflow workaround for 8168c chipset.

commit b5ba6d12bdac21bc0620a5089e0f24e362645efd upstream.

I found that one of the 8168c chipsets (concretely XID 1c4000c0) starts
generating RxFIFO overflow errors. The result is an infinite loop in
interrupt handler as the RxFIFOOver is handled only for ...MAC_VER_11.
With the workaround everything goes fine.

Signed-off-by: Ivan Vecera <ivecera@redhat.com>
Acked-by: Francois Romieu <romieu@fr.zoreil.com>
Cc: Hayes <hayeswang@realtek.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agonetfilter: arpt_mangle: fix return values of checkentry
Pablo Neira Ayuso [Tue, 1 Feb 2011 15:03:46 +0000 (16:03 +0100)]
netfilter: arpt_mangle: fix return values of checkentry

commit 9d0db8b6b1da9e3d4c696ef29449700c58d589db upstream.

In 135367b "netfilter: xtables: change xt_target.checkentry return type",
the type returned by checkentry was changed from boolean to int, but the
return values where not adjusted.

arptables: Input/output error

This broke arptables with the mangle target since it returns true
under success, which is interpreted by xtables as >0, thus
returning EIO.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agonet: don't allow CAP_NET_ADMIN to load non-netdev kernel modules
Vasiliy Kulikov [Tue, 1 Mar 2011 21:33:13 +0000 (00:33 +0300)]
net: don't allow CAP_NET_ADMIN to load non-netdev kernel modules

commit 8909c9ad8ff03611c9c96c9a92656213e4bb495b upstream.

Since a8f80e8ff94ecba629542d9b4b5f5a8ee3eb565c any process with
CAP_NET_ADMIN may load any module from /lib/modules/.  This doesn't mean
that CAP_NET_ADMIN is a superset of CAP_SYS_MODULE as modules are
limited to /lib/modules/**.  However, CAP_NET_ADMIN capability shouldn't
allow anybody load any module not related to networking.

This patch restricts an ability of autoloading modules to netdev modules
with explicit aliases.  This fixes CVE-2011-1019.

Arnd Bergmann suggested to leave untouched the old pre-v2.6.32 behavior
of loading netdev modules by name (without any prefix) for processes
with CAP_SYS_MODULE to maintain the compatibility with network scripts
that use autoloading netdev modules by aliases like "eth0", "wlan0".

Currently there are only three users of the feature in the upstream
kernel: ipip, ip_gre and sit.

    root@albatros:~# capsh --drop=$(seq -s, 0 11),$(seq -s, 13 34) --
    root@albatros:~# grep Cap /proc/$$/status
    CapInh: 0000000000000000
    CapPrm: fffffff800001000
    CapEff: fffffff800001000
    CapBnd: fffffff800001000
    root@albatros:~# modprobe xfs
    FATAL: Error inserting xfs
    (/lib/modules/2.6.38-rc6-00001-g2bf4ca3/kernel/fs/xfs/xfs.ko): Operation not permitted
    root@albatros:~# lsmod | grep xfs
    root@albatros:~# ifconfig xfs
    xfs: error fetching interface information: Device not found
    root@albatros:~# lsmod | grep xfs
    root@albatros:~# lsmod | grep sit
    root@albatros:~# ifconfig sit
    sit: error fetching interface information: Device not found
    root@albatros:~# lsmod | grep sit
    root@albatros:~# ifconfig sit0
    sit0      Link encap:IPv6-in-IPv4
      NOARP  MTU:1480  Metric:1

    root@albatros:~# lsmod | grep sit
    sit                    10457  0
    tunnel4                 2957  1 sit

For CAP_SYS_MODULE module loading is still relaxed:

    root@albatros:~# grep Cap /proc/$$/status
    CapInh: 0000000000000000
    CapPrm: ffffffffffffffff
    CapEff: ffffffffffffffff
    CapBnd: ffffffffffffffff
    root@albatros:~# ifconfig xfs
    xfs: error fetching interface information: Device not found
    root@albatros:~# lsmod | grep xfs
    xfs                   745319  0

Reference: https://lkml.org/lkml/2011/2/24/203

Signed-off-by: Vasiliy Kulikov <segoon@openwall.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Kees Cook <kees.cook@canonical.com>
Signed-off-by: James Morris <jmorris@namei.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoipvs: fix dst_lock locking on dest update
Julian Anastasov [Tue, 22 Feb 2011 08:40:25 +0000 (10:40 +0200)]
ipvs: fix dst_lock locking on dest update

commit ff75f40f44ae9b79d520bf32a05d35af74a805c0 upstream.

Fix dst_lock usage in __ip_vs_update_dest. We need
_bh locking because destination is updated in user context.
Can cause lockups on frequent destination updates.
Problem reported by Simon Kirby. Bug was introduced
in 2.6.37 from the "ipvs: changes for local real server"
change.

Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: Simon Horman <horms@verge.net.au>
Cc: Simon Kirby <sim@hostway.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoNFS: NFSv4 readdir loses entries
Chuck Lever [Fri, 28 Jan 2011 17:41:05 +0000 (12:41 -0500)]
NFS: NFSv4 readdir loses entries

commit d1205f87bbb8040c1408bbd9e0a720310b2b0b9b upstream.

On recent 2.6.38-rc kernels, connectathon basic test 6 fails on
NFSv4 mounts of OpenSolaris with something like:

> ./test6: readdir
>  ./test6: (/mnt/klimt/matisse.test) didn't read expected 'file.12' dir entry, pass 0
>  ./test6: (/mnt/klimt/matisse.test) didn't read expected 'file.82' dir entry, pass 0
>  ./test6: (/mnt/klimt/matisse.test) didn't read expected 'file.164' dir entry, pass 0
>  ./test6: (/mnt/klimt/matisse.test) Test failed with 3 errors
> basic tests failed
> Tests failed, leaving /mnt/klimt mounted
> [cel@matisse cthon04]$

I narrowed the problem down to nfs4_decode_dirent() reporting that the
decode buffer had overflowed while decoding the entries for those
missing files.

verify_attr_len() assumes both it's pointer arguments reside on the
same page.  When these arguments point to locations on two different
pages, verify_attr_len() can report false errors.  This can happen now
that a large NFSv4 readdir result can span pages.

We have reasonably good checking in nfs4_decode_dirent() anyway, so
it should be safe to simply remove the extra checking.

At a guess, this was introduced by commit 6650239a, "NFS: Don't use
vm_map_ram() in readdir".

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoHID: hid-mosart: ignore buttons report
Benjamin Tissoires [Tue, 14 Dec 2010 10:38:18 +0000 (11:38 +0100)]
HID: hid-mosart: ignore buttons report

commit ad6d42670279da8f33f633f8a96a67cd7ef3b1da upstream.

This commit allows the device to be recognized as a touchscreen, and not a
touchpad by xf86-input-evdev.

The device has 2 modes. The first one is an emulation of a touchscreen by
sending left and right button, and the second mode is the one used in
dual-touch (sending trackingID, touch and else).

That's why there is a hid report containing left and right buttons
(9000001 and 9000002). The point is that xorg relies on these fields to
determine if it's a touchpad or a touchscreen.
Clearing the report (return -1) makes xorg detecting it out of the box
as a quite pleasant (dual)touchscreen.

Signed-off-by: Benjamin Tissoires <benjamin.tissoires@enac.fr>
Acked-by: Chase Douglas <chase.douglas@canonical.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Cc: James Sharam <james.sharam@googlemail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agonfsd: wrong index used in inner loop
roel [Tue, 8 Mar 2011 21:32:26 +0000 (22:32 +0100)]
nfsd: wrong index used in inner loop

commit 3ec07aa9522e3d5e9d5ede7bef946756e623a0a0 upstream.

Index i was already used in the outer loop

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agopcc-cpufreq: don't load driver if get_freq fails during init.
Naga Chumbalkar [Wed, 9 Mar 2011 14:02:49 +0000 (14:02 +0000)]
pcc-cpufreq: don't load driver if get_freq fails during init.

commit 1f858ef2fbabdc5e645644010a31a40c32e397c9 upstream.

Return 0 on failure. This will cause the initialization of the driver
to fail and prevent the driver from loading if the BIOS cannot handle
the PCC interface command to "get frequency". Otherwise, the driver
will load and display a very high value like "4294967274" (which is
actually -EINVAL) for frequency:

# cat /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq
4294967274

Signed-off-by: Naga Chumbalkar <nagananda.chumbalkar@hp.com>
Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agonetfilter: nf_log: avoid oops in (un)bind with invalid nfproto values
Jan Engelhardt [Wed, 2 Mar 2011 11:10:13 +0000 (12:10 +0100)]
netfilter: nf_log: avoid oops in (un)bind with invalid nfproto values

commit 9ef0298a8e5730d9a46d640014c727f3b4152870 upstream.

Like many other places, we have to check that the array index is
within allowed limits, or otherwise, a kernel oops and other nastiness
can ensue when we access memory beyond the end of the array.

[ 5954.115381] BUG: unable to handle kernel paging request at 0000004000000000
[ 5954.120014] IP:  __find_logger+0x6f/0xa0
[ 5954.123979]  nf_log_bind_pf+0x2b/0x70
[ 5954.123979]  nfulnl_recv_config+0xc0/0x4a0 [nfnetlink_log]
[ 5954.123979]  nfnetlink_rcv_msg+0x12c/0x1b0 [nfnetlink]
...

The problem goes back to v2.6.30-rc1~1372~1342~31 where nf_log_bind
was decoupled from nf_log_register.

Reported-by: Miguel Di Ciurcio Filho <miguel.filho@gmail.com>,
  via irc.freenode.net/#netfilter
Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agomm: fix possible cause of a page_mapped BUG
Hugh Dickins [Thu, 24 Feb 2011 05:39:49 +0000 (21:39 -0800)]
mm: fix possible cause of a page_mapped BUG

commit a3e8cc643d22d2c8ed36b9be7d9c9ca21efcf7f7 upstream.

Robert Swiecki reported a BUG_ON(page_mapped) from a fuzzer, punching
a hole with madvise(,, MADV_REMOVE).  That path is under mutex, and
cannot be explained by lack of serialization in unmap_mapping_range().

Reviewing the code, I found one place where vm_truncate_count handling
should have been updated, when I switched at the last minute from one
way of managing the restart_addr to another: mremap move changes the
virtual addresses, so it ought to adjust the restart_addr.

But rather than exporting the notion of restart_addr from memory.c, or
converting to restart_pgoff throughout, simply reset vm_truncate_count
to 0 to force a rescan if mremap move races with preempted truncation.

We have no confirmation that this fixes Robert's BUG,
but it is a fix that's worth making anyway.

Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Kerin Millar <kerframil@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoixgbe: fix for 82599 erratum on Header Splitting
Don Skidmore [Tue, 18 Jan 2011 22:53:47 +0000 (22:53 +0000)]
ixgbe: fix for 82599 erratum on Header Splitting

commit a124339ad28389093ed15eca990d39c51c5736cc upstream.

We have found a hardware erratum on 82599 hardware that can lead to
unpredictable behavior when Header Splitting mode is enabled.  So
we are no longer enabling this feature on affected hardware.

Please see the 82599 Specification Update for more information.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Stephen Ko <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoath9k: Fix ath9k prevents CPU to enter C3 states
Mohammed Shafi Shajakhan [Fri, 4 Mar 2011 22:00:30 +0000 (23:00 +0100)]
ath9k: Fix ath9k prevents CPU to enter C3 states

This is a backport of upstream commit 0f5cd45960173ba5b36727decbb4a241cbd35ef9.

The DMA latency issue is observed only in Intel pinetrail platforms
but in the driver we had a default PM-QOS value of 55. This caused
unnecessary power consumption and battery drain in other platforms.

Remove the pm-qos thing in the driver code and address the throughput
issue in Intel pinetrail platfroms in user space using any one of
the scripts in below links:

http://www.kernel.org/pub/linux/kernel/people/mcgrof/scripts/cpudmalatency.c
http://johannes.sipsolutions.net/files/netlatency.c.txt

More details can be found in the following bugzilla link:

https://bugzilla.kernel.org/show_bug.cgi?id=27532

Signed-off-by: Thomas Bächler <thomas@archlinux.org>
Acked-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoRxRPC: Fix v1 keys
Anton Blanchard [Mon, 28 Feb 2011 03:27:53 +0000 (03:27 +0000)]
RxRPC: Fix v1 keys

commit f009918a1c1bbf8607b8aab3959876913a30193a upstream.

commit 339412841d7 (RxRPC: Allow key payloads to be passed in XDR form)
broke klog for me. I notice the v1 key struct had a kif_version field
added:

-struct rxkad_key {
-       u16     security_index;         /* RxRPC header security index */
-       u16     ticket_len;             /* length of ticket[] */
-       u32     expiry;                 /* time at which expires */
-       u32     kvno;                   /* key version number */
-       u8      session_key[8];         /* DES session key */
-       u8      ticket[0];              /* the encrypted ticket */
-};

+struct rxrpc_key_data_v1 {
+       u32             kif_version;            /* 1 */
+       u16             security_index;
+       u16             ticket_length;
+       u32             expiry;                 /* time_t */
+       u32             kvno;
+       u8              session_key[8];
+       u8              ticket[0];
+};

However the code in rxrpc_instantiate strips it away:

data += sizeof(kver);
datalen -= sizeof(kver);

Removing kif_version fixes my problem.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agonfs4: Ensure that ACL pages sent over NFS were not allocated from the slab (v3)
Neil Horman [Sat, 5 Mar 2011 00:26:03 +0000 (19:26 -0500)]
nfs4: Ensure that ACL pages sent over NFS were not allocated from the slab (v3)

commit e9e3d724e2145f5039b423c290ce2b2c3d8f94bc upstream.

The "bad_page()" page allocator sanity check was reported recently (call
chain as follows):

  bad_page+0x69/0x91
  free_hot_cold_page+0x81/0x144
  skb_release_data+0x5f/0x98
  __kfree_skb+0x11/0x1a
  tcp_ack+0x6a3/0x1868
  tcp_rcv_established+0x7a6/0x8b9
  tcp_v4_do_rcv+0x2a/0x2fa
  tcp_v4_rcv+0x9a2/0x9f6
  do_timer+0x2df/0x52c
  ip_local_deliver+0x19d/0x263
  ip_rcv+0x539/0x57c
  netif_receive_skb+0x470/0x49f
  :virtio_net:virtnet_poll+0x46b/0x5c5
  net_rx_action+0xac/0x1b3
  __do_softirq+0x89/0x133
  call_softirq+0x1c/0x28
  do_softirq+0x2c/0x7d
  do_IRQ+0xec/0xf5
  default_idle+0x0/0x50
  ret_from_intr+0x0/0xa
  default_idle+0x29/0x50
  cpu_idle+0x95/0xb8
  start_kernel+0x220/0x225
  _sinittext+0x22f/0x236

It occurs because an skb with a fraglist was freed from the tcp
retransmit queue when it was acked, but a page on that fraglist had
PG_Slab set (indicating it was allocated from the Slab allocator (which
means the free path above can't safely free it via put_page.

We tracked this back to an nfsv4 setacl operation, in which the nfs code
attempted to fill convert the passed in buffer to an array of pages in
__nfs4_proc_set_acl, which gets used by the skb->frags list in
xs_sendpages.  __nfs4_proc_set_acl just converts each page in the buffer
to a page struct via virt_to_page, but the vfs allocates the buffer via
kmalloc, meaning the PG_slab bit is set.  We can't create a buffer with
kmalloc and free it later in the tcp ack path with put_page, so we need
to either:

1) ensure that when we create the list of pages, no page struct has
   PG_Slab set

 or

2) not use a page list to send this data

Given that these buffers can be multiple pages and arbitrarily sized, I
think (1) is the right way to go.  I've written the below patch to
allocate a page from the buddy allocator directly and copy the data over
to it.  This ensures that we have a put_page free-able page for every
entry that winds up on an skb frag list, so it can be safely freed when
the frame is acked.  We do a put page on each entry after the
rpc_call_sync call so as to drop our own reference count to the page,
leaving only the ref count taken by tcp_sendpages.  This way the data
will be properly freed when the ack comes in

Successfully tested by myself to solve the above oops.

Note, as this is the result of a setacl operation that exceeded a page
of data, I think this amounts to a local DOS triggerable by an
uprivlidged user, so I'm CCing security on this as well.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
CC: Trond Myklebust <Trond.Myklebust@netapp.com>
CC: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agodrivers/misc/bmp085.c: add MODULE_DEVICE_TABLE
Axel Lin [Sat, 5 Mar 2011 01:36:22 +0000 (17:36 -0800)]
drivers/misc/bmp085.c: add MODULE_DEVICE_TABLE

commit 97e419a082461f8a3a0818834eb88ad41219a1da upstream.

The device table is required to load modules based on modaliases.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Cc: Shubhrajyoti D <shubhrajyoti@ti.com>
Cc: Christoph Mair <christoph.mair@gmail.com>
Cc: Jonathan Cameron <jic23@cam.ac.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoALSA: hda - Don't set to D3 in Cirrus errata init verbs
Takashi Iwai [Thu, 3 Mar 2011 13:54:19 +0000 (14:54 +0100)]
ALSA: hda - Don't set to D3 in Cirrus errata init verbs

commit 38c07641905c0db58e800ea974cd9158717c6610 upstream.

The errata init verbs for CS42xx codecs contain the verbs to set
the power-state of SPDIF nodes to D3, which seem to break the SPDIF
output on some MacBooks.  Since this is executed during the power-up
initialization, we shouldn't turn them down there.

Reported-by: Arun Raghavan <arun.raghavan@collabora.co.uk>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoALSA: HDA: Realtek: Fixup jack detection to input subsystem
David Henningsson [Mon, 28 Feb 2011 14:58:07 +0000 (15:58 +0100)]
ALSA: HDA: Realtek: Fixup jack detection to input subsystem

commit f0ce27996217d06207c8bfda1b1bbec2fbab48c6 upstream.

This patch fixes an error in the jack detection reporting,
causing the jack detection sometimes not to be reported
correctly to the input subsystem. It should apply to several
Realtek codecs.

Signed-off-by: David Henningsson <david.henningsson@canonical.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoASoC: Fix WM9081 platform data initialisation
Mark Brown [Tue, 1 Mar 2011 20:05:23 +0000 (20:05 +0000)]
ASoC: Fix WM9081 platform data initialisation

commit 3ee845acba58549578d03a46ed307c0a56c7f777 upstream.

It went AWOL in the multi-component conversion.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agokeyboard: integer underflow bug
Dan Carpenter [Thu, 3 Mar 2011 16:56:06 +0000 (17:56 +0100)]
keyboard: integer underflow bug

commit b652277b09d3d030cb074cc6a98ba80b34244c03 upstream.

The "ct" variable should be an unsigned int.  Both struct kbdiacrs
->kb_cnt and struct kbd_data ->accent_table_size are unsigned ints.

Making it signed causes a problem in KBDIACRUC because the user could
set the signed bit and cause a buffer overflow.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agovirtio: console: Don't access vqs if device was unplugged
Amit Shah [Fri, 4 Mar 2011 03:34:33 +0000 (14:04 +1030)]
virtio: console: Don't access vqs if device was unplugged

commit d7a62cd0332115d4c7c4689abea0d889a30d8349 upstream.

If a virtio-console device gets unplugged while a port is open, a
subsequent close() call on the port accesses vqs to free up buffers.
This can lead to a crash.

The buffers are already freed up as a result of the call to
unplug_ports() from virtcons_remove().  The fix is to simply not access
vq information if port->portdev is NULL.

Reported-by: juzhang <juzhang@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agocpuset: add a missing unlock in cpuset_write_resmask()
Li Zefan [Sat, 5 Mar 2011 01:36:21 +0000 (17:36 -0800)]
cpuset: add a missing unlock in cpuset_write_resmask()

commit b75f38d659e6fc747eda64cb72f3920e29dd44a4 upstream.

Don't forget to release cgroup_mutex if alloc_trial_cpuset() fails.

[akpm@linux-foundation.org: avoid multiple return points]
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoLinux 2.6.37.3 v2.6.37.3
Greg Kroah-Hartman [Mon, 7 Mar 2011 23:05:50 +0000 (15:05 -0800)]
Linux 2.6.37.3

13 years agoarp_notify: unconditionally send gratuitous ARP for NETDEV_NOTIFY_PEERS.
Ian Campbell [Fri, 11 Feb 2011 07:44:16 +0000 (07:44 +0000)]
arp_notify: unconditionally send gratuitous ARP for NETDEV_NOTIFY_PEERS.

commit d11327ad6695db8117c78d70611e71102ceec2ac upstream.

NETDEV_NOTIFY_PEER is an explicit request by the driver to send a link
notification while NETDEV_UP/NETDEV_CHANGEADDR generate link
notifications as a sort of side effect.

In the later cases the sysctl option is present because link
notification events can have undesired effects e.g. if the link is
flapping. I don't think this applies in the case of an explicit
request from a driver.

This patch makes NETDEV_NOTIFY_PEER unconditional, if preferred we
could add a new sysctl for this case which defaults to on.

This change causes Xen post-migration ARP notifications (which cause
switches to relearn their MAC tables etc) to be sent by default.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
[reported to solve hyperv live migration problem - gkh]
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Mike Surcouf <mike@surcouf.co.uk>
Cc: Hank Janssen <hjanssen@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoDNS: Fix a NULL pointer deref when trying to read an error key [CVE-2011-1076]
David Howells [Thu, 3 Mar 2011 11:28:58 +0000 (11:28 +0000)]
DNS: Fix a NULL pointer deref when trying to read an error key [CVE-2011-1076]

commit 1362fa078dae16776cd439791c6605b224ea6171 upstream.

When a DNS resolver key is instantiated with an error indication, attempts to
read that key will result in an oops because user_read() is expecting there to
be a payload - and there isn't one [CVE-2011-1076].

Give the DNS resolver key its own read handler that returns the error cached in
key->type_data.x[0] as an error rather than crashing.

Also make the kenter() at the beginning of dns_resolver_instantiate() limit the
amount of data it prints, since the data is not necessarily NUL-terminated.

The buggy code was added in:

commit 4a2d789267e00b5a1175ecd2ddefcc78b83fbf09
Author: Wang Lei <wang840925@gmail.com>
Date:   Wed Aug 11 09:37:58 2010 +0100
Subject: DNS: If the DNS server returns an error, allow that to be cached [ver #2]

This can trivially be reproduced by any user with the following program
compiled with -lkeyutils:

#include <stdlib.h>
#include <keyutils.h>
#include <err.h>
static char payload[] = "#dnserror=6";
int main()
{
key_serial_t key;
key = add_key("dns_resolver", "a", payload, sizeof(payload),
      KEY_SPEC_SESSION_KEYRING);
if (key == -1)
err(1, "add_key");
if (keyctl_read(key, NULL, 0) == -1)
err(1, "read_key");
return 0;
}

What should happen is that keyctl_read() reports error 6 (ENXIO) to the user:

dns-break: read_key: No such device or address

but instead the kernel oopses.

This cannot be reproduced with the 'keyutils add' or 'keyutils padd' commands
as both of those cut the data down below the NUL termination that must be
included in the data.  Without this dns_resolver_instantiate() will return
-EINVAL and the key will not be instantiated such that it can be read.

The oops looks like:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
IP: [<ffffffff811b99f7>] user_read+0x4f/0x8f
PGD 3bdf8067 PUD 385b9067 PMD 0
Oops: 0000 [#1] SMP
last sysfs file: /sys/devices/pci0000:00/0000:00:19.0/irq
CPU 0
Modules linked in:

Pid: 2150, comm: dns-break Not tainted 2.6.38-rc7-cachefs+ #468                  /DG965RY
RIP: 0010:[<ffffffff811b99f7>]  [<ffffffff811b99f7>] user_read+0x4f/0x8f
RSP: 0018:ffff88003bf47f08  EFLAGS: 00010246
RAX: 0000000000000001 RBX: ffff88003b5ea378 RCX: ffffffff81972368
RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff88003b5ea378
RBP: ffff88003bf47f28 R08: ffff88003be56620 R09: 0000000000000000
R10: 0000000000000395 R11: 0000000000000002 R12: 0000000000000000
R13: 0000000000000000 R14: 0000000000000000 R15: ffffffffffffffa1
FS:  00007feab5751700(0000) GS:ffff88003e000000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000000000010 CR3: 000000003de40000 CR4: 00000000000006f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process dns-break (pid: 2150, threadinfo ffff88003bf46000, task ffff88003be56090)
Stack:
 ffff88003b5ea378 ffff88003b5ea3a0 0000000000000000 0000000000000000
 ffff88003bf47f68 ffffffff811b708e ffff88003c442bc8 0000000000000000
 00000000004005a0 00007fffba368060 0000000000000000 0000000000000000
Call Trace:
 [<ffffffff811b708e>] keyctl_read_key+0xac/0xcf
 [<ffffffff811b7c07>] sys_keyctl+0x75/0xb6
 [<ffffffff81001f7b>] system_call_fastpath+0x16/0x1b
Code: 75 1f 48 83 7b 28 00 75 18 c6 05 58 2b fb 00 01 be bb 00 00 00 48 c7 c7 76 1c 75 81 e8 13 c2 e9 ff 4c 8b b3 e0 00 00 00 4d 85 ed <41> 0f b7 5e 10 74 2d 4d 85 e4 74 28 e8 98 79 ee ff 49 39 dd 48
RIP  [<ffffffff811b99f7>] user_read+0x4f/0x8f
 RSP <ffff88003bf47f08>
CR2: 0000000000000010

Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Jeff Layton <jlayton@redhat.com>
cc: Wang Lei <wang840925@gmail.com>
Signed-off-by: James Morris <jmorris@namei.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoe1000e: disable broken PHY wakeup for ICH10 LOMs, use MAC wakeup instead
Bruce Allan [Wed, 2 Feb 2011 09:30:36 +0000 (09:30 +0000)]
e1000e: disable broken PHY wakeup for ICH10 LOMs, use MAC wakeup instead

commit 4def99bbfd46e05c5e03b5b282cb4ee30e27ff19 upstream.

When support for 82577/82578 was added[1] in 2.6.31, PHY wakeup was in-
advertently enabled (even though it does not function properly) on ICH10
LOMs.  This patch makes it so that the ICH10 LOMs use MAC wakeup instead
as was done with the initial support for those devices (i.e. 82567LM-3,
82567LF-3 and 82567V-4).

[1] commit a4f58f5455ba0efda36fb33c37074922d1527a10

Reported-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agodccp: fix oops on Reset after close
Gerrit Renker [Wed, 2 Mar 2011 07:02:07 +0000 (23:02 -0800)]
dccp: fix oops on Reset after close

commit 720dc34bbbe9493c7bd48b2243058b4e447a929d upstream.

This fixes a bug in the order of dccp_rcv_state_process() that still permitted
reception even after closing the socket. A Reset after close thus causes a NULL
pointer dereference by not preventing operations on an already torn-down socket.

 dccp_v4_do_rcv()
|
| state other than OPEN
v
 dccp_rcv_state_process()
|
| DCCP_PKT_RESET
v
 dccp_rcv_reset()
|
v
 dccp_time_wait()

 WARNING: at net/ipv4/inet_timewait_sock.c:141 __inet_twsk_hashdance+0x48/0x128()
 Modules linked in: arc4 ecb carl9170 rt2870sta(C) mac80211 r8712u(C) crc_ccitt ah
 [<c0038850>] (unwind_backtrace+0x0/0xec) from [<c0055364>] (warn_slowpath_common)
 [<c0055364>] (warn_slowpath_common+0x4c/0x64) from [<c0055398>] (warn_slowpath_n)
 [<c0055398>] (warn_slowpath_null+0x1c/0x24) from [<c02b72d0>] (__inet_twsk_hashd)
 [<c02b72d0>] (__inet_twsk_hashdance+0x48/0x128) from [<c031caa0>] (dccp_time_wai)
 [<c031caa0>] (dccp_time_wait+0x40/0xc8) from [<c031c15c>] (dccp_rcv_state_proces)
 [<c031c15c>] (dccp_rcv_state_process+0x120/0x538) from [<c032609c>] (dccp_v4_do_)
 [<c032609c>] (dccp_v4_do_rcv+0x11c/0x14c) from [<c0286594>] (release_sock+0xac/0)
 [<c0286594>] (release_sock+0xac/0x110) from [<c031fd34>] (dccp_close+0x28c/0x380)
 [<c031fd34>] (dccp_close+0x28c/0x380) from [<c02d9a78>] (inet_release+0x64/0x70)

The fix is by testing the socket state first. Receiving a packet in Closed state
now also produces the required "No connection" Reset reply of RFC 4340, 8.3.1.

Reported-and-tested-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agor8169: disable ASPM
Stanislaw Gruszka [Tue, 22 Feb 2011 02:00:11 +0000 (02:00 +0000)]
r8169: disable ASPM

commit ba04c7c93bbcb48ce880cf75b6e9dffcd79d4c7b upstream.

For some time is known that ASPM is causing troubles on r8169, i.e. make
device randomly stop working without any errors in dmesg.

Currently Tomi Leppikangas reports that system with r8169 device hangs
with MCE errors when ASPM is enabled:
https://bugzilla.redhat.com/show_bug.cgi?id=642861#c4

Lets disable ASPM for r8169 devices at all, to avoid problems with
r8169 PCIe devices at least for some users.

Reported-by: Tomi Leppikangas <tomi.leppikangas@gmail.com>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agocarl9170: add Airlive X.USB a/b/g/n USBID
Jan Puk [Tue, 22 Feb 2011 13:49:43 +0000 (14:49 +0100)]
carl9170: add Airlive X.USB a/b/g/n USBID

commit c86664e5a285af1afa06416e450e7c4af04daa7c upstream.

"AirLive X.USB now works perfectly under a Linux
environment!"

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agonilfs2: fix regression that i-flag is not set on changeless checkpoints
Ryusuke Konishi [Mon, 28 Feb 2011 04:41:11 +0000 (13:41 +0900)]
nilfs2: fix regression that i-flag is not set on changeless checkpoints

commit 72746ac643928f6c3113b5aa783d8ea1b13949d2 upstream.

According to the report from Jiro SEKIBA titled "regression in
2.6.37?"  (Message-Id: <8739n8vs1f.wl%jir@sekiba.com>), on 2.6.37 and
later kernels, lscp command no longer displays "i" flag on checkpoints
that snapshot operations or garbage collection created.

This is a regression of nilfs2 checkpointing function, and it's
critical since it broke behavior of a part of nilfs2 applications.
For instance, snapshot manager of TimeBrowse gets to create
meaningless snapshots continuously; snapshot creation triggers another
checkpoint, but applications cannot distinguish whether the new
checkpoint contains meaningful changes or not without the i-flag.

This patch fixes the regression and brings that application behavior
back to normal.

Reported-by: Jiro SEKIBA <jir@unicus.jp>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Tested-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp>
Tested-by: Jiro SEKIBA <jir@unicus.jp>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agop54usb: add Senao NUB-350 usbid
Christian Lamparter [Sat, 26 Feb 2011 11:58:06 +0000 (12:58 +0100)]
p54usb: add Senao NUB-350 usbid

commit 2b799a6b25bb9f9fbc478782cd9503e8066ab618 upstream.

Reported-by: Mark Davis
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoath9k_htc: Fix an endian issue
Sujith Manoharan [Sun, 27 Feb 2011 03:53:52 +0000 (09:23 +0530)]
ath9k_htc: Fix an endian issue

commit 2c27392dc4d4f5ee8a3967a520b8f6cac0418031 upstream.

The stream length/tag fields have to be in little endian
format. Fixing this makes the driver work on big-endian
platforms.

Tested-by: raghunathan.kailasanathan@wipro.com
Signed-off-by: Sujith Manoharan <Sujith.Manoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoblock: kill loop_mutex
Petr Uzel [Thu, 3 Mar 2011 16:48:50 +0000 (11:48 -0500)]
block: kill loop_mutex

commit fd51469fb68b987032e46297e0a4fe9020063c20 upstream.

Following steps lead to deadlock in kernel:

dd if=/dev/zero of=img bs=512 count=1000
losetup -f img
mkfs.ext2 /dev/loop0
mount -t ext2 -o loop /dev/loop0 mnt
umount mnt/

Stacktrace:
[<c102ec04>] irq_exit+0x36/0x59
[<c101502c>] smp_apic_timer_interrupt+0x6b/0x75
[<c127f639>] apic_timer_interrupt+0x31/0x38
[<c101df88>] mutex_spin_on_owner+0x54/0x5b
[<fe2250e9>] lo_release+0x12/0x67 [loop]
[<c10c4eae>] __blkdev_put+0x7c/0x10c
[<c10a4da5>] fput+0xd5/0x1aa
[<fe2250cf>] loop_clr_fd+0x1a9/0x1b1 [loop]
[<fe225110>] lo_release+0x39/0x67 [loop]
[<c10c4eae>] __blkdev_put+0x7c/0x10c
[<c10a59d9>] deactivate_locked_super+0x17/0x36
[<c10b6f37>] sys_umount+0x27e/0x2a5
[<c10b6f69>] sys_oldumount+0xb/0xe
[<c1002897>] sysenter_do_call+0x12/0x26
[<ffffffff>] 0xffffffff

Regression since 2a48fc0ab24241755dc9, which introduced the private
loop_mutex as part of the BKL removal process.

As per [1], the mutex can be safely removed.

[1] http://www.gossamer-threads.com/lists/linux/kernel/1341930

Addresses: https://bugzilla.novell.com/show_bug.cgi?id=669394
Addresses: https://bugzilla.kernel.org/show_bug.cgi?id=29172

Signed-off-by: Petr Uzel <petr.uzel@suse.cz>
Reviewed-by: Nikanth Karthikesan <knikanth@suse.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoblock: blk-flush shouldn't call directly into q->request_fn() __blk_run_queue()
Tejun Heo [Wed, 2 Mar 2011 13:48:06 +0000 (08:48 -0500)]
block: blk-flush shouldn't call directly into q->request_fn() __blk_run_queue()

commit 255bb490c8c27eed484d538efe6ef6a7473bd3f6 upstream.

blk-flush decomposes a flush into sequence of multiple requests.  On
completion of a request, the next one is queued; however, block layer
must not implicitly call into q->request_fn() directly from completion
path.  This makes the queue behave unexpectedly when seen from the
drivers and violates the assumption that q->request_fn() is called
with process context + queue_lock.

This patch makes blk-flush the following two changes to make sure
q->request_fn() is not called directly from request completion path.

- blk_flush_complete_seq_end_io() now asks __blk_run_queue() to always
  use kblockd instead of calling directly into q->request_fn().

- queue_next_fseq() uses ELEVATOR_INSERT_REQUEUE instead of
  ELEVATOR_INSERT_FRONT so that elv_insert() doesn't try to unplug the
  request queue directly.

Reported by Jan in the following threads.

 http://thread.gmane.org/gmane.linux.ide/48778
 http://thread.gmane.org/gmane.linux.ide/48786

stable: applicable to v2.6.37.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Jan Beulich <JBeulich@novell.com>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoblock: add @force_kblockd to __blk_run_queue()
Tejun Heo [Wed, 2 Mar 2011 13:48:05 +0000 (08:48 -0500)]
block: add @force_kblockd to __blk_run_queue()

commit 1654e7411a1ad4999fe7890ef51d2a2bbb1fcf76 upstream.

__blk_run_queue() automatically either calls q->request_fn() directly
or schedules kblockd depending on whether the function is recursed.
blk-flush implementation needs to be able to explicitly choose
kblockd.  Add @force_kblockd.

All the current users are converted to specify %false for the
parameter and this patch doesn't introduce any behavior change.

stable: This is prerequisite for fixing ide oops caused by the new
        blk-flush implementation.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Jan Beulich <JBeulich@novell.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agoblk-throttle: Do not use kblockd workqueue for throtl work
Vivek Goyal [Tue, 1 Mar 2011 18:40:54 +0000 (13:40 -0500)]
blk-throttle: Do not use kblockd workqueue for throtl work

commit 450adcbe518ab3a3953d8475309525d22de77cba upstream.

o Dominik Klein reported a system hang issue while doing some blkio
  throttling testing.

  https://lkml.org/lkml/2011/2/24/173

o Some tracing revealed that CFQ was not dispatching any more jobs as
  queue unplug was not happening. And queue unplug was not happening
  because unplug work was not being called as there was one throttling
  work on same cpu which as not finished yet. And throttling work had not
  finished as it was tyring to dispatch a bio to CFQ but all the request
  descriptors were consume to it was put to sleep.

o So basically it is a cyclic dependecny between CFQ unplug work and
  throtl dispatch work. Tejun suggested that use separate workqueue for
  such cases.

o This patch uses a separate workqueue for throttle related work and
  does not rely on kblockd workqueue anymore.

Reported-by: Dominik Klein <dk@in-telegence.net>
Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
13 years agodrm/i915: fix memory corruption with GM965 and >4GB RAM
Jan Niehusmann [Thu, 3 Mar 2011 23:28:02 +0000 (00:28 +0100)]
drm/i915: fix memory corruption with GM965 and >4GB RAM

commit 6927faf30920b8c03dfa007e732642a1f1f20089 upstream.

On a Thinkpad x61s, I noticed some memory corruption when
plugging/unplugging the external VGA connection. The symptoms are that
4 bytes at the beginning of a page get overwritten by zeroes.
The address of the corruption varies when rebooting the machine, but
stays constant while it's running (so it's possible to repeatedly write
some data and then corrupt it again by plugging the cable).

Further investigation revealed that the corrupted address is
(dev_priv->status_page_dmah->busaddr & 0xffffffff), ie. the beginning of
the hardware status page of the i965 graphics card, cut to 32 bits.

So it seems that for some memory access, the hardware uses only 32 bit
addressing. If the hardware status page is located >4GB, this
corrupts unrelated memory.

Signed-off-by: Jan Niehusmann <jan@gondor.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>