platform/kernel/linux-3.10.git
10 years agodrm/i915: Preserve the DDI_A_4_LANES bit from the bios
Stéphane Marchesin [Fri, 12 Jul 2013 20:54:41 +0000 (13:54 -0700)]
drm/i915: Preserve the DDI_A_4_LANES bit from the bios

commit bcf53de4e60d9000b82f541d654529e2902a4c2c upstream.

Otherwise the DDI_A_4_LANES bit gets lost and we can't use > 2 lanes
on eDP. This fixes eDP on hsw with > 2 lanes.

Also s/port_reversal/saved_port_bits/ since the current name is
confusing.

Signed-off-by: Stéphane Marchesin <marcheu@chromium.org>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Zhouping Liu <zliu@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoxen-blkfront: use a different scatterlist for each request
Roger Pau Monne [Thu, 2 May 2013 08:58:50 +0000 (10:58 +0200)]
xen-blkfront: use a different scatterlist for each request

commit b7649158a0d241f8d53d13ff7441858539e16656 upstream.

In blkif_queue_request blkfront iterates over the scatterlist in order
to set the segments of the request, and in blkif_completion blkfront
iterates over the raw request, which makes it hard to know the exact
position of the source and destination memory positions.

This can be solved by allocating a scatterlist for each request, that
will be keep until the request is finished, allowing us to copy the
data back to the original memory without having to iterate over the
raw request.

Oracle-Bug: 16660413 - LARGE ASYNCHRONOUS READS APPEAR BROKEN ON 2.6.39-400
CC: stable@vger.kernel.org
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reported-and-Tested-by: Anne Milicia <anne.milicia@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/radeon: Disable dma rings for bo moves on r6xx
Alex Deucher [Thu, 11 Jul 2013 18:20:11 +0000 (14:20 -0400)]
drm/radeon: Disable dma rings for bo moves on r6xx

commit aeea40cbf9388fc829e66fa049f64d97fd72e118 upstream.

They still seem to cause instability on some r6xx parts.
As a follow up, we can switch to using CP DMA for bo
moves on r6xx as a lighter weight alternative to using
the 3D engine.

A version of this patch should also go to stable kernels.

Tested-by: J.N. <golden.fleeced@gmail.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoPCI: Retry allocation of only the resource type that failed
Yinghai Lu [Thu, 25 Jul 2013 13:31:38 +0000 (06:31 -0700)]
PCI: Retry allocation of only the resource type that failed

commit aa914f5ec25e4371ba18b312971314be1b9b1076 upstream.

Ben Herrenschmidt reported the following problem:

  - The bus has space for all desired MMIO resources, including optional
    space for SR-IOV devices
  - We attempt to allocate I/O port space, but it fails because the bus
    has no I/O space
  - Because of the I/O allocation failure, we retry MMIO allocation,
    requesting only the required space, without the optional SR-IOV space

This means we don't allocate the optional SR-IOV space, even though we
could.

This is related to 0c5be0cb0e ("PCI: Retry on IORESOURCE_IO type
allocations").

This patch changes how we handle allocation failures.  We will now retry
allocation of only the resource type that failed.  If MMIO allocation
fails, we'll retry only MMIO allocation.  If I/O port allocation fails,
we'll retry only I/O port allocation.

[bhelgaas: changelog]
Reference: https://lkml.kernel.org/r/1367712653.11982.19.camel@pasglop
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Tested-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoPCI: pciehp: Fix null pointer deref when hot-removing SR-IOV device
Yinghai Lu [Fri, 19 Jul 2013 19:14:16 +0000 (12:14 -0700)]
PCI: pciehp: Fix null pointer deref when hot-removing SR-IOV device

commit 29ed1f29b68a8395d5679b3c4e38352b617b3236 upstream.

Hot-removing a device with SR-IOV enabled causes a null pointer dereference
in v3.9 and v3.10.

This is a regression caused by ba518e3c17 ("PCI: pciehp: Iterate over all
devices in slot, not functions 0-7").  When we iterate over the
bus->devices list, we first remove the PF, which also removes all the VFs
from the list.  Then the list iterator blows up because more than just the
current entry was removed from the list.

ac205b7bb7 ("PCI: make sriov work with hotplug remove") works around a
similar problem in pci_stop_bus_devices() by iterating over the list in
reverse, so the VFs are stopped and removed from the list first, before the
PF.

This patch changes pciehp_unconfigure_device() to iterate over the list in
reverse, too.

[bhelgaas: bugzilla, changelog]
Reference: https://bugzilla.kernel.org/show_bug.cgi?id=60604
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Yijing Wang <wangyijing@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoRevert "cpuidle: Quickly notice prediction failure for repeat mode"
Rafael J. Wysocki [Fri, 26 Jul 2013 23:41:34 +0000 (01:41 +0200)]
Revert "cpuidle: Quickly notice prediction failure for repeat mode"

commit 148519120c6d1f19ad53349683aeae9f228b0b8d upstream.

Revert commit 69a37bea (cpuidle: Quickly notice prediction failure for
repeat mode), because it has been identified as the source of a
significant performance regression in v3.8 and later as explained by
Jeremy Eder:

  We believe we've identified a particular commit to the cpuidle code
  that seems to be impacting performance of variety of workloads.
  The simplest way to reproduce is using netperf TCP_RR test, so
  we're using that, on a pair of Sandy Bridge based servers.  We also
  have data from a large database setup where performance is also
  measurably/positively impacted, though that test data isn't easily
  share-able.

  Included below are test results from 3 test kernels:

  kernel       reverts
  -----------------------------------------------------------
  1) vanilla   upstream (no reverts)

  2) perfteam2 reverts e11538d1f03914eb92af5a1a378375c05ae8520c

  3) test      reverts 69a37beabf1f0a6705c08e879bdd5d82ff6486c4
                       e11538d1f03914eb92af5a1a378375c05ae8520c

  In summary, netperf TCP_RR numbers improve by approximately 4%
  after reverting 69a37beabf1f0a6705c08e879bdd5d82ff6486c4.  When
  69a37beabf1f0a6705c08e879bdd5d82ff6486c4 is included, C0 residency
  never seems to get above 40%.  Taking that patch out gets C0 near
  100% quite often, and performance increases.

  The below data are histograms representing the %c0 residency @
  1-second sample rates (using turbostat), while under netperf test.

  - If you look at the first 4 histograms, you can see %c0 residency
    almost entirely in the 30,40% bin.
  - The last pair, which reverts 69a37beabf1f0a6705c08e879bdd5d82ff6486c4,
    shows %c0 in the 80,90,100% bins.

  Below each kernel name are netperf TCP_RR trans/s numbers for the
  particular kernel that can be disclosed publicly, comparing the 3
  test kernels.  We ran a 4th test with the vanilla kernel where
  we've also set /dev/cpu_dma_latency=0 to show overall impact
  boosting single-threaded TCP_RR performance over 11% above
  baseline.

  3.10-rc2 vanilla RX + c0 lock (/dev/cpu_dma_latency=0):
  TCP_RR trans/s 54323.78

  -----------------------------------------------------------
  3.10-rc2 vanilla RX (no reverts)
  TCP_RR trans/s 48192.47

  Receiver %c0
      0.0000 -    10.0000 [     1]: *
     10.0000 -    20.0000 [     0]:
     20.0000 -    30.0000 [     0]:
     30.0000 -    40.0000 [    59]:
  ***********************************************************
     40.0000 -    50.0000 [     1]: *
     50.0000 -    60.0000 [     0]:
     60.0000 -    70.0000 [     0]:
     70.0000 -    80.0000 [     0]:
     80.0000 -    90.0000 [     0]:
     90.0000 -   100.0000 [     0]:

  Sender %c0
      0.0000 -    10.0000 [     1]: *
     10.0000 -    20.0000 [     0]:
     20.0000 -    30.0000 [     0]:
     30.0000 -    40.0000 [    11]: ***********
     40.0000 -    50.0000 [    49]:
  *************************************************
     50.0000 -    60.0000 [     0]:
     60.0000 -    70.0000 [     0]:
     70.0000 -    80.0000 [     0]:
     80.0000 -    90.0000 [     0]:
     90.0000 -   100.0000 [     0]:

  -----------------------------------------------------------
  3.10-rc2 perfteam2 RX (reverts commit
  e11538d1f03914eb92af5a1a378375c05ae8520c)
  TCP_RR trans/s 49698.69

  Receiver %c0
      0.0000 -    10.0000 [     1]: *
     10.0000 -    20.0000 [     1]: *
     20.0000 -    30.0000 [     0]:
     30.0000 -    40.0000 [    59]:
  ***********************************************************
     40.0000 -    50.0000 [     0]:
     50.0000 -    60.0000 [     0]:
     60.0000 -    70.0000 [     0]:
     70.0000 -    80.0000 [     0]:
     80.0000 -    90.0000 [     0]:
     90.0000 -   100.0000 [     0]:

  Sender %c0
      0.0000 -    10.0000 [     1]: *
     10.0000 -    20.0000 [     0]:
     20.0000 -    30.0000 [     0]:
     30.0000 -    40.0000 [     2]: **
     40.0000 -    50.0000 [    58]:
  **********************************************************
     50.0000 -    60.0000 [     0]:
     60.0000 -    70.0000 [     0]:
     70.0000 -    80.0000 [     0]:
     80.0000 -    90.0000 [     0]:
     90.0000 -   100.0000 [     0]:

  -----------------------------------------------------------
  3.10-rc2 test RX (reverts 69a37beabf1f0a6705c08e879bdd5d82ff6486c4
  and e11538d1f03914eb92af5a1a378375c05ae8520c)
  TCP_RR trans/s 47766.95

  Receiver %c0
      0.0000 -    10.0000 [     1]: *
     10.0000 -    20.0000 [     1]: *
     20.0000 -    30.0000 [     0]:
     30.0000 -    40.0000 [    27]: ***************************
     40.0000 -    50.0000 [     2]: **
     50.0000 -    60.0000 [     0]:
     60.0000 -    70.0000 [     2]: **
     70.0000 -    80.0000 [     0]:
     80.0000 -    90.0000 [     0]:
     90.0000 -   100.0000 [    28]: ****************************

  Sender:
      0.0000 -    10.0000 [     1]: *
     10.0000 -    20.0000 [     0]:
     20.0000 -    30.0000 [     0]:
     30.0000 -    40.0000 [    11]: ***********
     40.0000 -    50.0000 [     0]:
     50.0000 -    60.0000 [     1]: *
     60.0000 -    70.0000 [     0]:
     70.0000 -    80.0000 [     3]: ***
     80.0000 -    90.0000 [     7]: *******
     90.0000 -   100.0000 [    38]: **************************************

  These results demonstrate gaining back the tendency of the CPU to
  stay in more responsive, performant C-states (and thus yield
  measurably better performance), by reverting commit
  69a37beabf1f0a6705c08e879bdd5d82ff6486c4.

Requested-by: Jeremy Eder <jeder@redhat.com>
Tested-by: Len Brown <len.brown@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agocpufreq: Fix cpufreq driver module refcount balance after suspend/resume
Rafael J. Wysocki [Mon, 29 Jul 2013 22:32:00 +0000 (00:32 +0200)]
cpufreq: Fix cpufreq driver module refcount balance after suspend/resume

commit 2a99859932281ed6c2ecdd988855f8f6838f6743 upstream.

Since cpufreq_cpu_put() called by __cpufreq_remove_dev() drops the
driver module refcount, __cpufreq_remove_dev() causes that refcount
to become negative for the cpufreq driver after a suspend/resume
cycle.

This is not the only bad thing that happens there, however, because
kobject_put() should only be called for the policy kobject at this
point if the CPU is not the last one for that policy.

Namely, if the given CPU is the last one for that policy, the
policy kobject's refcount should be 1 at this point, as set by
cpufreq_add_dev_interface(), and only needs to be dropped once for
the kobject to go away.  This actually happens under the cpu == 1
check, so it need not be done before by cpufreq_cpu_put().

On the other hand, if the given CPU is not the last one for that
policy, this means that cpufreq_add_policy_cpu() has been called
at least once for that policy and cpufreq_cpu_get() has been
called for it too.  To balance that cpufreq_cpu_get(), we need to
call cpufreq_cpu_put() in that case.

Thus, to fix the described problem and keep the reference
counters balanced in both cases, move the cpufreq_cpu_get() call
in __cpufreq_remove_dev() to the code path executed only for
CPUs that share the policy with other CPUs.

Reported-and-tested-by: Toralf Förster <toralf.foerster@gmx.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoRevert "cpuidle: Quickly notice prediction failure in general case"
Rafael J. Wysocki [Fri, 26 Jul 2013 23:13:26 +0000 (01:13 +0200)]
Revert "cpuidle: Quickly notice prediction failure in general case"

commit 228b30234f258a193317874854eee1ca7807186e upstream.

Revert commit e11538d1 (cpuidle: Quickly notice prediction failure in
general case), since it depends on commit 69a37be (cpuidle: Quickly
notice prediction failure for repeat mode) that has been identified
as the source of a significant performance regression in v3.8 and
later.

Requested-by: Jeremy Eder <jeder@redhat.com>
Tested-by: Len Brown <len.brown@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoACPI / battery: Fix parsing _BIX return value
Lan Tianyu [Tue, 30 Jul 2013 12:00:42 +0000 (14:00 +0200)]
ACPI / battery: Fix parsing _BIX return value

commit 016d5baad04269e8559332df05f89bd95b52d6ad upstream.

The _BIX method returns extended battery info as a package.
According the ACPI spec (ACPI 5, Section 10.2.2.2), the first member
of that package should be "Revision".  However, the current ACPI
battery driver treats the first member as "Power Unit" which should
be the second member.  This causes the result of _BIX return data
parsing to be incorrect.

Fix this by adding a new member called 'revision' to struct
acpi_battery and adding the offsetof() information on it to
extended_info_offsets[] as the first row.

[rjw: Changelog]
Reported-and-tested-by: Jan Hoffmann <jan.christian.hoffmann@gmail.com>
References: http://bugzilla.kernel.org/show_bug.cgi?id=60519
Signed-off-by: Lan Tianyu <tianyu.lan@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agozram: protect sysfs handler from invalid memory access
Jiang Liu [Thu, 6 Jun 2013 16:07:27 +0000 (00:07 +0800)]
zram: protect sysfs handler from invalid memory access

commit 5863e10b441e7ea4b492f930f1be180a97d026f3 upstream.

Use zram->init_lock to protect access to zram->meta, otherwise it
may cause invalid memory access if zram->meta has been freed by
zram_reset_device().

This issue may be triggered by:
Thread 1:
while true; do cat mem_used_total; done
Thread 2:
while true; do echo 8M > disksize; echo 1 > reset; done

Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agozram: avoid access beyond the zram device
Jiang Liu [Thu, 6 Jun 2013 16:07:26 +0000 (00:07 +0800)]
zram: avoid access beyond the zram device

commit 12a7ad3b810e77137d0caf97a6dd97591e075b30 upstream.

Function valid_io_request() should verify the entire request are within
the zram device address range. Otherwise it may cause invalid memory
access when accessing/modifying zram->meta->table[index] because the
'index' is out of range. Then it may access non-exist memory, randomly
modify memory belong to other subsystems, which is hard to track down.

Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agozram: avoid double free in function zram_bvec_write()
Jiang Liu [Thu, 6 Jun 2013 16:07:25 +0000 (00:07 +0800)]
zram: avoid double free in function zram_bvec_write()

commit 65c484609a3b25c35e4edcd5f2c38f98f5226093 upstream.

When doing a patial write and the whole page is filled with zero,
zram_bvec_write() will free uncmem twice.

Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agozram: destroy all devices on error recovery path in zram_init()
Jiang Liu [Thu, 6 Jun 2013 16:07:24 +0000 (00:07 +0800)]
zram: destroy all devices on error recovery path in zram_init()

commit 39a9b8ac9333e4268ecff7da6c9d1ab3823ff243 upstream.

On error recovery path of zram_init(), it leaks the zram device object
causing the failure. So change create_device() to free allocated
resources on error path.

Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Acked-by: Jerome Marchand <jmarchan@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agozram: use zram->lock to protect zram_free_page() in swap free notify path
Jiang Liu [Thu, 6 Jun 2013 16:07:23 +0000 (00:07 +0800)]
zram: use zram->lock to protect zram_free_page() in swap free notify path

commit 57ab048532c0d975538cebd4456491b5c34248f4 upstream.

zram_slot_free_notify() is free-running without any protection from
concurrent operations. So there are race conditions between
zram_bvec_read()/zram_bvec_write() and zram_slot_free_notify(),
and possible consequences include:
1) Trigger BUG_ON(!handle) on zram_bvec_write() side.
2) Access to freed pages on zram_bvec_read() side.
3) Break some fields (bad_compress, good_compress, pages_stored)
   in zram->stats if the swap layer makes concurrently call to
   zram_slot_free_notify().

So enhance zram_slot_free_notify() to acquire writer lock on zram->lock
before calling zram_free_page().

Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agozram: avoid invalid memory access in zram_exit()
Jiang Liu [Thu, 6 Jun 2013 16:07:22 +0000 (00:07 +0800)]
zram: avoid invalid memory access in zram_exit()

commit 6030ea9b35971a4200062f010341ab832e878ac9 upstream.

Memory for zram->disk object may have already been freed after returning
from destroy_device(zram), then it's unsafe for zram_reset_device(zram)
to access zram->disk again.

We can't solve this bug by flipping the order of destroy_device(zram)
and zram_reset_device(zram), that will cause deadlock issues to the
zram sysfs handler.

So fix it by holding an extra reference to zram->disk before calling
destroy_device(zram).

Signed-off-by: Jiang Liu <jiang.liu@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomwifiex: fix wrong data rates in P2P client
Avinash Patil [Mon, 29 Jul 2013 23:32:38 +0000 (16:32 -0700)]
mwifiex: fix wrong data rates in P2P client

commit 237b2ac8ac89a6b0120decdd05c7bf4637deb98a upstream.

This patch fixes an issue wherein adhoc rates were being copied
into association request from P2P client.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomwifiex: check for bss_role instead of bss_mode for STA operations
Avinash Patil [Mon, 29 Jul 2013 23:32:37 +0000 (16:32 -0700)]
mwifiex: check for bss_role instead of bss_mode for STA operations

commit 953b3539ef9301b8ef73f4b6e2fd824b86aae65a upstream.

This patch fixes an issue wherein association would fail on P2P
interfaces. This happened because we are checking priv->mode
against NL80211_IFTYPE_STATION. While this check is correct for
infrastructure stations, it would fail P2P clients for which mode
is NL80211_IFTYPE_P2P_CLIENT.

Better check would be bss_role which has only 2 values: STA/AP.

Signed-off-by: Avinash Patil <patila@marvell.com>
Signed-off-by: Stone Piao <piaoyun@marvell.com>
Signed-off-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomwifiex: Add missing endian conversion.
Tomasz Moń [Tue, 23 Jul 2013 05:42:49 +0000 (07:42 +0200)]
mwifiex: Add missing endian conversion.

commit 83e612f632c3897be29ef02e0472f6d63e258378 upstream.

Both type and pkt_len variables are in host endian and these should be in
Little Endian in the payload.

Signed-off-by: Tomasz Moń <desowin@gmail.com>
Acked-by: Bing Zhao <bzhao@marvell.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agort2x00: fix stop queue
Stanislaw Gruszka [Sun, 28 Jul 2013 11:17:22 +0000 (13:17 +0200)]
rt2x00: fix stop queue

commit e2288b66fe7ff0288382b2af671b4da558b44472 upstream.

Since we clear QUEUE_STARTED in rt2x00queue_stop_queue(), following
call to rt2x00queue_pause_queue() reduce to noop, i.e we do not
stop queue in mac80211.

To fix that introduce rt2x00queue_pause_queue_nocheck() function,
which will stop queue in mac80211 directly.

Note that rt2x00_start_queue() explicitly set QUEUE_PAUSED bit.

Note also that reordering operations i.e. first call to
rt2x00queue_pause_queue() and then clear QUEUE_STARTED bit, will race
with rt2x00queue_unpause_queue(), so calling ieee80211_stop_queue()
directly is the only available solution to fix the problem without
major rework.

Signed-off-by: Stanislaw Gruszka <stf_xl@wp.pl>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agosvcrpc: fix kfree oops in gss-proxy code
J. Bruce Fields [Wed, 31 Jul 2013 18:11:14 +0000 (14:11 -0400)]
svcrpc: fix kfree oops in gss-proxy code

commit 743e217129f69aab074abe520a464fd0c6b1cca1 upstream.

mech_oid.data is an array, not kmalloc()'d memory.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agosvcrpc: fix gss_rpc_upcall create error
J. Bruce Fields [Mon, 10 Jun 2013 20:06:44 +0000 (16:06 -0400)]
svcrpc: fix gss_rpc_upcall create error

commit 9f96392b0ae6aefc02a9b900c3f4889dfafc8402 upstream.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agosvcrpc: fix gss-proxy xdr decoding oops
J. Bruce Fields [Fri, 7 Jun 2013 14:11:19 +0000 (10:11 -0400)]
svcrpc: fix gss-proxy xdr decoding oops

commit dc43376c26cef74226174a2394f37f2a3f8a8639 upstream.

Uninitialized stack data was being used as the destination for memcpy's.

Longer term we'll just delete some of this code; all we're doing is
skipping over xdr that we don't care about.

Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoBluetooth: fix wrong use of PTR_ERR() in btusb
Adam Lee [Wed, 10 Jul 2013 02:02:12 +0000 (10:02 +0800)]
Bluetooth: fix wrong use of PTR_ERR() in btusb

commit d9c78e9738ccd0017b10b8f44462aafb61904a4a upstream.

PTR_ERR() returns a signed long type value which is limited by IS_ERR(),
it must be a negative number whose range is [-MAX_ERRNO, 0).

The bug here returns negative numbers as error codes, then check it by
"if (ret < 0)", but -PTR_ERR() is actually positive. The wrong use here
leads to failure as below, even panic.

[   12.958920] Bluetooth: hci0 command 0xfc8e tx timeout
[   14.961765] Bluetooth: hci0 command 0xfc8e tx timeout
[   16.964688] Bluetooth: hci0 command 0xfc8e tx timeout
[   20.954501] Bluetooth: hci0 sending Intel patch command (0xfc8e) failed (-110)
[   22.957358] Bluetooth: hci0 command 0xfc8e tx timeout
[   30.948922] Bluetooth: hci0 sending Intel patch command (0xfc8e) failed (-110)
[   32.951780] Bluetooth: hci0 command 0xfc8e tx timeout
[   40.943359] Bluetooth: hci0 sending Intel patch command (0xfc8e) failed (-110)
[   42.946219] Bluetooth: hci0 command 0xfc8e tx timeout
[   50.937812] Bluetooth: hci0 sending Intel patch command (0xfc8e) failed (-110)
[   52.940670] Bluetooth: hci0 command 0xfc8e tx timeout
[   60.932236] Bluetooth: hci0 sending Intel patch command (0xfc8e) failed (-110)
[   62.935092] Bluetooth: hci0 command 0xfc8e tx timeout
[   70.926688] Bluetooth: hci0 sending Intel patch command (0xfc8e) failed (-110)
[   72.929545] Bluetooth: hci0 command 0xfc8e tx timeout
[   80.921111] Bluetooth: hci0 sending Intel patch command (0xfc8e) failed (-110)
[   82.923969] Bluetooth: hci0 command 0xfc2f tx timeout
[   90.915542] Bluetooth: hci0 sending Intel patch command (0xfc2f) failed (-110)
[   92.918406] Bluetooth: hci0 command 0xfc11 tx timeout
[  100.909955] Bluetooth: hci0 sending Intel patch command (0xfc11) failed (-110)
[  102.912858] Bluetooth: hci0 command 0xfc60 tx timeout
[  110.904394] Bluetooth: hci0 sending Intel patch command (0xfc60) failed (-110)
[  112.907293] Bluetooth: hci0 command 0xfc11 tx timeout
[  120.898831] Bluetooth: hci0 exiting Intel manufacturer mode failed (-110)
[  120.904757] bluetoothd[1030]: segfault at 4 ip 00007f8b2eb55236 sp 00007fff53ff6920 error 4 in bluetoothd[7f8b2eaff000+cb000]

Signed-off-by: Adam Lee <adam.lee@canonical.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoBluetooth: Add support for Mediatek Bluetooth device [0e8d:763f]
Cho, Yu-Chen [Tue, 4 Jun 2013 13:40:26 +0000 (21:40 +0800)]
Bluetooth: Add support for Mediatek Bluetooth device [0e8d:763f]

commit 178c059e7640aa8e50213400c6f3dde00189d979 upstream.

This patch adds support for Mediatek Bluetooth device

T:  Bus=02 Lev=01 Prnt=01 Port=03 Cnt=01 Dev#=  2 Spd=480  MxCh= 0
D:  Ver= 2.01 Cls=ef(misc ) Sub=02 Prot=01 MxPS=64 #Cfgs=  1
P:  Vendor=0e8d ProdID=763f Rev= 1.00
S:  Manufacturer=MediaTek
S:  Product=BT
S:  SerialNumber=1.0
C:* #Ifs= 2 Cfg#= 1 Atr=a0 MxPwr=450mA
A:  FirstIf#= 0 IfCount= 2 Cls=ff(vend.) Sub=ff Prot=ff
I:* If#= 0 Alt= 0 #EPs= 3 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
E:  Ad=81(I) Atr=03(Int.) MxPS=  16 Ivl=125us
E:  Ad=02(O) Atr=02(Bulk) MxPS= 512 Ivl=125us
E:  Ad=82(I) Atr=02(Bulk) MxPS= 512 Ivl=0ms
I:* If#= 1 Alt= 0 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
E:  Ad=03(O) Atr=01(Isoc) MxPS=   0 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=   0 Ivl=1ms
I:  If#= 1 Alt= 1 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
E:  Ad=03(O) Atr=01(Isoc) MxPS=   9 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=   9 Ivl=1ms
I:  If#= 1 Alt= 2 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
E:  Ad=03(O) Atr=01(Isoc) MxPS=  17 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  17 Ivl=1ms
I:  If#= 1 Alt= 3 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
E:  Ad=03(O) Atr=01(Isoc) MxPS=  25 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  25 Ivl=1ms
I:  If#= 1 Alt= 4 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
E:  Ad=03(O) Atr=01(Isoc) MxPS=  33 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  33 Ivl=1ms
I:  If#= 1 Alt= 5 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
E:  Ad=03(O) Atr=01(Isoc) MxPS=  49 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  49 Ivl=1ms
I:  If#= 1 Alt= 6 #EPs= 2 Cls=ff(vend.) Sub=ff Prot=ff Driver=(none)
E:  Ad=03(O) Atr=01(Isoc) MxPS=  63 Ivl=1ms
E:  Ad=83(I) Atr=01(Isoc) MxPS=  63 Ivl=1ms

Signed-off-by: Cho, Yu-Chen <acho@suse.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoBluetooth: Add support for Atheros [0cf3:e003]
AceLan Kao [Thu, 20 Jun 2013 05:38:45 +0000 (13:38 +0800)]
Bluetooth: Add support for Atheros [0cf3:e003]

commit 1d5b569ef85d013a775560a90050dc630614c045 upstream.

Add support for the AR9462 chip

T:  Bus=02 Lev=02 Prnt=02 Port=04 Cnt=01 Dev#=  4 Spd=12  MxCh= 0
D:  Ver= 1.10 Cls=e0(wlcon) Sub=01 Prot=01 MxPS=64 #Cfgs=  1
P:  Vendor=0cf3 ProdID=e003 Rev=00.02
C:  #Ifs= 2 Cfg#= 1 Atr=e0 MxPwr=100mA
I:  If#= 0 Alt= 0 #EPs= 3 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
I:  If#= 1 Alt= 0 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb

Signed-off-by: AceLan Kao <acelan.kao@canonical.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoBluetooth: Add support for Atheros [0cf3:3121]
AceLan Kao [Wed, 17 Jul 2013 03:27:40 +0000 (11:27 +0800)]
Bluetooth: Add support for Atheros [0cf3:3121]

commit 1ebd0b21ab14efb75950079840eac29afea2a26e upstream.

Add support for the AR3012 chip.

T:  Bus=03 Lev=01 Prnt=01 Port=06 Cnt=01 Dev#=  6 Spd=12  MxCh= 0
D:  Ver= 1.10 Cls=e0(wlcon) Sub=01 Prot=01 MxPS=64 #Cfgs=  1
P:  Vendor=0cf3 ProdID=3121 Rev=00.02
C:  #Ifs= 2 Cfg#= 1 Atr=e0 MxPwr=100mA
I:  If#= 0 Alt= 0 #EPs= 3 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
I:  If#= 1 Alt= 0 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb

Signed-off-by: AceLan Kao <acelan.kao@canonical.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoBluetooth: ath3k: Add support for ID 0x13d3/0x3402
Sujith Manoharan [Mon, 15 Jul 2013 03:59:03 +0000 (09:29 +0530)]
Bluetooth: ath3k: Add support for ID 0x13d3/0x3402

commit 5b77a1f3d7b7360dc2b7c6d2188d39b9f8432907 upstream.

T:  Bus=01 Lev=02 Prnt=02 Port=00 Cnt=01 Dev#=  5 Spd=12   MxCh= 0
D:  Ver= 1.10 Cls=e0(wlcon) Sub=01 Prot=01 MxPS=64 #Cfgs=  1
P:  Vendor=13d3 ProdID=3402 Rev= 0.02
S:  Manufacturer=Atheros Communications
S:  Product=Bluetooth USB Host Controller
S:  SerialNumber=Alaska Day 2006
C:* #Ifs= 2 Cfg#= 1 Atr=e0 MxPwr=100mA
I:* If#= 0 Alt= 0 #EPs= 3 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb

Bug: https://bugzilla.kernel.org/show_bug.cgi?id=59701

Signed-off-by: Sujith Manoharan <sujith@msujith.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoBluetooth: ath3k: don't use stack memory for DMA
Stanislaw Gruszka [Mon, 8 Jul 2013 08:27:23 +0000 (10:27 +0200)]
Bluetooth: ath3k: don't use stack memory for DMA

commit 517828a87994f41af6ae5a0f96f0f069f05baa81 upstream.

Memory allocated by vmalloc (including stack) can not be used for DMA,
i.e. data pointer on usb_control_msg() should not point to stack memory.

Resolves:
https://bugzilla.redhat.com/show_bug.cgi?id=977558

Reported-and-tested-by: Andy Lawrence <dr.diesel@gmail.com>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoBluetooth: ath3k: Add support for Fujitsu Lifebook UH5x2 [04c5:1330]
Thomas Loo [Wed, 3 Jul 2013 00:53:54 +0000 (02:53 +0200)]
Bluetooth: ath3k: Add support for Fujitsu Lifebook UH5x2 [04c5:1330]

commit 84eb2ae1807dd1467bf6f500fc69ae61f1907b75 upstream.

The Fujitsu Lifebook UH552/UH572 ships with a Qualcomm AR9462/AR3012
WLAN/BT-Combo card.
Add device ID to the ath3k driver to enable the bluetooth side of things.
Patch against v3.10.

T:  Bus=03 Lev=01 Prnt=01 Port=02 Cnt=01 Dev#=  3 Spd=12  MxCh= 0
D:  Ver= 1.10 Cls=e0(wlcon) Sub=01 Prot=01 MxPS=64 #Cfgs=  1
P:  Vendor=04c5 ProdID=1330 Rev=00.02
C:  #Ifs= 2 Cfg#= 1 Atr=e0 MxPwr=100mA
I:  If#= 0 Alt= 0 #EPs= 3 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb
I:  If#= 1 Alt= 0 #EPs= 2 Cls=e0(wlcon) Sub=01 Prot=01 Driver=btusb

Signed-off-by: Thomas Loo <tloo@saltstorm.net>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoath: wil6210: Fix build error
Larry Finger [Sun, 21 Jul 2013 02:46:48 +0000 (21:46 -0500)]
ath: wil6210: Fix build error

commit 5d21608a592a9afcac8d82c6478a564e911ce70b upstream.

Building driver wil6210 in 3.10 and 3.11 kernels yields the following errors:

  CC [M]  drivers/net/wireless/ath/wil6210/debugfs.o
drivers/net/wireless/ath/wil6210/debugfs.c: In function 'wil_print_ring':
drivers/net/wireless/ath/wil6210/debugfs.c:163:11: error: pointer targets in passing argument 5 of 'hex_dump_to_buffer' differ in signedness [-Werror=pointer-sign]
           false);
           ^
In file included from include/linux/kernel.h:13:0,
                 from include/linux/cache.h:4,
                 from include/linux/time.h:4,
                 from include/linux/stat.h:18,
                 from include/linux/module.h:10,
                 from drivers/net/wireless/ath/wil6210/debugfs.c:17:
include/linux/printk.h:361:13: note: expected 'char *' but argument is of type 'unsigned char *'
 extern void hex_dump_to_buffer(const void *buf, size_t len,
             ^
drivers/net/wireless/ath/wil6210/debugfs.c: In function 'wil_txdesc_debugfs_show':
drivers/net/wireless/ath/wil6210/debugfs.c:429:10: error: pointer targets in passing argument 5 of 'hex_dump_to_buffer' differ in signedness [-Werror=pointer-sign]
          sizeof(printbuf), false);
          ^
In file included from include/linux/kernel.h:13:0,
                 from include/linux/cache.h:4,
                 from include/linux/time.h:4,
                 from include/linux/stat.h:18,
                 from include/linux/module.h:10,
                 from drivers/net/wireless/ath/wil6210/debugfs.c:17:
include/linux/printk.h:361:13: note: expected 'char *' but argument is of type 'unsigned char *'
 extern void hex_dump_to_buffer(const void *buf, size_t len,
             ^
cc1: all warnings being treated as errors
make[5]: *** [drivers/net/wireless/ath/wil6210/debugfs.o] Error 1
make[4]: *** [drivers/net/wireless/ath/wil6210] Error 2
make[3]: *** [drivers/net/wireless/ath] Error 2
make[2]: *** [drivers/net/wireless] Error 2
make[1]: *** [drivers/net] Error 2
make: *** [drivers] Error 2

These errors are fixed by changing the type of the buffer from "unsigned char *" to "char *".

Reported-by: Thomas Fjellstrom <thomas@fjellstrom.ca>
Tested-by: Thomas Fjellstrom <thomas@fjellstrom.ca>
Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Cc: Thomas Fjellstrom <thomas@fjellstrom.ca>
Signed-off-by: Vladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoixgbe: Fix Tx Hang issue with lldpad on 82598EB
Jacob Keller [Fri, 26 Jul 2013 12:46:35 +0000 (05:46 -0700)]
ixgbe: Fix Tx Hang issue with lldpad on 82598EB

commit 1eb9ac14c34a948bf1538bfb9034e8ab29099a64 upstream.

This patch fixes an issue with the 82598EB device, where lldpad is causing Tx
Hangs on the card as soon as it attempts to configure DCB for the device. The
adapter will continually Tx hang and reset in a loop.

Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Tested-by: Jack Morgan <jack.morgan@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomac80211: fix monitor interface suspend crash regression
Stanislaw Gruszka [Tue, 23 Jul 2013 11:56:50 +0000 (13:56 +0200)]
mac80211: fix monitor interface suspend crash regression

commit cd34f647a78e7f2296fcb72392b9e5c832793e65 upstream.

My commit:

commit 12e7f517029dad819c45eca9ca01fdb9ba57616b
Author: Stanislaw Gruszka <sgruszka@redhat.com>
Date:   Thu Feb 28 10:55:26 2013 +0100

    mac80211: cleanup generic suspend/resume procedures

removed check for deleting MONITOR and AP_VLAN when suspend. That can
cause a crash (i.e. in iwlagn_mac_remove_interface()) since we remove
interface in the driver that we did not add before.

Reference:
http://marc.info/?l=linux-kernel&m=137391815113860&w=2

Bisected-by: Ortwin Glück <odi@odi.ch>
Reported-and-tested-by: Ortwin Glück <odi@odi.ch>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomac80211: fix ethtool stats for non-station interfaces
Johannes Berg [Mon, 8 Jul 2013 08:43:31 +0000 (10:43 +0200)]
mac80211: fix ethtool stats for non-station interfaces

commit e13bae4f807401729b3f27c7e882a96b8b292809 upstream.

As reported in https://bugzilla.kernel.org/show_bug.cgi?id=60514,
the station loop never initialises 'sinfo' and therefore adds up
a stack values, leaking stack information (the number of times it
adds values is easily obtained another way.)

Fix this by initialising the sinfo for each station to add.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomac80211: fix duplicate retransmission detection
Johannes Berg [Thu, 11 Jul 2013 20:33:26 +0000 (22:33 +0200)]
mac80211: fix duplicate retransmission detection

commit 6b0f32745dcfba01d7be33acd1b40306c7a914c6 upstream.

The duplicate retransmission detection code in mac80211
erroneously attempts to do the check for every frame,
even frames that don't have a sequence control field or
that don't use it (QoS-Null frames.)

This is problematic because it causes the code to access
data beyond the end of the SKB and depending on the data
there will drop packets erroneously.

Correct the code to not do duplicate detection for such
frames.

I found this error while testing AP powersave, it lead
to retransmitted PS-Poll frames being dropped entirely
as the data beyond the end of the SKB was always zero.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomac80211/minstrel_ht: fix cck rate sampling
Felix Fietkau [Fri, 28 Jun 2013 19:04:35 +0000 (21:04 +0200)]
mac80211/minstrel_ht: fix cck rate sampling

commit 1cd158573951f737fbc878a35cb5eb47bf9af3d5 upstream.

The CCK group needs special treatment to set the right flags and rate
index. Add this missing check to prevent setting broken rates for tx
packets.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomac80211/minstrel: fix NULL pointer dereference issue
Felix Fietkau [Mon, 15 Jul 2013 12:35:06 +0000 (14:35 +0200)]
mac80211/minstrel: fix NULL pointer dereference issue

commit 5c9fc93bc9bc417418fc1b6366833ae6a07b804d upstream.

When priv_sta == NULL, mi->prev_sample is dereferenced too early. Move
the assignment further down, after the rate_control_send_low call.

Reported-by: Krzysztof Mazur <krzysiek@podlesie.net>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agonl80211: fix mgmt tx status and testmode reporting for netns
Michal Kazior [Tue, 25 Jun 2013 07:17:17 +0000 (09:17 +0200)]
nl80211: fix mgmt tx status and testmode reporting for netns

commit a0ec570f4f69c4cb700d743a915096c2c8f56a99 upstream.

These two events were sent to the default network
namespace.

This caused AP mode in a non-default netns to not
work correctly. Mgmt tx status was multicasted to
a different (default) netns instead of the one the
AP was in.

Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoath9k_htc: reboot firmware if it was loaded
Oleksij Rempel [Fri, 19 Jul 2013 18:16:18 +0000 (20:16 +0200)]
ath9k_htc: reboot firmware if it was loaded

commit 4928bd2ef8ece262f4f314630219999a91eaa440 upstream.

Currently ath9k_htc will reboot firmware only if interface was
ever started. Which lead to the problem in case where interface
was never started but module need to be reloaded.

This patch will partially fix bug "ath9k_htc: Target is unresponsive"
https://github.com/qca/open-ath9k-htc-firmware/issues/1

Reproduction case:
- plug adapter
- make sure nothing will touch it. Stop Networkmanager or blacklist mac address of this adapter.
- rmmod ath9k_htc; sleep 1; modprobe ath9k_htc

Signed-off-by: Oleksij Rempel <linux@rempel-privat.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoath9k_htc: do some initial hardware configuration
Oleksij Rempel [Fri, 19 Jul 2013 18:16:17 +0000 (20:16 +0200)]
ath9k_htc: do some initial hardware configuration

commit dc2a87f519a4d8cb376ab54f22b6b98a943b51ce upstream.

Currently we configure harwdare and clock, only after
interface start. In this case, if we reload module or
reboot PC without configuring adapter, firmware will freeze.
There is no software way to reset adpter.

This patch add initial configuration and set it in
disabled state, to avoid this freeze. Behaviour of this patch
should be similar to: ifconfig wlan0 up; ifconfig wlan0 down.

Bug: https://github.com/qca/open-ath9k-htc-firmware/issues/1
Tested-by: Bo Shi <cnshibo@gmail.com>
Signed-off-by: Oleksij Rempel <linux@rempel-privat.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoiwlwifi: mvm: fix flushing not started aggregation sessions
Johannes Berg [Wed, 24 Jul 2013 11:55:51 +0000 (13:55 +0200)]
iwlwifi: mvm: fix flushing not started aggregation sessions

commit b6658ff80c43bcf84be0bbe371c88af1452e7776 upstream.

When a not fully started aggregation session is destroyed
and flushed, we get a warning, e.g.

  WARNING: at drivers/net/wireless/iwlwifi/pcie/tx.c:1142 iwl_trans_pcie_txq_disable+0x11c/0x160
  queue 16 not used
  Modules linked in: [...]
  Pid: 5135, comm: hostapd Tainted: G        W  O 3.5.0 #10
  Call Trace:
  wlan0: driver sets block=0 for sta 00:03:7f:10:44:d3
   [<ffffffff81036492>] warn_slowpath_common+0x72/0xa0
   [<ffffffff81036577>] warn_slowpath_fmt+0x47/0x50
   [<ffffffffa0368d6c>] iwl_trans_pcie_txq_disable+0x11c/0x160 [iwlwifi]
   [<ffffffffa03a2099>] iwl_mvm_sta_tx_agg_flush+0xe9/0x150 [iwlmvm]
   [<ffffffffa0396c43>] iwl_mvm_mac_ampdu_action+0xf3/0x1e0 [iwlmvm]
   [<ffffffffa0293ad3>] ___ieee80211_stop_tx_ba_session+0x193/0x920 [mac80211]
   [<ffffffffa0294ed8>] __ieee80211_stop_tx_ba_session+0x48/0x70 [mac80211]
   [<ffffffffa029159f>] ieee80211_sta_tear_down_BA_sessions+0x4f/0x80 [mac80211]
   [<ffffffffa028a686>] __sta_info_destroy+0x66/0x370 [mac80211]
   [<ffffffffa028abb4>] sta_info_destroy_addr_bss+0x44/0x70 [mac80211]
   [<ffffffffa02a3e26>] ieee80211_del_station+0x26/0x50 [mac80211]
   [<ffffffffa01e6395>] nl80211_del_station+0x85/0x200 [cfg80211]

when a station deauthenticated from us without fully setting
up the aggregation session.

Fix this by checking the aggregation state before removing
the hardware queue.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoiwlwifi: add DELL SKU for 5150 HMC
Emmanuel Grumbach [Thu, 18 Jul 2013 16:11:26 +0000 (19:11 +0300)]
iwlwifi: add DELL SKU for 5150 HMC

commit a1923f1d4723e5757cefdd60f7c7ab30e472007a upstream.

This SKU was missing in the list of supported devices

https://bugzilla.kernel.org/show_bug.cgi?id=60577

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoiwlwifi: mvm: refuse connection to APs with BI < 16
Johannes Berg [Thu, 4 Jul 2013 13:55:29 +0000 (15:55 +0200)]
iwlwifi: mvm: refuse connection to APs with BI < 16

commit 48bc13072109ea58465542aa1ee31b4e1065468a upstream.

Due to a firmware bug, it crashes when the beacon interval
is smaller than 16. Avoid this by refusing the station state
change creating the AP station, causing mac80211 to abandon
the attempt to connect to the AP, and eventually wpa_s to
blacklist it.

Reviewed-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoiwlwifi: mvm: fix bug in scan ssid
David Spinadel [Thu, 4 Jul 2013 12:17:48 +0000 (15:17 +0300)]
iwlwifi: mvm: fix bug in scan ssid

commit fe04e83706037802c502ea41c0d1799ec35fc0d7 upstream.

Increment index in each iteration. Without this increment we are
overriding the added SSIDs and we will send only the last SSId
and (n_ssids - 1) broadcast probes.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoiwlwifi: mvm: fix L2P BA ressources leak
Emmanuel Grumbach [Tue, 2 Jul 2013 10:35:35 +0000 (13:35 +0300)]
iwlwifi: mvm: fix L2P BA ressources leak

commit 93a426673fbfeae7fa6b27008828e2ac4c08dbee upstream.

We didn't release the Rx AMPDU ressources properly.
This bug led to firmware assert after 16 BA sessions.

Signed-off-by: Emmanuel Grumbach <emmanuel.grumbach@intel.com>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: mos7840: fix pointer casts
Johan Hovold [Sat, 27 Jul 2013 11:34:42 +0000 (13:34 +0200)]
USB: mos7840: fix pointer casts

commit 683a0e4d7971c3186dc4d429027debfe309129aa upstream.

Silence compiler warnings on 64-bit systems introduced by commit
05cf0dec ("USB: mos7840: fix race in led handling") which uses the
usb-serial data pointer to temporarily store the device type during
probe but failed to add the required casts.

[gregkh - change uintptr_t to unsigned long]

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: mos7840: fix race in led handling
Johan Hovold [Fri, 26 Jul 2013 09:55:19 +0000 (11:55 +0200)]
USB: mos7840: fix race in led handling

commit 05cf0dec5ccc696a7636c84b265b477173498156 upstream.

Fix race in LED handling introduced by commit 0eafe4de ("USB: serial:
mos7840: add support for MCS7810 devices") which reused the port control
urb for manipulating the LED without making sure that the urb is not
already in use. This could lead to the control urb being manipulated
while in flight.

Fix by adding a dedicated LED urb and ctrlrequest along with a LED-busy
flag to handle concurrency.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: mos7840: fix device-type detection
Johan Hovold [Fri, 26 Jul 2013 09:55:18 +0000 (11:55 +0200)]
USB: mos7840: fix device-type detection

commit 40c24f2893ba0ba7df485871f6aac0c197ceef5b upstream.

Fix race in device-type detection introduced by commit 0eafe4de ("USB:
serial: mos7840: add support for MCS7810 devices") which used a static
variable to hold the device type.

Move type detection to probe and use serial data to store the device
type.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: mos7840: fix race in register handling
Johan Hovold [Fri, 26 Jul 2013 09:55:17 +0000 (11:55 +0200)]
USB: mos7840: fix race in register handling

commit d8a083cc746664916d9d36ed9e4d08a29525f245 upstream.

Fix race in mos7840_get_reg which unconditionally manipulated the
control urb (which may already be in use) by adding a control-urb busy
flag.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodma: pl330: Fix cyclic transfers
Lars-Peter Clausen [Tue, 23 Jul 2013 08:24:50 +0000 (10:24 +0200)]
dma: pl330: Fix cyclic transfers

commit fc51446021f42aca8906e701fc2292965aafcb15 upstream.

Allocate a descriptor for each period of a cyclic transfer, not just the first.
Also since the callback needs to be called for each finished period make sure to
initialize the callback and callback_param fields of each descriptor in a cyclic
transfer.

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoserial/mxs-auart: increase time to wait for transmitter to become idle
Uwe Kleine-König [Fri, 28 Jun 2013 09:49:41 +0000 (11:49 +0200)]
serial/mxs-auart: increase time to wait for transmitter to become idle

commit 079a036f4283e2b0e5c26080b8c5112bc0cc1831 upstream.

Without this patch the driver waits ~1 ms for the UART to become idle. At
115200n8 this time is (theoretically) enough to transfer 11.5 characters
(= 115200 bits/s / (10 Bits/char) * 1ms). As the mxs-auart has a fifo size
of 16 characters the clock is gated too early. The problem is worse for
lower baud rates.

This only happens to really shut down the transmitter in the middle of a
transfer if /dev/ttyAPPx isn't opened in userspace (e.g. by a getty) but
was at least once (because the bootloader doesn't disable the transmitter).

So increase the timeout to 20 ms which should be enough for 9600n8, too.
Moreover skip gating the clock if the timeout is elapsed.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoserial: arc_uart: Fix module alias
Axel Lin [Sun, 21 Jul 2013 02:14:15 +0000 (10:14 +0800)]
serial: arc_uart: Fix module alias

commit d5a12ea7a9e58d9e5c19d25cb668aadb396423ec upstream.

Platform drivers use "platform:" prefix in module alias.
Also use DRIVER_NAME in MODULE_ALIAS to make module autoloading work.

Signed-off-by: Axel Lin <axel.lin@ingics.com>
Acked-by: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoserial/mxs-auart: fix race condition in interrupt handler
Uwe Kleine-König [Thu, 4 Jul 2013 09:28:51 +0000 (11:28 +0200)]
serial/mxs-auart: fix race condition in interrupt handler

commit d970d7fe65adff5efe75b4a73c4ffc9be57089f7 upstream.

The handler needs to ack the pending events before actually handling them.
Otherwise a new event might come in after it it considered non-pending or
handled and is acked then without being handled. So this event is only
noticed when the next interrupt happens.

Without this patch an i.MX28 based machine running an rt-patched kernel
regularly hangs during boot.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoALSA: compress: fix the return value for SNDRV_COMPRESS_VERSION
Vinod Koul [Mon, 29 Jul 2013 09:40:22 +0000 (15:10 +0530)]
ALSA: compress: fix the return value for SNDRV_COMPRESS_VERSION

commit a8d30608eaed6cc759b8e2e8a8bbbb42591f797f upstream.

the return value of SNDRV_COMPRESS_VERSION always return default -ENOTTY as the
return value was never updated for this call
assign return value from put_user()

Reported-by: Haynes <hgeorge@codeaurora.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoALSA: hda - Fix missing fixup for Mac Mini with STAC9221
Takashi Iwai [Thu, 1 Aug 2013 06:38:27 +0000 (08:38 +0200)]
ALSA: hda - Fix missing fixup for Mac Mini with STAC9221

commit 697aebab78a88c6b164cfb74d19b86817d2ccd82 upstream.

A fixup for Apple Mac Mini was lost during the adaption to the generic
parser because the fallback for the generic ID 8384:7680 was dropped,
and it resulted in the silence output (and maybe other problems).

Unfortunately, just adding the missing subsystem ID wasn't enough, in
this case.  The subsystem ID of this machine is 0000:0100 (what Apple
thought...?), and since snd_hda_pick_fixup() doesn't take the vendor
id zero into account, the driver ignored this entry.  Now it's fixed
to regard the vendor id zero as a valid value.

Reported-and-tested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agohwmon: (max6697) fix MAX6581 ideality
Vivien Didelot [Tue, 30 Jul 2013 21:14:34 +0000 (17:14 -0400)]
hwmon: (max6697) fix MAX6581 ideality

commit 5c52add19733eb36d8619713312f5604efef3502 upstream.

Without this patch, the values for ideality (register 0x4b) and ideality
selection mask (register 0x4c) are inverted.

Signed-off-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoparisc: Fix interrupt routing for C8000 serial ports
Thomas Bogendoerfer [Tue, 30 Jul 2013 00:02:16 +0000 (02:02 +0200)]
parisc: Fix interrupt routing for C8000 serial ports

commit dd5e6d6a3db09b16b7c222943977865eead88cc3 upstream.

We can't use dev->mod_index for selecting the interrupt routing entry,
because it's not an index into interrupt routing table. It will be even
wrong on a machine with 2 CPUs (4 cores). But all needed information is
contained in the PAT entries for the serial ports. mod[0] contains the
iosapic address and mod_info has some indications for the interrupt
input (at least it looks like it). This patch implements the searching
for the right iosapic and uses this interrupt input information.

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoparisc: Fix cache routines to ignore vma's with an invalid pfn
John David Anglin [Tue, 23 Jul 2013 16:27:52 +0000 (12:27 -0400)]
parisc: Fix cache routines to ignore vma's with an invalid pfn

commit 50861f5a02dbf939c27d35a26c472885e2844188 upstream.

The parisc architecture does not have a pte special bit. As a result,
special mappings are handled with the VM_PFNMAP and VM_MIXEDMAP flags.
VM_MIXEDMAP mappings may or may not have a "struct page" backing. When
pfn_valid() is false, there is no "struct page" backing. Otherwise, they
are treated as normal pages.

The FireGL driver uses the VM_MIXEDMAP without a backing "struct page".
This treatment caused a panic due to a TLB data miss in
update_mmu_cache. This appeared to be in the code generated for
page_address(). We were in fact using a very circular bit of code to
determine the physical address of the PFN in various cache routines.
This wasn't valid when there was no "struct page" backing.  The needed
address can in fact be determined simply from the PFN itself without
using the "struct page".

The attached patch updates update_mmu_cache(), flush_cache_mm(),
flush_cache_range() and flush_cache_page() to check pfn_valid() and to
directly compute the PFN physical and virtual addresses.

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoparisc: agp/parisc-agp: allow binding of user memory to the AGP GART
Alex Ivanov [Wed, 10 Jul 2013 19:14:55 +0000 (21:14 +0200)]
parisc: agp/parisc-agp: allow binding of user memory to the AGP GART

commit 06f0cce43a32bd2357cea1d8733bba48693d556b upstream.

Allow binding of user memory to the AGP GART on systems with HP
Quicksilver AGP bus. This resolves 'bind memory failed' error seen in
dmesg:

 [29.365973] [TTM] AGP Bind memory failed.
 …
 [29.367030] [drm] Forcing AGP to PCI mode

The system doesn't more fail to bind the memory, and hence not falling
back to the PCI mode (if other failures aren't detected).

This is just a simple write down from the following patches:
agp/amd-k7: Allow binding user memory to the AGP GART
agp/hp-agp: Allow binding user memory to the AGP GART

Signed-off-by: Alex Ivanov <gnidorah@p0n4ik.tk>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agopowerpc: VPHN topology change updates all siblings
Robert Jennings [Thu, 25 Jul 2013 01:13:21 +0000 (20:13 -0500)]
powerpc: VPHN topology change updates all siblings

commit 3be7db6ab45b21345386d1a466da133b19cde5e4 upstream.

When an associativity level change is found for one thread, the
siblings threads need to be updated as well.  This is done today
for PRRN in stage_topology_update() but is missing for VPHN in
update_cpu_associativity_changes_mask().  This patch will correctly
update all thread siblings during a topology change.

Without this patch a topology update can result in a CPU in
init_sched_groups_power() getting stuck indefinitely in a loop.

This loop is built in build_sched_groups(). As a result of the thread
moving to a node separate from its siblings the struct sched_group will
have its next pointer set to point to itself rather than the sched_group
struct of the next thread.  This happens because we have a domain without
the SD_OVERLAP flag, which is correct, and a topology that doesn't conform
with reality (threads on the same core assigned to different numa nodes).
When this list is traversed by init_sched_groups_power() it will reach
the thread's sched_group structure and loop indefinitely; the cpu will
be stuck at this point.

The bug was exposed when VPHN was enabled in commit b7abef0 (v3.9).

Reported-by: Jan Stancek <jstancek@redhat.com>
Signed-off-by: Robert Jennings <rcj@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: 7791/1: a.out: remove partial a.out support
Will Deacon [Thu, 25 Jul 2013 10:44:48 +0000 (11:44 +0100)]
ARM: 7791/1: a.out: remove partial a.out support

commit acfdd4b1f7590d02e9bae3b73bdbbc4a31b05d38 upstream.

a.out support on ARM requires that argc, argv and envp are passed in
r0-r2 respectively, which requires hacking load_aout_binary to
prevent argc being clobbered by the return code. Whilst mainline kernels
do set the registers up in start_thread, the aout loader has never
carried the hack in mainline.

Initialising the registers in this way actually goes against the libc
expectations for ELF binaries, where argc, argv and envp are passed on
the stack, with r0 being used to hold a pointer to an exit function for
cleaning up after the dynamic linker if required. If the pointer is
NULL, then it is ignored. When execing an ELF binary, Linux currently
zeroes r0, then sets it to argc and then finally clobbers it with the
return value of the execve syscall, so we actually end up with:

r0 = 0
stack[0] = argc
r1 = stack[1] = argv
r2 = stack[2] = envp

libc treats r1 and r2 as undefined. The clobbering of r0 by sys_execve
works for user-spawned threads, but when executing an ELF binary from a
kernel thread (via call_usermodehelper), the execve is performed on the
ret_from_fork path, which restores r0 from the saved pt_regs, resulting
in argc being presented to the C library. This has horrible consequences
when the application exits, since we have an exit function registered
using argc, resulting in a jump to hyperspace.

This patch solves the problem by removing the partial a.out support from
arch/arm/ altogether.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Cc: Ashish Sangwan <ashishsangwan2@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: 7790/1: Fix deferred mm switch on VIVT processors
Catalin Marinas [Tue, 23 Jul 2013 15:15:36 +0000 (16:15 +0100)]
ARM: 7790/1: Fix deferred mm switch on VIVT processors

commit bdae73cd374e28db544fdd9b77de689a36e3c129 upstream.

As of commit b9d4d42ad9 (ARM: Remove __ARCH_WANT_INTERRUPTS_ON_CTXSW on
pre-ARMv6 CPUs), the mm switching on VIVT processors is done in the
finish_arch_post_lock_switch() function to avoid whole cache flushing
with interrupts disabled. The need for deferred mm switch is stored as a
thread flag (TIF_SWITCH_MM). However, with preemption enabled, we can
have another thread switch before finish_arch_post_lock_switch(). If the
new thread has the same mm as the previous 'next' thread, the scheduler
will not call switch_mm() and the TIF_SWITCH_MM flag won't be set for
the new thread.

This patch moves the switch pending flag to the mm_context_t structure
since this is specific to the mm rather than thread.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Reported-by: Marc Kleine-Budde <mkl@pengutronix.de>
Tested-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: 7784/1: mm: ensure SMP alternates assemble to exactly 4 bytes with Thumb-2
Will Deacon [Mon, 15 Jul 2013 13:26:19 +0000 (14:26 +0100)]
ARM: 7784/1: mm: ensure SMP alternates assemble to exactly 4 bytes with Thumb-2

commit bf3f0f332f76a85ff3a0b393aaded5a8533769c0 upstream.

Commit ae8a8b9553bd ("ARM: 7691/1: mm: kill unused TLB_CAN_READ_FROM_L1_CACHE
and use ALT_SMP instead") added early function returns for page table
cache flushing operations on ARMv7 SMP CPUs.

Unfortunately, when targetting Thumb-2, these `mov pc, lr' sequences
assemble to 2 bytes which can lead to corruption of the instruction
stream after code patching.

This patch fixes the alternates to use wide (32-bit) instructions for
Thumb-2, therefore ensuring that the patching code works correctly.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agopowerpc/windfarm: Fix noisy slots-fan on Xserve (rm31)
Aaro Koskinen [Sun, 21 Jul 2013 00:30:11 +0000 (03:30 +0300)]
powerpc/windfarm: Fix noisy slots-fan on Xserve (rm31)

commit fe956a1d4081ce1a959f87df397a15e252201f10 upstream.

slots-fan on G5 Xserve is always running at full speed with windfarm_rm31
driver, resulting in a very high acoustic noise level. It seems the fan
parameters are incorrect, and have been copied from the Drive Bay fan
(RPM, not present on rm31) of the legacy therm_pm72 driver. This patch
changes the parameters to match the Slots fan (PWM) of therm_pm72. With
the patch, slots-fan speed drops from 99% to 19% during normal use,
and slots-temp settle to ~42'C.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: fix nommu builds with 48be69a02 (ARM: move signal handlers into a vdso-like...
Russell King [Sat, 3 Aug 2013 09:39:51 +0000 (10:39 +0100)]
ARM: fix nommu builds with 48be69a02 (ARM: move signal handlers into a vdso-like page)

commit 8c0cc8a5d90bc7373a7a9e7f7a40eb41f51e03fc upstream.

Olof reports that noMMU builds error out with:

arch/arm/kernel/signal.c: In function 'setup_return':
arch/arm/kernel/signal.c:413:25: error: 'mm_context_t' has no member named 'sigpage'

This shows one of the evilnesses of IS_ENABLED().  Get rid of it here
and replace it with #ifdef's - and as no noMMU platform can make use
of sigpage, depend on CONIFG_MMU not CONFIG_ARM_MPU.

Reported-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: fix a cockup in 48be69a02 (ARM: move signal handlers into a vdso-like page)
Russell King [Sat, 3 Aug 2013 09:30:05 +0000 (10:30 +0100)]
ARM: fix a cockup in 48be69a02 (ARM: move signal handlers into a vdso-like page)

commit e0d407564b532d978b03ceccebd224a05d02f111 upstream.

Unfortunately, I never committed the fix to a nasty oops which can
occur as a result of that commit:

------------[ cut here ]------------
kernel BUG at /home/olof/work/batch/include/linux/mm.h:414!
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP ARM
Modules linked in:
CPU: 0 PID: 490 Comm: killall5 Not tainted 3.11.0-rc3-00288-gabe0308 #53
task: e90acac0 ti: e9be8000 task.ti: e9be8000
PC is at special_mapping_fault+0xa4/0xc4
LR is at __do_fault+0x68/0x48c

This doesn't show up unless you do quite a bit of testing; a simple
boot test does not do this, so all my nightly tests were passing fine.

The reason for this is that install_special_mapping() expects the
page array to stick around, and as this was only inserting one page
which was stored on the kernel stack, that's why this was blowing up.

Reported-by: Olof Johansson <olof@lixom.net>
Tested-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: make vectors page inaccessible from userspace
Russell King [Wed, 31 Jul 2013 20:58:56 +0000 (21:58 +0100)]
ARM: make vectors page inaccessible from userspace

commit a5463cd3435475386cbbe7b06e01292ac169d36f upstream.

If kuser helpers are not provided by the kernel, disable user access to
the vectors page.  With the kuser helpers gone, there is no reason for
this page to be visible to userspace.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: move signal handlers into a vdso-like page
Russell King [Tue, 23 Jul 2013 23:29:18 +0000 (00:29 +0100)]
ARM: move signal handlers into a vdso-like page

commit 48be69a026b2c17350a5ef18a1959a919f60be7d upstream.

Move the signal handlers into a VDSO page rather than keeping them in
the vectors page.  This allows us to place them randomly within this
page, and also map the page at a random location within userspace
further protecting these code fragments from ROP attacks.  The new
VDSO page is also poisoned in the same way as the vector page.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: allow kuser helpers to be removed from the vector page
Russell King [Tue, 23 Jul 2013 17:37:00 +0000 (18:37 +0100)]
ARM: allow kuser helpers to be removed from the vector page

commit f6f91b0d9fd971c630cef908dde8fe8795aefbf8 upstream.

Provide a kernel configuration option to allow the kernel user helpers
to be removed from the vector page, thereby preventing their use with
ROP (return orientated programming) attacks.  This option is only
visible for CPU architectures which natively support all the operations
which kernel user helpers would normally provide, and must be enabled
with caution.

Acked-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: update FIQ support for relocation of vectors
Russell King [Tue, 9 Jul 2013 00:03:17 +0000 (01:03 +0100)]
ARM: update FIQ support for relocation of vectors

commit e39e3f3ebfef03450cf7bfa7a974a8c61f7980c8 upstream.

FIQ should no longer copy the FIQ code into the user visible vector
page.  Instead, it should use the hidden page.  This change makes
that happen.

Acked-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: use linker magic for vectors and vector stubs
Russell King [Thu, 4 Jul 2013 11:03:31 +0000 (12:03 +0100)]
ARM: use linker magic for vectors and vector stubs

commit b9b32bf70f2fb710b07c94e13afbc729afe221da upstream.

Use linker magic to create the vectors and vector stubs: we can tell the
linker to place them at an appropriate VMA, but keep the LMA within the
kernel.  This gets rid of some unnecessary symbol manipulation, and
have the linker calculate the relocations appropriately.

Acked-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: move vector stubs
Russell King [Thu, 4 Jul 2013 10:40:32 +0000 (11:40 +0100)]
ARM: move vector stubs

commit 19accfd373847ac3d10623c5d20f948846299741 upstream.

Move the machine vector stubs into the page above the vector page,
which we can prevent from being visible to userspace.  Also move
the reset stub, and place the swi vector at a location that the
'ldr' can get to it.

This hides pointers into the kernel which could give valuable
information to attackers, and reduces the number of exploitable
instructions at a fixed address.

Acked-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: poison memory between kuser helpers
Russell King [Thu, 4 Jul 2013 10:32:04 +0000 (11:32 +0100)]
ARM: poison memory between kuser helpers

commit 5b43e7a383d69381ffe53423e46dd0fafae07da3 upstream.

Poison the memory between each kuser helper.  This ensures that any
branch between the kuser helpers will be appropriately trapped.

Acked-by: Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoARM: poison the vectors page
Russell King [Thu, 4 Jul 2013 10:00:23 +0000 (11:00 +0100)]
ARM: poison the vectors page

commit f928d4f2a86f46b030fa0850385b4391fc2b5918 upstream.

Fill the empty regions of the vectors page with an exception generating
instruction.  This ensures that any inappropriate branch to the vector
page is appropriately trapped, rather than just encountering some code
to execute.  (The vectors page was filled with zero before, which
corresponds with the "andeq r0, r0, r0" instruction - a no-op.)

Acked-by Nicolas Pitre <nico@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoLinux 3.10.5 v3.10.5
Greg Kroah-Hartman [Sun, 4 Aug 2013 08:51:49 +0000 (16:51 +0800)]
Linux 3.10.5

10 years agox86: Fix /proc/mtrr with base/size more than 44bits
Yinghai Lu [Thu, 13 Jun 2013 22:33:35 +0000 (15:33 -0700)]
x86: Fix /proc/mtrr with base/size more than 44bits

commit d5c78673b1b28467354c2c30c3d4f003666ff385 upstream.

On one sytem that mtrr range is more then 44bits, in dmesg we have
[    0.000000] MTRR default type: write-back
[    0.000000] MTRR fixed ranges enabled:
[    0.000000]   00000-9FFFF write-back
[    0.000000]   A0000-BFFFF uncachable
[    0.000000]   C0000-DFFFF write-through
[    0.000000]   E0000-FFFFF write-protect
[    0.000000] MTRR variable ranges enabled:
[    0.000000]   0 [000080000000-0000FFFFFFFF] mask 3FFF80000000 uncachable
[    0.000000]   1 [380000000000-38FFFFFFFFFF] mask 3F0000000000 uncachable
[    0.000000]   2 [000099000000-000099FFFFFF] mask 3FFFFF000000 write-through
[    0.000000]   3 [00009A000000-00009AFFFFFF] mask 3FFFFF000000 write-through
[    0.000000]   4 [381FFA000000-381FFBFFFFFF] mask 3FFFFE000000 write-through
[    0.000000]   5 [381FFC000000-381FFC0FFFFF] mask 3FFFFFF00000 write-through
[    0.000000]   6 [0000AD000000-0000ADFFFFFF] mask 3FFFFF000000 write-through
[    0.000000]   7 [0000BD000000-0000BDFFFFFF] mask 3FFFFF000000 write-through
[    0.000000]   8 disabled
[    0.000000]   9 disabled

but /proc/mtrr report wrong:
reg00: base=0x080000000 ( 2048MB), size= 2048MB, count=1: uncachable
reg01: base=0x80000000000 (8388608MB), size=1048576MB, count=1: uncachable
reg02: base=0x099000000 ( 2448MB), size=   16MB, count=1: write-through
reg03: base=0x09a000000 ( 2464MB), size=   16MB, count=1: write-through
reg04: base=0x81ffa000000 (8519584MB), size=   32MB, count=1: write-through
reg05: base=0x81ffc000000 (8519616MB), size=    1MB, count=1: write-through
reg06: base=0x0ad000000 ( 2768MB), size=   16MB, count=1: write-through
reg07: base=0x0bd000000 ( 3024MB), size=   16MB, count=1: write-through
reg08: base=0x09b000000 ( 2480MB), size=   16MB, count=1: write-combining

so bit 44 and bit 45 get cut off.

We have problems in arch/x86/kernel/cpu/mtrr/generic.c::generic_get_mtrr().
1. for base, we miss cast base_lo to 64bit before shifting.
Fix that by adding u64 casting.

2. for size, it only can handle 44 bits aka 32bits + page_shift
Fix that with 64bit mask instead of 32bit mask_lo, then range could be
more than 44bits.
At the same time, we need to update size_or_mask for old cpus that does
support cpuid 0x80000008 to get phys_addr. Need to set high 32bits
to all 1s, otherwise will not get correct size for them.

Also fix mtrr_add_page: it should check base and (base + size - 1)
instead of base and size, as base and size could be small but
base + size could bigger enough to be out of boundary. We can
use boot_cpu_data.x86_phys_bits directly to avoid size_or_mask.

So When are we going to have size more than 44bits? that is 16TiB.

after patch we have right ouput:
reg00: base=0x080000000 ( 2048MB), size= 2048MB, count=1: uncachable
reg01: base=0x380000000000 (58720256MB), size=1048576MB, count=1: uncachable
reg02: base=0x099000000 ( 2448MB), size=   16MB, count=1: write-through
reg03: base=0x09a000000 ( 2464MB), size=   16MB, count=1: write-through
reg04: base=0x381ffa000000 (58851232MB), size=   32MB, count=1: write-through
reg05: base=0x381ffc000000 (58851264MB), size=    1MB, count=1: write-through
reg06: base=0x0ad000000 ( 2768MB), size=   16MB, count=1: write-through
reg07: base=0x0bd000000 ( 3024MB), size=   16MB, count=1: write-through
reg08: base=0x09b000000 ( 2480MB), size=   16MB, count=1: write-combining

-v2: simply checking in mtrr_add_page according to hpa.

[ hpa: This probably wants to go into -stable only after having sat in
  mainline for a bit.  It is not a regression. ]

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Link: http://lkml.kernel.org/r/1371162815-29931-1-git-send-email-yinghai@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: Correct obj->mm_list link to dev_priv->dev_priv->mm.inactive_list
Xiong Zhang [Fri, 5 Jul 2013 10:53:29 +0000 (18:53 +0800)]
drm/i915: Correct obj->mm_list link to dev_priv->dev_priv->mm.inactive_list

commit 067556084a0e412013af6b0250a3143ae5afde6d upstream.

obj->mm_list link to dev_priv->mm.inactive_list/active_list
obj->global_list link to dev_priv->mm.unbound_list/bound_list

This regression has been introduced in

commit 93927ca52a55c23e0a6a305e7e9082e8411ac9fa
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Thu Jan 10 18:03:00 2013 +0100

    drm/i915: Revert shrinker changes from "Track unbound pages"

Cc: stable@vger.kernel.org
Signed-off-by: Xiong Zhang <xiong.y.zhang@intel.com>
[danvet: Add regression notice.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Zhouping Liu <zliu@redhat.com>
10 years agoperf tools: Revert regression in configuration of Python support
Michael Witten [Wed, 17 Apr 2013 02:23:16 +0000 (02:23 +0000)]
perf tools: Revert regression in configuration of Python support

commit a363a9da65d253fa7354ce5fd630f4f94df934cc upstream.

Among other things, the following:

  commit 31160d7feab786c991780d7f0ce2755a469e0e5e
  Date:   Tue Jan 8 16:22:36 2013 -0500
  perf tools: Fix GNU make v3.80 compatibility issue

attempts to aid the user by tapping into an existing error message,
as described in the commit message:

  ... Also fix an issue where _get_attempt was called with only
  one argument. This prevented the error message from printing
  the name of the variable that can be used to fix the problem.

or more precisely:

  -$(if $($(1)),$(call _ge_attempt,$($(1)),$(1)),$(call _ge_attempt,$(2)))
  +$(if $($(1)),$(call _ge_attempt,$($(1)),$(1)),$(call _ge_attempt,$(2),$(1)))

However, The "missing" argument was in fact missing on purpose; it's
absence is a signal that the error message should be skipped, because
the failure would be due to the default value, not any user-supplied
value.  This can be seen in how `_ge_attempt' uses `gea_err' (in the
config/utilities.mak file):

  _ge_attempt = $(if $(get-executable),$(get-executable),$(_gea_warn)$(call _gea_err,$(2)))
  _gea_warn = $(warning The path '$(1)' is not executable.)
  _gea_err  = $(if $(1),$(error Please set '$(1)' appropriately))

That is, because the argument is no longer missing, the value `$(1)'
(associated with `_gea_err') always evaluates to true, thus always
triggering the error condition that is meant to be reserved for
only the case when a user explicitly supplies an invalid value.

Concretely, the result is a regression in the Makefile's configuration
of python support; rather than gracefully disable support when the
relevant executables cannot be found according to default values, the
build process halts in error as though the user explicitly supplied
the values.

This new commit simply reverts the offending one-line change.

Reported-by: Pekka Enberg <penberg@kernel.org>
Link: http://lkml.kernel.org/r/CAOJsxLHv17Ys3M7P5q25imkUxQW6LE_vABxh1N3Tt7Mv6Ho4iw@mail.gmail.com
Signed-off-by: Michael Witten <mfwitten@gmail.com>
Cc: Mark Brown <broonie@sirena.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoiscsi-target: Fix iscsit_sequence_cmd reject handling for iser
Nicholas Bellinger [Tue, 30 Jul 2013 04:04:02 +0000 (04:04 +0000)]
iscsi-target: Fix iscsit_sequence_cmd reject handling for iser

commit 561bf15892375597ee59d473a704a3e634c4f311 upstream

This patch moves ISCSI_OP_REJECT failures into iscsit_sequence_cmd()
in order to avoid external iscsit_reject_cmd() reject usage for all
PDU types.

It also updates PDU specific handlers for traditional iscsi-target
code to not reset the session after posting a ISCSI_OP_REJECT during
setup.

(v2: Fix CMDSN_LOWER_THAN_EXP for ISCSI_OP_SCSI to call
     target_put_sess_cmd() after iscsit_sequence_cmd() failure)

Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoiscsi-target: Fix iscsit_add_reject* usage for iser
Nicholas Bellinger [Tue, 30 Jul 2013 04:04:01 +0000 (04:04 +0000)]
iscsi-target: Fix iscsit_add_reject* usage for iser

commit ba159914086f06532079fc15141f46ffe7e04a41 upstream

This patch changes iscsit_add_reject() + iscsit_add_reject_from_cmd()
usage to not sleep on iscsi_cmd->reject_comp to address a free-after-use
usage bug in v3.10 with iser-target code.

It saves ->reject_reason for use within iscsit_build_reject() so the
correct value for both transport cases.  It also drops the legacy
fail_conn parameter usage throughput iscsi-target code and adds
two iscsit_add_reject_cmd() and iscsit_reject_cmd helper functions,
along with various small cleanups.

(v2: Re-enable target_put_sess_cmd() to be called from
     iscsit_add_reject_from_cmd() for rejects invoked after
     target_get_sess_cmd() has been called)

Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Cc: Or Gerlitz <ogerlitz@mellanox.com>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoradeon kms: do not flush uninitialized hotplug work
Sergey Senozhatsky [Sun, 14 Jul 2013 11:03:27 +0000 (14:03 +0300)]
radeon kms: do not flush uninitialized hotplug work

commit a01c34f72e7cd2624570818f579b5ab464f93de2 upstream.

Fix a warning from lockdep caused by calling flush_work() for
uninitialized hotplug work. Initialize hotplug_work, audio_work
and reset_work upon successful radeon_irq_kms_init() completion
and thus perform hotplug flush_work only when rdev->irq.installed
is true.

[    4.790019] [drm] Loading CEDAR Microcode
[    4.790943] r600_cp: Failed to load firmware "radeon/CEDAR_smc.bin"
[    4.791152] [drm:evergreen_startup] *ERROR* Failed to load firmware!
[    4.791330] radeon 0000:01:00.0: disabling GPU acceleration

[    4.792633] INFO: trying to register non-static key.
[    4.792792] the code is fine but needs lockdep annotation.
[    4.792953] turning off the locking correctness validator.

[    4.793114] CPU: 2 PID: 1 Comm: swapper/0 Not tainted 3.11.0-rc0-dbg-10676-gfe56456-dirty #1816
[    4.793314] Hardware name: Acer             Aspire 5741G    /Aspire 5741G    , BIOS V1.20 02/08/2011
[    4.793507]  ffffffff821fd810 ffff8801530b9a18 ffffffff8160434e 0000000000000002
[    4.794155]  ffff8801530b9ad8 ffffffff810b8404 ffff8801530b0798 ffff8801530b0000
[    4.794789]  ffff8801530b9b00 0000000000000046 00000000000004c0 ffffffff00000000
[    4.795418] Call Trace:
[    4.795573]  [<ffffffff8160434e>] dump_stack+0x4e/0x82
[    4.795731]  [<ffffffff810b8404>] __lock_acquire+0x1a64/0x1d30
[    4.795893]  [<ffffffff814a87f0>] ? dev_vprintk_emit+0x50/0x60
[    4.796034]  [<ffffffff810b8fb4>] lock_acquire+0xa4/0x200
[    4.796216]  [<ffffffff8106cd75>] ? flush_work+0x5/0x280
[    4.796375]  [<ffffffff8106cdad>] flush_work+0x3d/0x280
[    4.796520]  [<ffffffff8106cd75>] ? flush_work+0x5/0x280
[    4.796682]  [<ffffffff810b659d>] ? trace_hardirqs_on_caller+0xfd/0x1c0
[    4.796862]  [<ffffffff8131d775>] ? delay_tsc+0x95/0xf0
[    4.797024]  [<ffffffff8141bb8b>] radeon_irq_kms_fini+0x2b/0x70
[    4.797186]  [<ffffffff814557c9>] evergreen_init+0x2a9/0x2e0
[    4.797347]  [<ffffffff813ebb1f>] radeon_device_init+0x5ef/0x700
[    4.797511]  [<ffffffff81335bc7>] ? pci_find_capability+0x47/0x50
[    4.797672]  [<ffffffff813edaed>] radeon_driver_load_kms+0x8d/0x150
[    4.797843]  [<ffffffff813ce426>] drm_get_pci_dev+0x166/0x280
[    4.798007]  [<ffffffff8116cff5>] ? kfree+0xf5/0x2e0
[    4.798168]  [<ffffffff813ea298>] ? radeon_pci_probe+0x98/0xd0
[    4.798329]  [<ffffffff813ea2aa>] radeon_pci_probe+0xaa/0xd0
[    4.798489]  [<ffffffff81339404>] pci_device_probe+0x84/0xe0
[    4.798644]  [<ffffffff814ac7d6>] driver_probe_device+0x76/0x240
[    4.798805]  [<ffffffff814aca73>] __driver_attach+0x93/0xa0
[    4.798948]  [<ffffffff814ac9e0>] ? __device_attach+0x40/0x40
[    4.799126]  [<ffffffff814aa82b>] bus_for_each_dev+0x6b/0xb0
[    4.799272]  [<ffffffff814ac2be>] driver_attach+0x1e/0x20
[    4.799434]  [<ffffffff814abec0>] bus_add_driver+0x1f0/0x280
[    4.799596]  [<ffffffff814ad0e4>] driver_register+0x74/0x150
[    4.799758]  [<ffffffff8133923d>] __pci_register_driver+0x5d/0x60
[    4.799936]  [<ffffffff81d16efc>] ? ttm_init+0x67/0x67
[    4.800081]  [<ffffffff813ce655>] drm_pci_init+0x115/0x130
[    4.800243]  [<ffffffff81d16efc>] ? ttm_init+0x67/0x67
[    4.800405]  [<ffffffff81d16f98>] radeon_init+0x9c/0xba
[    4.800586]  [<ffffffff810002ca>] do_one_initcall+0xfa/0x150
[    4.800746]  [<ffffffff81073f60>] ? parse_args+0x120/0x330
[    4.800909]  [<ffffffff81cdafae>] kernel_init_freeable+0x111/0x191
[    4.801052]  [<ffffffff81cda87a>] ? do_early_param+0x88/0x88
[    4.801233]  [<ffffffff815fb670>] ? rest_init+0x140/0x140
[    4.801393]  [<ffffffff815fb67e>] kernel_init+0xe/0x180
[    4.801556]  [<ffffffff8160dcac>] ret_from_fork+0x7c/0xb0
[    4.801718]  [<ffffffff815fb670>] ? rest_init+0x140/0x140

Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoxen/evtchn: avoid a deadlock when unbinding an event channel
David Vrabel [Fri, 19 Jul 2013 14:51:58 +0000 (15:51 +0100)]
xen/evtchn: avoid a deadlock when unbinding an event channel

commit 179fbd5a45f0d4034cc6fd37b8d367a3b79663c4 upstream.

Unbinding an event channel (either with the ioctl or when the evtchn
device is closed) may deadlock because disable_irq() is called with
port_user_lock held which is also locked by the interrupt handler.

Think of the IOCTL_EVTCHN_UNBIND is being serviced, the routine has
just taken the lock, and an interrupt happens. The evtchn_interrupt
is invoked, tries to take the lock and spins forever.

A quick glance at the code shows that the spinlock is a local IRQ
variant. Unfortunately that does not help as "disable_irq() waits for
the interrupt handler on all CPUs to stop running.  If the irq occurs
on another VCPU, it tries to take port_user_lock and can't because
the unbind ioctl is holding it." (from David). Hence we cannot
depend on the said spinlock to protect us. We could make it a system
wide IRQ disable spinlock but there is a better way.

We can piggyback on the fact that the existence of the spinlock is
to make get_port_user() checks be up-to-date. And we can alter those
checks to not depend on the spin lock (as it's protected by u->bind_mutex
in the ioctl) and can remove the unnecessary locking (this is
IOCTL_EVTCHN_UNBIND) path.

In the interrupt handler we cannot use the mutex, but we do not
need it.

"The unbind disables the irq before making the port user stale, so when
you clear it you are guaranteed that the interrupt handler that might
use that port cannot be running." (from David).

Hence this patch removes the spinlock usage on the teardown path
and piggybacks on disable_irq happening before we muck with the
get_port_user() data. This ensures that the interrupt handler will
never run on stale data.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
[v1: Expanded the commit description a bit]
Signed-off-by: Jonghwan Choi <jhbird.choi@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agolivelock avoidance in sget()
Al Viro [Fri, 19 Jul 2013 23:13:55 +0000 (03:13 +0400)]
livelock avoidance in sget()

commit acfec9a5a892f98461f52ed5770de99a3e571ae2 upstream.

Eric Sandeen has found a nasty livelock in sget() - take a mount(2) about
to fail.  The superblock is on ->fs_supers, ->s_umount is held exclusive,
->s_active is 1.  Along comes two more processes, trying to mount the same
thing; sget() in each is picking that superblock, bumping ->s_count and
trying to grab ->s_umount.  ->s_active is 3 now.  Original mount(2)
finally gets to deactivate_locked_super() on failure; ->s_active is 2,
superblock is still ->fs_supers because shutdown will *not* happen until
->s_active hits 0.  ->s_umount is dropped and now we have two processes
chasing each other:
s_active = 2, A acquired ->s_umount, B blocked
A sees that the damn thing is stillborn, does deactivate_locked_super()
s_active = 1, A drops ->s_umount, B gets it
A restarts the search and finds the same superblock.  And bumps it ->s_active.
s_active = 2, B holds ->s_umount, A blocked on trying to get it
... and we are in the earlier situation with A and B switched places.

The root cause, of course, is that ->s_active should not grow until we'd
got MS_BORN.  Then failing ->mount() will have deactivate_locked_super()
shut the damn thing down.  Fortunately, it's easy to do - the key point
is that grab_super() is called only for superblocks currently on ->fs_supers,
so it can bump ->s_count and grab ->s_umount first, then check MS_BORN and
bump ->s_active; we must never increment ->s_count for superblocks past
->kill_sb(), but grab_super() is never called for those.

The bug is pretty old; we would've caught it by now, if not for accidental
exclusion between sget() for block filesystems; the things like cgroup or
e.g. mtd-based filesystems don't have anything of that sort, so they get
bitten.  The right way to deal with that is obviously to fix sget()...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agotty_port: Fix refcounting leak in tty_port_tty_hangup()
Gianluca Anzolin [Thu, 25 Jul 2013 05:26:16 +0000 (07:26 +0200)]
tty_port: Fix refcounting leak in tty_port_tty_hangup()

commit 1d9e689c934bd5ecb0f273c6c65e0655c5cfee5f upstream.

The function tty_port_tty_hangup() could leak a reference to the tty_struct:

        struct tty_struct *tty = tty_port_tty_get(port);

        if (tty && (!check_clocal || !C_CLOCAL(tty))) {
                tty_hangup(tty);
                tty_kref_put(tty);
        }

If tty != NULL and the second condition is false we never call tty_kref_put and
the reference is leaked.

Fix by always calling tty_kref_put() which accepts a NULL argument.

The patch fixes a regression introduced by commit aa27a094.

Acked-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signed-off-by: Gianluca Anzolin <gianluca@sottospazio.it>
Acked-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agomm: mempolicy: fix mbind_range() && vma_adjust() interaction
Oleg Nesterov [Wed, 31 Jul 2013 20:53:28 +0000 (13:53 -0700)]
mm: mempolicy: fix mbind_range() && vma_adjust() interaction

commit 3964acd0dbec123aa0a621973a2a0580034b4788 upstream.

vma_adjust() does vma_set_policy(vma, vma_policy(next)) and this
is doubly wrong:

1. This leaks vma->vm_policy if it is not NULL and not equal to
   next->vm_policy.

   This can happen if vma_merge() expands "area", not prev (case 8).

2. This sets the wrong policy if vma_merge() joins prev and area,
   area is the vma the caller needs to update and it still has the
   old policy.

Revert commit 1444f92c8498 ("mm: merging memory blocks resets
mempolicy") which introduced these problems.

Change mbind_range() to recheck mpol_equal() after vma_merge() to fix
the problem that commit tried to address.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Steven T Hampson <steven.t.hampson@intel.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Andi Kleen <andi@firstfloor.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agousb: gadget: udc-core: fix the typo of udc state attribute
Rong Wang [Sun, 28 Jul 2013 15:01:35 +0000 (23:01 +0800)]
usb: gadget: udc-core: fix the typo of udc state attribute

commit 1894870eb4240399fabc6f0cb8c6fff4e6edbe83 upstream.

The name of udc state attribute file under sysfs is registered as
"state", while usb_gadget_set_state take it as "status" when it's
going to update. This patch fixes the typo.

Signed-off-by: Rong Wang <Rong.Wang@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Felipe Balbi <balbi@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: serial: ftdi_sio: add more RT Systems ftdi devices
Rick Farina (Zero_Chaos) [Mon, 29 Jul 2013 19:17:59 +0000 (15:17 -0400)]
USB: serial: ftdi_sio: add more RT Systems ftdi devices

commit fed1f1ed90bce42ea010e2904cbc04e7b8304940 upstream.

RT Systems makes many usb serial cables based on the ftdi_sio driver for
programming various amateur radios.  This patch is a full listing of
their current product offerings and should allow these cables to all
be recognized.

Signed-off-by: Rick Farina (Zero_Chaos) <zerochaos@gentoo.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agortlwifi: Initialize power-setting callback for USB devices
Larry Finger [Fri, 28 Jun 2013 14:12:53 +0000 (09:12 -0500)]
rtlwifi: Initialize power-setting callback for USB devices

commit bcfb879432094c267c35a7ff75d953d3a66c193a upstream.

Commit a269913c5 entitled "rtlwifi: Rework rtl_lps_leave() and
rtl_lps_enter() to use work queue" has two bugs for USB drivers.
Firstly, the work queue in question was not initialized. Secondly,
the callback routine used by this queue is contained within the
file used for PCI devices. As a result, it is not available for
architectures without PCI hardware.

Signed-off-by: Larry Finger <Larry.Finger@lwfinger.net>
Reported-by: Richard Genoud <richard.genoud@gmail.com>
Tested-by: Richard Genoud <richard.genoud@gmail.com>
Cc: Richard Genoud <richard.genoud@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/radeon/atom: initialize more atom interpretor elements to 0
Alex Deucher [Tue, 30 Jul 2013 04:22:53 +0000 (00:22 -0400)]
drm/radeon/atom: initialize more atom interpretor elements to 0

commit 42a21826dc54583cdb79cc8477732e911ac9c376 upstream.

The ProcessAuxChannel table on some rv635 boards assumes
the divmul members are initialized to 0 otherwise we get
an invalid fb offset since it has a bad mask set when
setting the fb base.  While here initialize all the
atom interpretor elements to 0.

Fixes:
https://bugzilla.kernel.org/show_bug.cgi?id=60639

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/radeon: fix audio dto programming on DCE4+
Alex Deucher [Fri, 26 Jul 2013 17:26:05 +0000 (13:26 -0400)]
drm/radeon: fix audio dto programming on DCE4+

commit 7d61d835824f73dc4097b51f800382467c8049c5 upstream.

We need to set the dto source before setting the
dividers otherwise we may get stability problems
with the dto leading to audio playback problems.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/nouveau: fix semaphore dmabuf obj
Maarten Lankhorst [Tue, 23 Jul 2013 13:49:39 +0000 (15:49 +0200)]
drm/nouveau: fix semaphore dmabuf obj

commit 7a7da592cbb22a1d360638dbecc393470c5effe3 upstream.

Fixes some dmabuf object errors on nv50 chipset and below.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: fix missed hunk after GT access breakage
Ben Widawsky [Tue, 30 Jul 2013 23:27:57 +0000 (16:27 -0700)]
drm/i915: fix missed hunk after GT access breakage

commit e1b4d3036c07ff137955fb1c0197ab62534f46ec upstream.

Upon some code refactoring, a hunk was missed. This was fixed for
next, but missed the current trees, and hasn't yet been merged by Dave
Airlie. It is fixed in:
commit 907b28c56ea40629aa6595ddfa414ec2fc7da41c
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Fri Jul 19 20:36:52 2013 +0100

    drm/i915: Colocate all GT access routines in the same file

It is introduced by:
commit 181d1b9e31c668259d3798c521672afb8edd355c
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date:   Sun Jul 21 13:16:24 2013 +0200

    drm/i915: fix up gt init sequence fallout

Reported-by: Dave Jones <davej@redhat.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: fix up gt init sequence fallout
Daniel Vetter [Sun, 21 Jul 2013 11:16:24 +0000 (13:16 +0200)]
drm/i915: fix up gt init sequence fallout

commit 181d1b9e31c668259d3798c521672afb8edd355c upstream.

The regression fix for gen6+ rps fallout

commit 7dcd2677ea912573d9ed4bcd629b0023b2d11505
Author: Konstantin Khlebnikov <khlebnikov@openvz.org>
Date:   Wed Jul 17 10:22:58 2013 +0400

    drm/i915: fix long-standing SNB regression in power consumption after resume

unintentionally also changed the init sequence ordering between
gt_init and gt_reset - we need to reset BIOS damage like leftover
forcewake references before we run our own code. Otherwise we can get
nasty dmesg noise like

[drm:__gen6_gt_force_wake_mt_get] *ERROR* Timed out waiting for forcewake old ack to clear.

again. Since _reset suggests that we first need to have stuff
initialized (which isn't the case here) call it sanitze instead.

While at it also block out the rps disable introduced by the above
commit on ilk: We don't have any knowledge of ilk rps being broken in
similar ways. And the disable functions uses the default hw state
which is only read out when we're enabling rps. So essentially we've
been writing random grabage into that register.

Reported-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: Serialize almost all register access
Chris Wilson [Fri, 19 Jul 2013 19:36:51 +0000 (20:36 +0100)]
drm/i915: Serialize almost all register access

commit a7cd1b8fea2f341b626b255d9898a5ca5fabbf0a upstream.

In theory, the different register blocks were meant to be only ever
touched when holding either the struct_mutex, mode_config.lock or even a
specific localised lock. This does not seem to be the case, and the
hardware reacts extremely badly if we attempt to concurrently access two
registers within the same cacheline.

The HSD suggests that we only need to do this workaround for display
range registers. However, upon review we need to serialize the multiple
stages in our register write functions - if only for preemption
protection.

Irrespective of the hardware requirements, the current io functions are
a little too loose with respect to the combination of pre- and
post-condition testing that we do in conjunction with the actual io. As
a result, we may be pre-empted and generate both false-postive and
false-negative errors.

Note well that this is a "90%" solution, there remains a few direct
users of ioread/iowrite which will be fixed up in the next few patches.
Since they are more invasive and that this simple change will prevent
almost all lockups on Haswell, we kept this patch simple to facilitate
backporting to stable.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=63914
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: quirk no PCH_PWM_ENABLE for Dell XPS13 backlight
Kamal Mostafa [Fri, 19 Jul 2013 22:02:01 +0000 (15:02 -0700)]
drm/i915: quirk no PCH_PWM_ENABLE for Dell XPS13 backlight

commit e85843bec6c2ea7c10ec61238396891cc2b753a9 upstream.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=47941
BugLink: https://bugs.launchpad.net/bugs/1163720
BugLink: https://bugs.launchpad.net/bugs/1162026
Some machines suffer from non-functional backlight controls if
BLM_PCH_PWM_ENABLE is set, so provide a quirk to avoid doing so.
Apply this quirk to Dell XPS 13 models.

Tested-by: Eric Griffith <EGriffith92@gmail.com>
Tested-by: Kent Baxley <kent.baxley@canonical.com>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: correctly restore fences with objects attached
Daniel Vetter [Wed, 17 Jul 2013 12:51:28 +0000 (14:51 +0200)]
drm/i915: correctly restore fences with objects attached

commit 94a335dba34ff47cad3d6d0c29b452d43a1be3c8 upstream.

To avoid stalls we delay tiling changes and especially hold of
committing the new fence state for as long as possible.
Synchronization points are in the execbuf code and in our gtt fault
handler.

Unfortunately we've missed that tricky detail when adding proper fence
restore code in

commit 19b2dbde5732170a03bd82cc8bd442cf88d856f7
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Jun 12 10:15:12 2013 +0100

    drm/i915: Restore fences after resume and GPU resets

The result was that we've restored fences for objects with no tiling,
since the object<->fence link still existed after resume. Now that
wouldn't have been too bad since any subsequent access would have
fixed things up, but if we've changed from tiled to untiled real havoc
happened:

The tiling stride is stored -1 in the fence register, so a stride of 0
resulted in all 1s in the top 32bits, and so a completely bogus fence
spanning everything from the start of the object to the top of the
GTT. The tell-tale in the register dumps looks like:

                 FENCE START 2: 0x0214d001
                 FENCE END 2: 0xfffff3ff

Bit 11 isn't set since the hw doesn't store it, even when writing all
1s (at least on my snb here).

To prevent such a gaffle in the future add a sanity check for fences
with an untiled object attached in i915_gem_write_fence.

v2: Fix the WARN, spotted by Chris.

v3: Trying to reuse get_fences looked ugly and obfuscated the code.
Instead reuse update_fence and to make it really dtrt also move the
fence dirty state clearing into update_fence.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=60530
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Stéphane Marchesin <marcheu@chromium.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested-by: Matthew Garrett <matthew.garrett@nebula.com>
Tested-by: Björn Bidar <theodorstormgrade@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: Fix dereferencing invalid connectors in is_crtc_connector_off()
Chris Wilson [Wed, 17 Jul 2013 11:14:40 +0000 (12:14 +0100)]
drm/i915: Fix dereferencing invalid connectors in is_crtc_connector_off()

commit 2e57f47d317dd035b18634b0c602272529368fcc upstream.

In commit e3de42b68478a8c95dd27520e9adead2af9477a5
Author: Imre Deak <imre.deak@intel.com>
Date:   Fri May 3 19:44:07 2013 +0200

    drm/i915: force full modeset if the connector is in DPMS OFF mode

a new function was added that walked over the set of connectors to see
if any of the currently associated CRTC was switched off. This function
walked an array of connectors, rather than the array of pointers to
connectors contained in the drm_mode_set - i.e. it was dereferencing far
past the end of the first connector. This only becomes an issue if we
attempt to use a clone mode (i.e. more than one connector per CRTC) such
that set->num_connectors > 1.

Reported-by: Timo Aaltonen <tjaalton@ubuntu.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=65927
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Imre Deak <imre.deak@intel.com>
Cc: Egbert Eich <eich@suse.de>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: fix long-standing SNB regression in power consumption after resume v2
Konstantin Khlebnikov [Wed, 17 Jul 2013 06:22:58 +0000 (10:22 +0400)]
drm/i915: fix long-standing SNB regression in power consumption after resume v2

commit 7dcd2677ea912573d9ed4bcd629b0023b2d11505 upstream.

This patch fixes regression in power consumtion of sandy bridge gpu, which
exists since v3.6 Sometimes after resuming from s2ram gpu starts thinking that
it's extremely busy. After that it never reaches rc6 state.

Bug exists since kernel v3.6:

commit b4ae3f22d238617ca11610b29fde16cf8c0bc6e0
Author: Jesse Barnes <jbarnes@virtuousgeek.org>
Date:   Thu Jun 14 11:04:48 2012 -0700

    drm/i915: load boot context at driver init time

For some reason RC6 is already enabled at the beginning of resuming process.
Following initliaztion breaks some internal state and confuses RPS engine.
This patch disables RC6 at the beginnig of resume and initialization.

I've rearranged initialization sequence, because intel_disable_gt_powersave()
needs initialized force_wake_get/put and some locks from the dev_priv.

Note: The culprit in the initialization sequence seems to be the write
to MBCTL added in the above mentioned commit. The first version of
this patch just held a forcewake reference across the clock gating
init functions, which seems to have been enought to gather quite a few
positive test reports. But since that smelled a bit like ad-hoc
duct-tape v2 now just disables rps/rc6 across the entire hw setup.

[danvet: Add note about v1 vs. v2 of this patch and use standard
layout for the commit citation. Also add the tested-bys from v1 and a cc:
stable.]

References https://bugs.freedesktop.org/show_bug.cgi?id=54089
References https://bugzilla.kernel.org/show_bug.cgi?id=58971
References https://patchwork.kernel.org/patch/2827634/ (patch v1)

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Tested-by: Alexander Kaltsas <alexkaltsas@gmail.com> (v1)
Tested-by: rocko <rockorequin@hotmail.com> (v1)
Tested-by: JohnMB <johnmbryant@sky.com> (v1)
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: Fix incoherence with fence updates on Sandybridge+
Chris Wilson [Wed, 10 Jul 2013 12:36:23 +0000 (13:36 +0100)]
drm/i915: Fix incoherence with fence updates on Sandybridge+

commit d18b9619034230b6f945e215276425636ca401fe upstream.

This hopefully fixes the root cause behind the workaround added in

commit 25ff1195f8a0b3724541ae7bbe331b4296de9c06
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Thu Apr 4 21:31:03 2013 +0100

    drm/i915: Workaround incoherence between fences and LLC across multiple CPUs

Thanks to further investigation by Jon Bloomfield, he realised that
the 64-bit register might be broken up by the hardware into two 32-bit
writes (a problem we have encountered elsewhere). This non-atomicity
would then cause an issue where a second thread would see an
intermediate register state (new high dword, old low dword), and this
register would randomly be used in preference to its own thread register.
This would cause the second thread to read from and write into a fairly
random tiled location.  Breaking the operation into 3 explicit 32-bit
updates (first disable the fence, poke the upper bits, then poke the lower
bits and enable) ensures that, given proper serialisation between the
32-bit register write and the memory transfer, that the fence value is
always consistent.

Armed with this knowledge, we can explain how the previous workaround
work. The key to the corruption is that a second thread sees an
erroneous fence register that conflicts and overrides its own. By
serialising the fence update across all CPUs, we have a small window
where no GTT access is occurring and so hide the potential corruption.
This also leads to the conclusion that the earlier workaround was
incomplete.

v2: Be overly paranoid about the order in which fence updates become
visible to the GPU to make really sure that we turn the fence off before
doing the update, and then only switch the fence on afterwards.

Signed-off-by: Jon Bloomfield <jon.bloomfield@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Carsten Emde <C.Emde@osadl.org>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoPartially revert "drm/i915: unconditionally use mt forcewake on hsw/ivb"
Guenter Roeck [Tue, 9 Jul 2013 23:00:31 +0000 (16:00 -0700)]
Partially revert "drm/i915: unconditionally use mt forcewake on hsw/ivb"

commit c11e5f35ab490bd30591563816fbc83526521777 upstream.

This patch partially reverts commit 36ec8f877481449bdfa072e6adf2060869e2b970 for
IvyBridge CPUs.

The original commit results in repeated 'Timed out waiting for forcewake old
ack to clear' messages on a Supermicro C7H61 board (BIOS version 2.00 and 2.00b)
with i7-3770K CPU. It ultimately results in a hangup if the system is highly
loaded. Reverting the commit for IvyBridge CPUs fixes the issue.

Issue a warning if the CPU is IvyBridge and mt forcewake is disabled, since
this condition can result in secondary issues.

v2: Only revert patch for Ivybridge CPUs
    Issue info message if mt forcewake is disabled on Ivybridge

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=60541
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=66139
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agodrm/i915: Fix write-read race with multiple rings
Chris Wilson [Tue, 9 Jul 2013 08:22:39 +0000 (09:22 +0100)]
drm/i915: Fix write-read race with multiple rings

commit 02978ff57a5bdfbf703d2bc5a4d933a53ede3144 upstream.

Daniel noticed a problem where is we wrote to an object with ring A in
the middle of a very long running batch, then executed a quick batch on
ring B before a batch that reads from the same object, its obj->ring would
now point to ring B, but its last_write_seqno would be still relative to
ring A. This would allow for the user to read from the object before the
GPU had completed the write, as set_domain would only check that ring B
had passed the last_write_seqno.

To fix this simply (and inelegantly), we bump the last_write_seqno when
switching rings so that the last_write_seqno is always relative to the
current obj->ring.

This fixes igt/tests/gem_write_read_ring_switch.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
[danvet: Add note about the newly created igt which exercises this bug.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>