]> Git Repo - linux.git/log
linux.git
9 years agox86/nmi/64: Make the "NMI executing" variable more consistent
Andy Lutomirski [Wed, 15 Jul 2015 17:29:40 +0000 (10:29 -0700)]
x86/nmi/64: Make the "NMI executing" variable more consistent

Currently, "NMI executing" is one the first time an outermost
NMI hits repeat_nmi and zero thereafter.  Change it to be zero
each time for consistency.

This is intended to help NMI handling fail harder if it's buggy.

Signed-off-by: Andy Lutomirski <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
9 years agox86/nmi/64: Minor asm simplification
Andy Lutomirski [Wed, 15 Jul 2015 17:29:39 +0000 (10:29 -0700)]
x86/nmi/64: Minor asm simplification

Replace LEA; MOV with an equivalent SUB.  This saves one
instruction.

Signed-off-by: Andy Lutomirski <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
9 years agox86/nmi/64: Use DF to avoid userspace RSP confusing nested NMI detection
Andy Lutomirski [Wed, 15 Jul 2015 17:29:38 +0000 (10:29 -0700)]
x86/nmi/64: Use DF to avoid userspace RSP confusing nested NMI detection

We have a tricky bug in the nested NMI code: if we see RSP
pointing to the NMI stack on NMI entry from kernel mode, we
assume that we are executing a nested NMI.

This isn't quite true.  A malicious userspace program can point
RSP at the NMI stack, issue SYSCALL, and arrange for an NMI to
happen while RSP is still pointing at the NMI stack.

Fix it with a sneaky trick.  Set DF in the region of code that
the RSP check is intended to detect.  IRET will clear DF
atomically.

( Note: other than paravirt, there's little need for all this
  complexity. We could check RIP instead of RSP. )

Signed-off-by: Andy Lutomirski <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
9 years agox86/nmi/64: Reorder nested NMI checks
Andy Lutomirski [Wed, 15 Jul 2015 17:29:37 +0000 (10:29 -0700)]
x86/nmi/64: Reorder nested NMI checks

Check the repeat_nmi .. end_repeat_nmi special case first.  The
next patch will rework the RSP check and, as a side effect, the
RSP check will no longer detect repeat_nmi .. end_repeat_nmi, so
we'll need this ordering of the checks.

Note: this is more subtle than it appears.  The check for
repeat_nmi .. end_repeat_nmi jumps straight out of the NMI code
instead of adjusting the "iret" frame to force a repeat.  This
is necessary, because the code between repeat_nmi and
end_repeat_nmi sets "NMI executing" and then writes to the
"iret" frame itself.  If a nested NMI comes in and modifies the
"iret" frame while repeat_nmi is also modifying it, we'll end up
with garbage.  The old code got this right, as does the new
code, but the new code is a bit more explicit.

If we were to move the check right after the "NMI executing"
check, then we'd get it wrong and have random crashes.

( Because the "NMI executing" check would jump to the code that would
  modify the "iret" frame without checking if the interrupted NMI was
  currently modifying it. )

Signed-off-by: Andy Lutomirski <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
9 years agox86/nmi/64: Improve nested NMI comments
Andy Lutomirski [Wed, 15 Jul 2015 17:29:36 +0000 (10:29 -0700)]
x86/nmi/64: Improve nested NMI comments

I found the nested NMI documentation to be difficult to follow.
Improve the comments.

Signed-off-by: Andy Lutomirski <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
9 years agox86/nmi/64: Switch stacks on userspace NMI entry
Andy Lutomirski [Wed, 15 Jul 2015 17:29:35 +0000 (10:29 -0700)]
x86/nmi/64: Switch stacks on userspace NMI entry

Returning to userspace is tricky: IRET can fail, and ESPFIX can
rearrange the stack prior to IRET.

The NMI nesting fixup relies on a precise stack layout and
atomic IRET.  Rather than trying to teach the NMI nesting fixup
to handle ESPFIX and failed IRET, punt: run NMIs that came from
user mode on the normal kernel stack.

This will make some nested NMIs visible to C code, but the C
code is okay with that.

As a side effect, this should speed up perf: it eliminates an
RDMSR when NMIs come from user mode.

Signed-off-by: Andy Lutomirski <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Reviewed-by: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
9 years agox86/nmi/64: Remove asm code that saves CR2
Andy Lutomirski [Wed, 15 Jul 2015 17:29:34 +0000 (10:29 -0700)]
x86/nmi/64: Remove asm code that saves CR2

Now that do_nmi saves CR2, we don't need to save it in asm.

Signed-off-by: Andy Lutomirski <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Acked-by: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
9 years agox86/nmi: Enable nested do_nmi() handling for 64-bit kernels
Andy Lutomirski [Wed, 15 Jul 2015 17:29:33 +0000 (10:29 -0700)]
x86/nmi: Enable nested do_nmi() handling for 64-bit kernels

32-bit kernels handle nested NMIs in C.  Enable the exact same
handling on 64-bit kernels as well.  This isn't currently
necessary, but it will become necessary once the asm code starts
allowing limited nesting.

Signed-off-by: Andy Lutomirski <[email protected]>
Reviewed-by: Steven Rostedt <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: [email protected]
Signed-off-by: Ingo Molnar <[email protected]>
9 years agoirqchip/gicv3-its: Fix mapping of LPIs to collections
Marc Zyngier [Fri, 17 Jul 2015 09:46:42 +0000 (10:46 +0100)]
irqchip/gicv3-its: Fix mapping of LPIs to collections

The GICv3 ITS architecture allows a given [DevID, EventID] pair to be
translated to a [LPI, Collection] pair, where DevID is the device writing
the MSI, EventID is the payload being written, LPI is the actual
interrupt number, and Collection is roughly equivalent to a target CPU.

Each LPI can be mapped to a separate collection, but the ITS driver
insists on maintaining the collection on a device basis, instead of doing
it on a per interrupt basis.

This is obviously flawed, and this patch fixes it by adding a per interrupt
index that indicates which collection number is in use.

Reported-by: Ian Campbell <[email protected]>
Signed-off-by: Marc Zyngier <[email protected]>
Cc: <[email protected]>
Cc: Jason Cooper <[email protected]>
Cc: [email protected] # 4.1, 4.0
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
9 years agogenirq: Prevent resend to interrupts marked IRQ_NESTED_THREAD
Thomas Gleixner [Thu, 16 Jul 2015 12:10:17 +0000 (14:10 +0200)]
genirq: Prevent resend to interrupts marked IRQ_NESTED_THREAD

The resend mechanism happily calls the interrupt handler of interrupts
which are marked IRQ_NESTED_THREAD from softirq context. This can
result in crashes because the interrupt handler is not the proper way
to invoke the device handlers. They must be invoked via
handle_nested_irq.

Prevent the resend even if the interrupt has no valid parent irq
set. Its better to have a lost interrupt than a crashing machine.

Reported-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: [email protected]
9 years agomac80211: shut down interfaces before destroying interface list
Johannes Berg [Wed, 8 Jul 2015 12:41:50 +0000 (15:41 +0300)]
mac80211: shut down interfaces before destroying interface list

If the hardware is unregistered while interfaces are up, mac80211 will
unregister all interfaces, which in turns causes mac80211 to be called
again to remove them all from the driver and eventually shut down the
hardware.

During this shutdown, however, it's currently already unsafe to iterate
the list of interfaces atomically, as the list is manipulated in an
unsafe manner. This puts an undue burden on the driver - it must stop
all its activities before calling ieee80211_unregister_hw(), while in
the normal stop path it can do all cleanup in the stop method. If, for
example, it's using the iteration during RX for some reason, it would
have to stop RX before unregistering to avoid crashes.

Fix this problem by closing all interfaces before unregistering them.
This will cause the driver stop to have completed before we manipulate
the interface list, and after the driver is stopped *and* has called
ieee80211_unregister_hw() it really musn't be iterating any more as
the memory will be freed as well.

Signed-off-by: Johannes Berg <[email protected]>
9 years agomac80211: wowlan: enable powersave if suspend while ps-polling
Chaitanya T K [Wed, 10 Jun 2015 13:42:31 +0000 (19:12 +0530)]
mac80211: wowlan: enable powersave if suspend while ps-polling

If for any reason we're in the middle of PS-polling or awake after
TX due to dynamic powersave while going to suspend, go back to save
power. This might cause a response frame to get lost, but since we
can't really wait for it while going to suspend that's still better
than not enabling powersave which would cause higher power usage
during (and possibly even after) suspend.

Note that this really only affects the very few drivers that use
the powersave implementation in mac80211.

Signed-off-by: Chaitanya T K <[email protected]>
[rewrite misleading commit log]
Signed-off-by: Johannes Berg <[email protected]>
9 years agomac80211: don't clear all tx flags when requeing
Michal Kazior [Thu, 2 Jul 2015 07:59:56 +0000 (09:59 +0200)]
mac80211: don't clear all tx flags when requeing

When acting as AP and a PS-Poll frame is received
associated station is marked as one in a Service
Period. This state is kept until Tx status for
released frame is reported. While a station is in
Service Period PS-Poll frames are ignored.

However if PS-Poll was received during A-MPDU
teardown it was possible to have the to-be
released frame re-queued back to pending queue.
In such case the frame was stripped of 2 important
flags:

 (a) IEEE80211_TX_CTL_NO_PS_BUFFER
 (b) IEEE80211_TX_STATUS_EOSP

Stripping of (a) led to the frame that was to be
released to be queued back to ps_tx_buf queue. If
station remained to use only PS-Poll frames the
re-queued frame (and new ones) was never actually
transmitted because mac80211 would ignore
subsequent PS-Poll frames due to station being in
Service Period. There was nothing left to clear
the Service Period bit (no xmit -> no tx status ->
no SP end), i.e. the AP would have the station
stuck in Service Period. Beacon TIM would
repeatedly prompt station to poll for frames but
it would get none.

Once (a) is not stripped (b) becomes important
because it's the main condition to clear the
Service Period bit of the station when Tx status
for the released frame is reported back.

This problem was observed with ath9k acting as P2P
GO in some testing scenarios but isn't limited to
it. AP operation with mac80211 based Tx A-MPDU
control combined with clients using PS-Poll frames
is subject to this race.

Signed-off-by: Michal Kazior <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
9 years agomac80211: clear subdir_stations when removing debugfs
Tom Hughes [Mon, 29 Jun 2015 18:41:49 +0000 (19:41 +0100)]
mac80211: clear subdir_stations when removing debugfs

If we don't do this, and we then fail to recreate the debugfs
directory during a mode change, then we will fail later trying
to add stations to this now bogus directory:

BUG: unable to handle kernel NULL pointer dereference at 0000006c
IP: [<c0a92202>] mutex_lock+0x12/0x30
Call Trace:
[<c0678ab4>] start_creating+0x44/0xc0
[<c0679203>] debugfs_create_dir+0x13/0xf0
[<f8a938ae>] ieee80211_sta_debugfs_add+0x6e/0x490 [mac80211]

Cc: [email protected]
Signed-off-by: Tom Hughes <[email protected]>
Signed-off-by: Johannes Berg <[email protected]>
9 years agoMerge tag 'iwlwifi-for-kalle-2015-06-12' of https://git.kernel.org/pub/scm/linux...
Kalle Valo [Fri, 17 Jul 2015 08:19:08 +0000 (11:19 +0300)]
Merge tag 'iwlwifi-for-kalle-2015-06-12' of https://git.kernel.org/pub/scm/linux/kernel/git/iwlwifi/iwlwifi-fixes

* bug fixes specific for 8000 series
* fix a crash in time events
* fix a crash in PCIe transport
* fix BT Coex code that prevented association on certain
        devices (3160).
* revert the new RBD allocation model because it introduced
a bug when running on weak VM setups.
* a new device IDs

9 years agodrm/ttm: improve uncached page deallocation.
Jérôme Glisse [Thu, 9 Jul 2015 18:19:30 +0000 (14:19 -0400)]
drm/ttm: improve uncached page deallocation.

Calls to set_memory_wb() incure heavy TLB flush and IPI cost. To
minimize those wait until pool grow beyond batch size before
draining the pool.

Signed-off-by: Jérôme Glisse <[email protected]>
Reviewed-by: Mario Kleiner <[email protected]>
Reviewed-and-Tested-by: Michel Dänzer <[email protected]>
Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
9 years agodrm/ttm: fix uncached page deallocation to properly fill page pool v3.
Jérôme Glisse [Thu, 9 Jul 2015 18:19:29 +0000 (14:19 -0400)]
drm/ttm: fix uncached page deallocation to properly fill page pool v3.

Current code never allowed the page pool to actualy fill in anyway.
This fix it, so that we only start freeing page from the pool when
we go over the pool size.

Changed since v1:
  - Move the page batching optimization to its separate patch.

Changed since v2:
  - Do not remove code part of the batching optimization with
    this patch.
  - Better commit message.

Signed-off-by: Jérôme Glisse <[email protected]>
Reviewed-by: Mario Kleiner <[email protected]>
Reviewed-and-Tested-by: Michel Dänzer <[email protected]>
Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
Cc: Thomas Hellstrom <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Signed-off-by: Dave Airlie <[email protected]>
9 years agopinctrl: samsung: Remove old unused defines
Krzysztof Kozlowski [Mon, 15 Jun 2015 04:54:29 +0000 (13:54 +0900)]
pinctrl: samsung: Remove old unused defines

Since 9a2c1c3b91aa ("pinctrl: samsung: Allow grouping multiple
pinmux/pinconf nodes") the defines for GPIO group and function names are
not used anywhere in the driver.

Signed-off-by: Krzysztof Kozlowski <[email protected]>
Inspired-by: Dan Carpenter <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agoravb: do not invalidate cache for RX buffer twice
Sergei Shtylyov [Tue, 14 Jul 2015 21:56:52 +0000 (00:56 +0300)]
ravb: do not invalidate cache for RX buffer twice

First, dma_sync_single_for_cpu() shouldn't have been called in the first place
(it's a streaming DMA API), dma_unmap_single() should have been called instead.
Second, dma_unmap_single() call after handing the buffer to napi_gro_receive()
makes little sense.  Moreover desc->dptr might not be valid at this point.

Signed-off-by: Sergei Shtylyov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoMerge tag 'pm+acpi-4.2-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 17 Jul 2015 04:10:53 +0000 (21:10 -0700)]
Merge tag 'pm+acpi-4.2-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm

Pull power management and ACPI fixes from Rafael Wysocki:
 "These fix two bugs in the cpufreq core (including one recent
  regression), fix a 4.0 PCI regression related to the ACPI resources
  management and quieten an RCU-related lockdep complaint about a
  tracepoint in the suspend-to-idle code.

  Specifics:

   - Fix a recently introduced issue in the cpufreq policy object
     reinitialization that leads to CPU offline/online breakage (Viresh
     Kumar)

   - Make it possible to access frequency tables of offline CPUs which
     is needed by thermal management code among other things (Viresh
     Kumar)

   - Fix an ACPI resource management regression introduced during the
     4.0 cycle that may cause incorrect resource validation results to
     appear in 32-bit x86 kernels due to silent truncation of 64-bit
     values to 32-bit (Jiang Liu)

   - Fix up an RCU-related lockdep complaint about suspicious RCU usage
     in idle caused by using a suspend tracepoint in the core suspend-
     to-idle code (Rafael J Wysocki)"

* tag 'pm+acpi-4.2-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI / PCI: Fix regressions caused by resource_size_t overflow with 32-bit kernel
  cpufreq: Allow freq_table to be obtained for offline CPUs
  cpufreq: Initialize the governor again while restoring policy
  suspend-to-idle: Prevent RCU from complaining about tick_freeze()

9 years agoMerge tag 'platform-drivers-x86-v4.2-3' of git://git.infradead.org/users/dvhart/linux...
Linus Torvalds [Fri, 17 Jul 2015 03:57:25 +0000 (20:57 -0700)]
Merge tag 'platform-drivers-x86-v4.2-3' of git://git.infradead.org/users/dvhart/linux-platform-drivers-x86

Pull x86 platform driver fixes from Darren Hart:
 "Fix SMBIOS call handling and hwswitch state coherency in the
  dell-laptop driver.  Cleanups for intel_*_ipc drivers.  Details:

  dell-laptop:
   - Do not cache hwswitch state
   - Check return value of each SMBIOS call
   - Clear buffer before each SMBIOS call

  intel_scu_ipc:
   - Move local memory initialization out of a mutex

  intel_pmc_ipc:
   - Update kerneldoc formatting
   - Fix compiler casting warnings"

* tag 'platform-drivers-x86-v4.2-3' of git://git.infradead.org/users/dvhart/linux-platform-drivers-x86:
  intel_scu_ipc: move local memory initialization out of a mutex
  intel_pmc_ipc: Update kerneldoc formatting
  dell-laptop: Do not cache hwswitch state
  dell-laptop: Check return value of each SMBIOS call
  dell-laptop: Clear buffer before each SMBIOS call
  intel_pmc_ipc: Fix compiler casting warnings

9 years agoMerge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu
Linus Torvalds [Fri, 17 Jul 2015 03:49:09 +0000 (20:49 -0700)]
Merge branch 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu

Pull m68knommu/coldfire fixes from Greg Ungerer:
 "Contains build fixes and updates for the ColdFire defconfigs.

  Specifically there is a couple of fixes that address problems building
  allnoconfig.  Also fix for enabling PCI bus on the M54xx family of
  ColdFire"

* 'for-next' of git://git.kernel.org/pub/scm/linux/kernel/git/gerg/m68knommu:
  m68k: enable PCI support for m5475evb defconfig
  m68k: fix io functions for ColdFire/MMU/PCI case
  m68knommu: update defconfig for ColdFire m5475evb
  m68knommu: update defconfig for ColdFire m5407c3
  m68knommu: update defconfig for ColdFire m5307c3
  m68knommu: update defconfig for ColdFire m5275evb
  m68knommu: update defconfig for ColdFire m5272c3
  m68knommu: update defconfig for ColdFire m5249evb
  m68knommu: update defconfig for m5208evb
  m68knommu: make ColdFire SoC selection a choice
  m68knommu: improve the clock configuration defaults
  m68knommu: force setting of CONFIG_CLOCK_FREQ for ColdFire

9 years agodm cache: avoid calls to prealloc_free_structs() if possible
Mike Snitzer [Fri, 17 Jul 2015 01:48:55 +0000 (21:48 -0400)]
dm cache: avoid calls to prealloc_free_structs() if possible

If no work was performed then prealloc_data_structs() wasn't ever called
so there isn't any need to call prealloc_free_structs().

Signed-off-by: Mike Snitzer <[email protected]>
9 years agodm cache: avoid preallocation if no work in writeback_some_dirty_blocks()
Mike Snitzer [Fri, 17 Jul 2015 01:26:10 +0000 (21:26 -0400)]
dm cache: avoid preallocation if no work in writeback_some_dirty_blocks()

Refactor writeback_some_dirty_blocks() to avoid prealloc_data_structs()
if the policy doesn't have any dirty blocks ready for writeback.

Signed-off-by: Mike Snitzer <[email protected]>
9 years agodm cache: do not wake_worker() in free_migration()
Mike Snitzer [Fri, 17 Jul 2015 01:16:31 +0000 (21:16 -0400)]
dm cache: do not wake_worker() in free_migration()

All methods that queue work call wake_worker() as you'd expect.
E.g. cell_defer, defer_bio, quiesce_migration (which is called by
writeback, promote, demote_then_promote, invalidate, discard, etc).

Signed-off-by: Mike Snitzer <[email protected]>
9 years agoMerge branch 'drm-rockchip-2015-07-13' of https://github.com/markyzq/kernel-drm-rockc...
Dave Airlie [Fri, 17 Jul 2015 00:25:02 +0000 (10:25 +1000)]
Merge branch 'drm-rockchip-2015-07-13' of https://github.com/markyzq/kernel-drm-rockchip into drm-fixes

misc rockchip fixes.

* 'drm-rockchip-2015-07-13' of https://github.com/markyzq/kernel-drm-rockchip:
  drm/rockchip: vop: remove hardware cursor window
  drm/rockchip: vop: switch cursor plane to window 3
  drm/rockchip: Drop owner assignment from platform_driver
  drm/rockchip: use drm_gem_mmap helpers
  drm/rockchip: only call drm_fb_helper_hotplug_event if fb_helper present
  drm/rockchip: Add BGR formats to VOP

9 years agoMerge tag 'imx-drm-fixes-2015-07-14' of git://git.pengutronix.de/git/pza/linux into...
Dave Airlie [Fri, 17 Jul 2015 00:24:34 +0000 (10:24 +1000)]
Merge tag 'imx-drm-fixes-2015-07-14' of git://git.pengutronix.de/git/pza/linux into drm-fixes

imx-drm: fixes for parallel-display, imx-tve, and ipu-common

These patches fix the parallel-display driver to use the standard OF
graph bindings for connecting a drm_panel via device tree instead of
an undocumented, driver specific device tree property, take care to
disable all IPU interrupts before setting up the irq chip to fix a
kexec lockup, and fix VGA output on i.MX53-QSB boards by correcting
the media bus format set by the imx-tve driver.

* tag 'imx-drm-fixes-2015-07-14' of git://git.pengutronix.de/git/pza/linux:
  drm/imx: tve: fix media bus format for VGA output
  GPU: ipu: fix lockup caused by pending chained interrupts
  drm/imx: parallel-display: fix drm_panel support

9 years agoMerge branch 'drm-armada-fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm into...
Dave Airlie [Fri, 17 Jul 2015 00:06:04 +0000 (10:06 +1000)]
Merge branch 'drm-armada-fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm into drm-fixes

A range of fixes for the Armada DRM driver:
- A missing wakeup could result in overlay frames being delayed, causing
  video playback to hiccup.
- Avoid unmapping a dma-buf attachment which was never mapped
- Fix the overlay when partially off the screen by switching to the
  drm_plane_helper_check_update() helper and using the calculated
  coordinates to set the start address.
- Remove an incorrect assignment to crtc->mode - which should be the
  unadjusted mode.
- Fix a missing call to drm_plane_cleanup() in the overlay code.

* 'drm-armada-fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
  drm/armada: avoid saving the adjusted mode to crtc->mode
  drm/armada: fix overlay when partially off-screen
  drm/armada: convert overlay to use drm_plane_helper_check_update()
  drm/armada: fix gem object free after failed prime import
  drm/armada: fix incorrect overlay plane cleanup
  drm/armada: fix missing overlay wake-up

9 years agoMerge tag 'drm-intel-fixes-2015-07-15' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Thu, 16 Jul 2015 23:57:16 +0000 (09:57 +1000)]
Merge tag 'drm-intel-fixes-2015-07-15' of git://anongit.freedesktop.org/drm-intel into drm-fixes

Next batch of i915 fixes. Note that the compat32 patch here needs the drm
core one to be actually useful, I'll send you that one with a separate
drm-fixes pull request. One revert because a fix in -rc2 did break
existing userspace.

* tag 'drm-intel-fixes-2015-07-15' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: Do not call intel_crtc_disable if the crtc is already disabled.
  Revert "drm/i915: Declare the swizzling unknown for L-shaped configurations"
  drm/i915: Forward all core DRM ioctls to core compat handling
  drm/i915: fix oops in primary_check_plane
  drm/i915: remove unused has_dma_mapping flag
  drm/i915: Fix missing return warning for !CONFIG_DEBUGFS
  drm/i915: avoid leaking DMA mappings
  drm/i915: Snapshot seqno of most recently submitted request.
  drm/i915: Store device pointer in contexts for late tracepoint usafe

9 years agoMerge tag 'topic/drm-fixes-2015-07-16' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Thu, 16 Jul 2015 23:56:00 +0000 (09:56 +1000)]
Merge tag 'topic/drm-fixes-2015-07-16' of git://anongit.freedesktop.org/drm-intel into drm-fixes

Ok next attempt at drm-fixes pull. Big thing really is just the compat32
one for addfb2.1.

* tag 'topic/drm-fixes-2015-07-16' of git://anongit.freedesktop.org/drm-intel:
  drm: Provide compat ioctl for addfb2.1
  Documentation: drm: Fix tablulation in KMS properties table
  drm: add a check for x/y in drm_mode_setcrtc
  drm/rockchip: use drm_gem_mmap helpers

9 years agoMerge branch 'drm-fixes-4.2' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Thu, 16 Jul 2015 23:39:23 +0000 (09:39 +1000)]
Merge branch 'drm-fixes-4.2' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

More radeon and amdgpu fixes for 4.2.  Mostly amdgpu bug fixes.

* 'drm-fixes-4.2' of git://people.freedesktop.org/~agd5f/linux:
  drm/amdgpu/dce8: Re-set VBLANK interrupt state when enabling a CRTC
  drm/radeon/ci: silence a harmless PCC warning
  drm/amdgpu/cz: silence some dpm debug output
  drm/amdgpu/cz: store the forced dpm level
  drm/amdgpu/cz: unforce dpm levels before forcing to low/high
  drm/amdgpu: remove bogus check in gfx8 rb setup
  drm/amdgpu: set proper index/data pair for smc regs on CZ (v2)
  drm/amdgpu: disable the IP module if early_init returns -ENOENT (v2)
  drm/amdgpu: stop context leak in the error path
  drm/amdgpu: validate the context id in the dependencies
  drm/radeon: fix user ptr race condition
  drm/radeon: Don't flush the GART TLB if rdev->gart.ptr == NULL
  drm/radeon: add a dpm quirk for Sapphire Radeon R9 270X 2GB GDDR5

9 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Thu, 16 Jul 2015 23:38:08 +0000 (16:38 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "A collection of fixes from the last few weeks that should go into the
  current series.  This contains:

   - Various fixes for the per-blkcg policy data, fixing regressions
     since 4.1.  From Arianna and Tejun

   - Code cleanup for bcache closure macros from me.  Really just
     flushing this out, it's been sitting in another branch for months

   - FIELD_SIZEOF cleanup from Maninder Singh

   - bio integrity oops fix from Mike

   - Timeout regression fix for blk-mq from Ming Lei"

* 'for-linus' of git://git.kernel.dk/linux-block:
  blk-mq: set default timeout as 30 seconds
  NVMe: Reread partitions on metadata formats
  bcache: don't embed 'return' statements in closure macros
  blkcg: fix blkcg_policy_data allocation bug
  blkcg: implement all_blkcgs list
  blkcg: blkcg_css_alloc() should grab blkcg_pol_mutex while iterating blkcg_policy[]
  blkcg: allow blkcg_pol_mutex to be grabbed from cgroup [file] methods
  block/blk-cgroup.c: free per-blkcg data when freeing the blkcg
  block: use FIELD_SIZEOF to calculate size of a field
  bio integrity: do not assume bio_integrity_pool exists if bioset exists

9 years agoMerge tag 'jfs-4.2' of git://github.com/kleikamp/linux-shaggy
Linus Torvalds [Thu, 16 Jul 2015 23:28:28 +0000 (16:28 -0700)]
Merge tag 'jfs-4.2' of git://github.com/kleikamp/linux-shaggy

Pull jfs fixes from David Kleikamp:
 "A couple trivial fixes and an error path fix"

* tag 'jfs-4.2' of git://github.com/kleikamp/linux-shaggy:
  jfs: clean up jfs_rename and fix out of order unlock
  jfs: fix indentation on if statement
  jfs: removed a prohibited space after opening parenthesis

9 years agoARM: keystone: dts: rename pcie nodes to help override status
Murali Karicheri [Thu, 16 Jul 2015 22:13:03 +0000 (18:13 -0400)]
ARM: keystone: dts: rename pcie nodes to help override status

Now that PCIe DT binding is disabled in SoC specific DTS,
we need a way to override it in a board specific DTS. So
rename the PCIe nodes accordingly.

Signed-off-by: Murali Karicheri <[email protected]>
Signed-off-by: Santosh Shilimkar <[email protected]>
9 years agoARM: keystone: dts: fix dt bindings for PCIe
Murali Karicheri [Thu, 16 Jul 2015 22:12:57 +0000 (18:12 -0400)]
ARM: keystone: dts: fix dt bindings for PCIe

Currently PCIe DT bindings are broken. PCIe driver can't function
without having a SerDes driver that provide the phy configuration.
On K2E EVM, this causes problem since the EVM has Marvell SATA
controller present and with default values in the SerDes register,
it seems to pass the PCIe link check, but causes issues since
the configuration is not correct. The manifestation is that when
EVM is booted with NFS rootfs, the boot hangs. We shouldn't enable
PCIe on this EVM since to work, SerDes driver has to be present as
well. So by default, the PCIe DT binding should be disabled in SoC
specific DTS. It can be enabled in the board specific DTS when the
SerDes device driver is also present.

So fix the status of PCIe DT bindings in the SoC specific DTS to
"disabled". To enable PCIe, the status should be set to "ok" in
the EVM DTS file when SerDes driver support becomes available in
the upstream tree.

Signed-off-by: Murali Karicheri <[email protected]>
Signed-off-by: Santosh Shilimkar <[email protected]>
9 years agoMerge branches 'pm-cpuidle', 'pm-cpufreq' and 'acpi-resources'
Rafael J. Wysocki [Thu, 16 Jul 2015 21:47:19 +0000 (23:47 +0200)]
Merge branches 'pm-cpuidle', 'pm-cpufreq' and 'acpi-resources'

* pm-cpuidle:
  suspend-to-idle: Prevent RCU from complaining about tick_freeze()

* pm-cpufreq:
  cpufreq: Allow freq_table to be obtained for offline CPUs
  cpufreq: Initialize the governor again while restoring policy

* acpi-resources:
  ACPI / PCI: Fix regressions caused by resource_size_t overflow with 32-bit kernel

9 years agoARM: pxa: fix dm9000 platform data regression
Robert Jarzmik [Sat, 11 Jul 2015 19:33:06 +0000 (21:33 +0200)]
ARM: pxa: fix dm9000 platform data regression

Since dm9000 driver added support for a vcc regulator, platform data
based platforms have their ethernet broken, as the regulator claiming
returns -EPROBE_DEFER and prevents dm9000 loading.

This patch fixes this for all pxa boards using dm9000, by using the
specific regulator_has_full_constraints() function.

This was discovered and tested on the cm-x300 board.

Fixes: 7994fe55a4a2 ("dm9000: Add regulator and reset support to dm9000")
Signed-off-by: Robert Jarzmik <[email protected]>
Acked-by: Igor Grinberg <[email protected]>
9 years agoregulator: core: Fix memory leak in regulator_resolve_supply()
Javier Martinez Canillas [Wed, 15 Jul 2015 14:10:29 +0000 (16:10 +0200)]
regulator: core: Fix memory leak in regulator_resolve_supply()

The regulator_resolve_supply() function calls set_supply() which in turn
calls create_regulator() to allocate a supply regulator.

If an error occurs after set_supply() succeeded, the allocated regulator
has to be freed before propagating the error code.

Signed-off-by: Javier Martinez Canillas <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
9 years agoregulator: core: Increase refcount for regulator supply's module
Javier Martinez Canillas [Wed, 15 Jul 2015 14:10:28 +0000 (16:10 +0200)]
regulator: core: Increase refcount for regulator supply's module

When a regulator is unregistered with regulator_unregister(), a call to
regulator_put() is made for its input supply if there is one. This does
a module_put() to decrement the refcount of the module that owns the
supply but there isn't a corresponding try_module_get() in set_supply()
to make the calls balanced.

Signed-off-by: Javier Martinez Canillas <[email protected]>
Signed-off-by: Mark Brown <[email protected]>
9 years agoInput: elantech - force resolution of 31 u/mm
Peter Hutterer [Thu, 16 Jul 2015 17:38:15 +0000 (10:38 -0700)]
Input: elantech - force resolution of 31 u/mm

All Elantech touchpads pre-v4 with dynamic resolution queries have a fixed
resolution of 800dpi -> 31.49 units/mm. Set this statically, so userspace
does not have to guess.

Signed-off-by: Peter Hutterer <[email protected]>
Reviewed-by: Hans de Goede <[email protected]>
Reviewed-by: David Herrmann <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
9 years agoInput: zforce - don't overwrite the stack
Oleksij Rempel [Mon, 13 Jul 2015 16:54:42 +0000 (09:54 -0700)]
Input: zforce - don't overwrite the stack

If we get a corrupted packet with PAYLOAD_LENGTH > FRAME_MAXSIZE, we
will silently overwrite the stack.

Cc: [email protected]
Signed-off-by: Oleksij Rempel <[email protected]>
Signed-off-by: Dirk Behme <[email protected]>
Signed-off-by: Dmitry Torokhov <[email protected]>
9 years agodrm/amdgpu/dce8: Re-set VBLANK interrupt state when enabling a CRTC
Michel Dänzer [Thu, 9 Jul 2015 09:24:24 +0000 (18:24 +0900)]
drm/amdgpu/dce8: Re-set VBLANK interrupt state when enabling a CRTC

Something (ATOM BIOS?) seems to be clobbering the LB_INTERRUPT_MASK
register while the CRTC is off, which caused e.g. glxgears or
gnome-shell to hang after a modeset.

Signed-off-by: Michel Dänzer <[email protected]>
Reviewed-by: Jammy Zhou <[email protected]>
9 years agodrm/radeon/ci: silence a harmless PCC warning
Alex Deucher [Thu, 16 Jul 2015 14:17:09 +0000 (10:17 -0400)]
drm/radeon/ci: silence a harmless PCC warning

This has been a source of confusion.  Make it debug only.

Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]
9 years agodrm/amdgpu/cz: silence some dpm debug output
Alex Deucher [Tue, 14 Jul 2015 21:37:48 +0000 (17:37 -0400)]
drm/amdgpu/cz: silence some dpm debug output

Reviewed-by: Jammy Zhou <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
9 years agodrm/amdgpu/cz: store the forced dpm level
Alex Deucher [Tue, 14 Jul 2015 16:29:00 +0000 (12:29 -0400)]
drm/amdgpu/cz: store the forced dpm level

So the selected forced level shows up properly in sysfs.

Reviewed-by: Jammy Zhou <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
9 years agodrm/amdgpu/cz: unforce dpm levels before forcing to low/high
Alex Deucher [Tue, 14 Jul 2015 16:26:41 +0000 (12:26 -0400)]
drm/amdgpu/cz: unforce dpm levels before forcing to low/high

This is necessary to properly reset the min/max limits before
clamping them otherwise we may get improper clamping depending
on what what was the last forced level.

Reviewed-by: Jammy Zhou <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
9 years agodrm/amdgpu: remove bogus check in gfx8 rb setup
Alex Deucher [Fri, 10 Jul 2015 21:05:31 +0000 (17:05 -0400)]
drm/amdgpu: remove bogus check in gfx8 rb setup

Always respect the harvest configuration as is.

Acked-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
9 years agodrm/amdgpu: set proper index/data pair for smc regs on CZ (v2)
Alex Deucher [Fri, 10 Jul 2015 20:21:10 +0000 (16:21 -0400)]
drm/amdgpu: set proper index/data pair for smc regs on CZ (v2)

v2: squash in later fix

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
9 years agodrm/amdgpu: disable the IP module if early_init returns -ENOENT (v2)
Alex Deucher [Fri, 10 Jul 2015 17:59:44 +0000 (13:59 -0400)]
drm/amdgpu: disable the IP module if early_init returns -ENOENT (v2)

If we run into a part with a harvest configuration where the entire
IP block is unusable, just disable the IP block.

v2: fix logic as noted by Christian

Reviewed-by: Christian König <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
9 years agodrm/amdgpu: stop context leak in the error path
Christian König [Tue, 7 Jul 2015 15:27:03 +0000 (17:27 +0200)]
drm/amdgpu: stop context leak in the error path

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Jammy Zhou <[email protected]>
9 years agodrm/amdgpu: validate the context id in the dependencies
Christian König [Mon, 6 Jul 2015 17:42:10 +0000 (19:42 +0200)]
drm/amdgpu: validate the context id in the dependencies

Just to make sure userspace don't send nonsense to the kernel.

Signed-off-by: Christian König <[email protected]>
Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Jammy Zhou <[email protected]>
9 years agodrm/radeon: fix user ptr race condition
Christian König [Tue, 14 Jul 2015 13:58:30 +0000 (15:58 +0200)]
drm/radeon: fix user ptr race condition

Port of amdgpu patch 9298e52f8b51d1e4acd68f502832f3a97f8cf892.

Signed-off-by: Christian König <[email protected]>
CC: [email protected]
Signed-off-by: Alex Deucher <[email protected]>
9 years agodrm/radeon: Don't flush the GART TLB if rdev->gart.ptr == NULL
Michel Dänzer [Fri, 3 Jul 2015 01:02:27 +0000 (10:02 +0900)]
drm/radeon: Don't flush the GART TLB if rdev->gart.ptr == NULL

This can be the case when the GPU is powered off, e.g. via vgaswitcheroo
or runpm. When the GPU is powered up again, radeon_gart_table_vram_pin
flushes the TLB after setting rdev->gart.ptr to non-NULL.

Fixes panic on powering off R7xx GPUs.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=61529
Reviewed-by: Christian König <[email protected]>
Signed-off-by: Michel Dänzer <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]
9 years agodrm/radeon: add a dpm quirk for Sapphire Radeon R9 270X 2GB GDDR5
Alex Deucher [Fri, 10 Jul 2015 01:08:17 +0000 (21:08 -0400)]
drm/radeon: add a dpm quirk for Sapphire Radeon R9 270X 2GB GDDR5

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=76490

Signed-off-by: Alex Deucher <[email protected]>
Cc: [email protected]
9 years agolibata: Do not blacklist M510DC
Martin K. Petersen [Thu, 16 Jul 2015 01:03:23 +0000 (21:03 -0400)]
libata: Do not blacklist M510DC

A new Micron drive was just announced, once again recycling the first
part of the model string. Add an underscore to the M510/M550 pattern to
avoid picking up the new DC drive.

Signed-off-by: Martin K. Petersen <[email protected]>
Cc: <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
9 years agoblk-mq: set default timeout as 30 seconds
Ming Lei [Thu, 16 Jul 2015 11:53:22 +0000 (19:53 +0800)]
blk-mq: set default timeout as 30 seconds

It is reasonable to set default timeout of request as 30 seconds instead of
30000 ticks, which may be 300 seconds if HZ is 100, for example, some arm64
based systems may choose 100 HZ.

Signed-off-by: Ming Lei <[email protected]>
Fixes: c76cbbcf4044 ("blk-mq: put blk_queue_rq_timeout together in blk_mq_init_queue()"
Signed-off-by: Jens Axboe <[email protected]>
9 years agodm cache: display 'needs_check' in status if it is set
Mike Snitzer [Wed, 15 Jul 2015 15:42:59 +0000 (11:42 -0400)]
dm cache: display 'needs_check' in status if it is set

There is currently no way to see that the needs_check flag has been set
in the metadata.  Display 'needs_check' in the cache status if it is set
in the cache metadata.

Also, update cache documentation.

Signed-off-by: Mike Snitzer <[email protected]>
9 years agodm thin: display 'needs_check' in status if it is set
Mike Snitzer [Wed, 15 Jul 2015 15:40:24 +0000 (11:40 -0400)]
dm thin: display 'needs_check' in status if it is set

There is currently no way to see that the needs_check flag has been set
in the metadata.  Display 'needs_check' in the thin-pool status if it is
set in the thinp metadata.

Also, update thinp documentation.

Signed-off-by: Mike Snitzer <[email protected]>
9 years agodm thin: stay in out-of-data-space mode once no_space_timeout expires
Mike Snitzer [Wed, 15 Jul 2015 20:52:04 +0000 (16:52 -0400)]
dm thin: stay in out-of-data-space mode once no_space_timeout expires

This fixes an issue where running out of data space would cause the
thin-pool's metadata to become read-only.  There was no reason to make
metadata read-only -- calling set_pool_mode() with PM_READ_ONLY was a
misguided way to error all queued and future write IOs.  We can
accomplish the same by degrading from PM_OUT_OF_DATA_SPACE to
PM_OUT_OF_DATA_SPACE with error_if_no_space enabled.

Otherwise, the use of PM_READ_ONLY could cause a race where commit() was
started before the PM_READ_ONLY transition but dm_pool_commit_metadata()
would go on to fail because the block manager had transitioned to
read-only.  The return of -EPERM from dm_pool_commit_metadata(), due to
attempting to commit while in read-only mode, caused the thin-pool to
set 'needs_check' because a metadata_operation_failed().  This needless
cascade of failures makes life for users more difficult than needed.

Reported-by: Vivek Goyal <[email protected]>
Signed-off-by: Mike Snitzer <[email protected]>
9 years agoscsi: fix host max depth checking for the 'queue_depth' sysfs interface
Jens Axboe [Mon, 13 Jul 2015 14:24:39 +0000 (08:24 -0600)]
scsi: fix host max depth checking for the 'queue_depth' sysfs interface

Commit 1e6f2416044c0 changed the scsi sysfs 'queue_depth' code to
rejects depths higher than the scsi host template setting. But lots
of hosts set this to 1, and update the settings in the scsi host
when the controller/devices probing happens.

This breaks (at least) mpt2sas and mpt3sas runtime setting of queue
depth, returning EINVAL for all settings but '1'. And once it's set to
1, there's no way to go back up.

Cc: [email protected]
Fixes: 1e6f2416044c0 "scsi: don't allow setting of queue_depth bigger than can_queue"
Signed-off-by: Jens Axboe <[email protected]>
Reviewed-by: Martin K. Petersen <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
Signed-off-by: James Bottomley <[email protected]>
9 years agogpio: pca953x: fix nested irqs rescheduling
Grygorii Strashko [Tue, 7 Jul 2015 14:34:49 +0000 (17:34 +0300)]
gpio: pca953x: fix nested irqs rescheduling

pca953x interrupt controller functionality is implemented using
nested threaded IRQs which require parent_irq to be configured
properly otherwise below warning can be seen if IRQ core
will try re-schedule nested IRQ:

------------[ cut here ]------------
WARNING: CPU: 1 PID: 12 at kernel/irq/manage.c:696 irq_nested_primary_handler+0x30/0x38()
Primary handler called for nested irq 301
Modules linked in: uinput ipv6 smsc95xx usbnet mii imx2_wdt etnaviv(C) matrix_keypad matrix_keymap ar1021_i2c
CPU: 1 PID: 12 Comm: ksoftirqd/1 Tainted: G        WC    4.1.1 #9
Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
Backtrace:
[<c0013298>] (dump_backtrace) from [<c0013488>] (show_stack+0x20/0x24)
[<c0013468>] (show_stack) from [<c05743c4>] (dump_stack+0x70/0xc0)
[<c0574354>] (dump_stack) from [<c002b7b8>] (warn_slowpath_common+0x88/0xc0)
[<c002b730>] (warn_slowpath_common) from [<c002b8ac>] (warn_slowpath_fmt+0x40/0x48)
[<c002b870>] (warn_slowpath_fmt) from [<c0075798>] (irq_nested_primary_handler+0x30/0x38)
[<c0075768>] (irq_nested_primary_handler) from [<c0075200>] (handle_irq_event_percpu+0x70/0x2d0)
[<c0075190>] (handle_irq_event_percpu) from [<c00754ac>] (handle_irq_event+0x4c/0x6c)
[<c0075460>] (handle_irq_event) from [<c0078204>] (handle_simple_irq+0xa4/0xc8)
[<c0078160>] (handle_simple_irq) from [<c0077cd4>] (resend_irqs+0x50/0x7c)
[<c0077c84>] (resend_irqs) from [<c002f99c>] (tasklet_action+0x94/0x140)
[<c002f908>] (tasklet_action) from [<c002eea8>] (__do_softirq+0xa0/0x3c8)
[<c002ee08>] (__do_softirq) from [<c002f208>] (run_ksoftirqd+0x38/0x54)
[<c002f1d0>] (run_ksoftirqd) from [<c004b1e4>] (smpboot_thread_fn+0x1f8/0x2f0)
[<c004afec>] (smpboot_thread_fn) from [<c0047744>] (kthread+0xe8/0x104)
[<c004765c>] (kthread) from [<c000fac8>] (ret_from_fork+0x14/0x2c)
---[ end trace 96052cda48865769 ]---

The issue was reported and described in details by Lothar Waßmann and
Christian Gmeiner in https://lkml.org/lkml/2014/9/9/123.

Fix it by adding missed call of gpiochip_set_chained_irqchip()
so GPIO IRQ chip helpers will set parent_irq for nested IRQs
properly.

Reported-by: Lothar Waßmann <[email protected]>
Tested-by: Christian Gmeiner <[email protected]>
Signed-off-by: Grygorii Strashko <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agost: null pointer dereference panic caused by use after kref_put by st_open
Seymour, Shane M [Thu, 2 Jul 2015 12:01:10 +0000 (12:01 +0000)]
st: null pointer dereference panic caused by use after kref_put by st_open

Two SLES11 SP3 servers encountered similar crashes simultaneously
following some kind of SAN/tape target issue:

...
qla2xxx [0000:81:00.0]-801c:3: Abort command issued nexus=3:0:2 --  1 2002.
qla2xxx [0000:81:00.0]-801c:3: Abort command issued nexus=3:0:2 --  1 2002.
qla2xxx [0000:81:00.0]-8009:3: DEVICE RESET ISSUED nexus=3:0:2 cmd=ffff882f89c2c7c0.
qla2xxx [0000:81:00.0]-800c:3: do_reset failed for cmd=ffff882f89c2c7c0.
qla2xxx [0000:81:00.0]-800f:3: DEVICE RESET FAILED: Task management failed nexus=3:0:2 cmd=ffff882f89c2c7c0.
qla2xxx [0000:81:00.0]-8009:3: TARGET RESET ISSUED nexus=3:0:2 cmd=ffff882f89c2c7c0.
qla2xxx [0000:81:00.0]-800c:3: do_reset failed for cmd=ffff882f89c2c7c0.
qla2xxx [0000:81:00.0]-800f:3: TARGET RESET FAILED: Task management failed nexus=3:0:2 cmd=ffff882f89c2c7c0.
qla2xxx [0000:81:00.0]-8012:3: BUS RESET ISSUED nexus=3:0:2.
qla2xxx [0000:81:00.0]-802b:3: BUS RESET SUCCEEDED nexus=3:0:2.
qla2xxx [0000:81:00.0]-505f:3: Link is operational (8 Gbps).
qla2xxx [0000:81:00.0]-8018:3: ADAPTER RESET ISSUED nexus=3:0:2.
qla2xxx [0000:81:00.0]-00af:3: Performing ISP error recovery - ha=ffff88bf04d18000.
 rport-3:0-0: blocked FC remote port time out: removing target and saving binding
qla2xxx [0000:81:00.0]-505f:3: Link is operational (8 Gbps).
qla2xxx [0000:81:00.0]-8017:3: ADAPTER RESET SUCCEEDED nexus=3:0:2.
 rport-2:0-0: blocked FC remote port time out: removing target and saving binding
sg_rq_end_io: device detached
BUG: unable to handle kernel NULL pointer dereference at 00000000000002a8
IP: [<ffffffff8133b268>] __pm_runtime_idle+0x28/0x90
PGD 7e6586f067 PUD 7e5af06067 PMD 0 [1739975.390354] Oops: 0002 [#1] SMP
CPU 0
...
Supported: No, Proprietary modules are loaded [1739975.390463]
Pid: 27965, comm: ABCD Tainted: PF           X 3.0.101-0.29-default #1 HP ProLiant DL580 Gen8
RIP: 0010:[<ffffffff8133b268>]  [<ffffffff8133b268>] __pm_runtime_idle+0x28/0x90
RSP: 0018:ffff8839dc1e7c68  EFLAGS: 00010202
RAX: 0000000000000000 RBX: ffff883f0592fc00 RCX: 0000000000000090
RDX: 0000000000000000 RSI: 0000000000000004 RDI: 0000000000000138
RBP: 0000000000000138 R08: 0000000000000010 R09: ffffffff81bd39d0
R10: 00000000000009c0 R11: ffffffff81025790 R12: 0000000000000001
R13: ffff883022212b80 R14: 0000000000000004 R15: ffff883022212b80
FS:  00007f8e54560720(0000) GS:ffff88407f800000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
CR2: 00000000000002a8 CR3: 0000007e6ced6000 CR4: 00000000001407f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Process ABCD (pid: 27965, threadinfo ffff8839dc1e6000, task ffff883592e0c640)
Stack:
 ffff883f0592fc00 00000000fffffffa 0000000000000001 ffff883022212b80
 ffff883eff772400 ffffffffa03fa309 0000000000000000 0000000000000000
 ffffffffa04003a0 ffff883f063196c0 ffff887f0379a930 ffffffff8115ea1e
Call Trace:
 [<ffffffffa03fa309>] st_open+0x129/0x240 [st]
 [<ffffffff8115ea1e>] chrdev_open+0x13e/0x200
 [<ffffffff811588a8>] __dentry_open+0x198/0x310
 [<ffffffff81167d74>] do_last+0x1f4/0x800
 [<ffffffff81168fe9>] path_openat+0xd9/0x420
 [<ffffffff8116946c>] do_filp_open+0x4c/0xc0
 [<ffffffff8115a00f>] do_sys_open+0x17f/0x250
 [<ffffffff81468d92>] system_call_fastpath+0x16/0x1b
 [<00007f8e4f617fd0>] 0x7f8e4f617fcf
Code: eb d3 90 48 83 ec 28 40 f6 c6 04 48 89 6c 24 08 4c 89 74 24 20 48 89 fd 48 89 1c 24 4c 89 64 24 10 41 89 f6 4c 89 6c 24 18 74 11 <f0> ff 8f 70 01 00 00 0f 94 c0 45 31 ed 84 c0 74 2b 4c 8d a5 a0
RIP  [<ffffffff8133b268>] __pm_runtime_idle+0x28/0x90
 RSP <ffff8839dc1e7c68>
CR2: 00000000000002a8

Analysis reveals the cause of the crash to be due to STp->device
being NULL. The pointer was NULLed via scsi_tape_put(STp) when it
calls scsi_tape_release(). In st_open() we jump to err_out after
scsi_block_when_processing_errors() completes and returns the
device as offline (sdev_state was SDEV_DEL):

1180 /* Open the device. Needs to take the BKL only because of incrementing the SCSI host
1181    module count. */
1182 static int st_open(struct inode *inode, struct file *filp)
1183 {
1184         int i, retval = (-EIO);
1185         int resumed = 0;
1186         struct scsi_tape *STp;
1187         struct st_partstat *STps;
1188         int dev = TAPE_NR(inode);
1189         char *name;
...
1217         if (scsi_autopm_get_device(STp->device) < 0) {
1218                 retval = -EIO;
1219                 goto err_out;
1220         }
1221         resumed = 1;
1222         if (!scsi_block_when_processing_errors(STp->device)) {
1223                 retval = (-ENXIO);
1224                 goto err_out;
1225         }
...
1264  err_out:
1265         normalize_buffer(STp->buffer);
1266         spin_lock(&st_use_lock);
1267         STp->in_use = 0;
1268         spin_unlock(&st_use_lock);
1269         scsi_tape_put(STp); <-- STp->device = 0 after this
1270         if (resumed)
1271                 scsi_autopm_put_device(STp->device);
1272         return retval;

The ref count for the struct scsi_tape had already been reduced
to 1 when the .remove method of the st module had been called.
The kref_put() in scsi_tape_put() caused scsi_tape_release()
to be called:

0266 static void scsi_tape_put(struct scsi_tape *STp)
0267 {
0268         struct scsi_device *sdev = STp->device;
0269
0270         mutex_lock(&st_ref_mutex);
0271         kref_put(&STp->kref, scsi_tape_release); <-- calls this
0272         scsi_device_put(sdev);
0273         mutex_unlock(&st_ref_mutex);
0274 }

In scsi_tape_release() the struct scsi_device in the struct
scsi_tape gets set to NULL:

4273 static void scsi_tape_release(struct kref *kref)
4274 {
4275         struct scsi_tape *tpnt = to_scsi_tape(kref);
4276         struct gendisk *disk = tpnt->disk;
4277
4278         tpnt->device = NULL; <<<---- where the dev is nulled
4279
4280         if (tpnt->buffer) {
4281                 normalize_buffer(tpnt->buffer);
4282                 kfree(tpnt->buffer->reserved_pages);
4283                 kfree(tpnt->buffer);
4284         }
4285
4286         disk->private_data = NULL;
4287         put_disk(disk);
4288         kfree(tpnt);
4289         return;
4290 }

Although the problem was reported on SLES11.3 the problem appears
in linux-next as well.

The crash is fixed by reordering the code so we no longer access
the struct scsi_tape after the kref_put() is done on it in st_open().

Signed-off-by: Shane Seymour <[email protected]>
Signed-off-by: Darren Lavender <[email protected]>
Reviewed-by: Johannes Thumshirn <[email protected]>
Acked-by: Kai Mäkisara <[email protected]>
Cc: [email protected]
Signed-off-by: James Bottomley <[email protected]>
9 years agogpio: omap: prevent module from being unloaded while in use
Grygorii Strashko [Thu, 25 Jun 2015 15:13:33 +0000 (18:13 +0300)]
gpio: omap: prevent module from being unloaded while in use

OMAP GPIO driver allowed to be built as loadable module, but it
doesn't set owner field in GPIO chip structure. As result,
module_get/put() API is not working and it's possible to unload
OMAP driver while in use:

  omap_gpio 48051000.gpio: REMOVING GPIOCHIP WITH GPIOS STILL REQUESTED

Hence, add missing configuration.

Cc: Tony Lindgren <[email protected]>
Fixes: cac089f9026e ('gpio: omap: Allow building as a loadable module')
Signed-off-by: Grygorii Strashko <[email protected]>
Acked-by: Alexandre Courbot <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agogpio: max732x: Add missing dev reference to gpiochip
Marek Vasut [Tue, 30 Jun 2015 18:04:18 +0000 (20:04 +0200)]
gpio: max732x: Add missing dev reference to gpiochip

In case the gpiochip doesn't have the .dev field set, as is the case
in here, it is not possible to reference this device in DT as a GPIO
controller. A good example of this problem is that gpio-leds can not
be used when connected to this chip, the gpio-leds driver bails out
with -EPROBE_DEFER.

Fix this problem by setting the .dev field of the gpio_chip to the
parent i2c device.

Signed-off-by: Marek Vasut <[email protected]>
Cc: Alexandre Courbot <[email protected]>
Cc: Linus Walleij <[email protected]>
Cc: Mans Rullgard <[email protected]>
Cc: Olaf Mandel <[email protected]>
Cc: Semen Protsenko <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agogpio/xilinx: Use correct address when setting initial values.
Raphaël Teysseyre [Wed, 24 Jun 2015 07:19:45 +0000 (09:19 +0200)]
gpio/xilinx: Use correct address when setting initial values.

xgpio_save_regs() is used in this driver to setup the initial
values of the registers in the hardware.

The relevant registers at that time are:
0x0 -> channel 0 data (32 bits, one for each GPIO on this channel).
0x4 -> channel 0 tri, controls in/out status for each GPIO of this channel.
0x8 -> channel 1 data
0xC -> channel 1 tri

gpio-xilinx.c defines these:
XGPIO_DATA_OFFSET   (0x0)
XGPIO_TRI_OFFSET    (0x4)
XGPIO_CHANNEL_OFFSET 0x8

Before this patch, the "data" register value of channel 1 was written
at 0x4 intead of 0x8 (overwriting the channel 0 "tri" register),
and the "tri" register value for channel 1 was written at 0x8 instead of 0xC.

Signed-off-by: Raphaël Teysseyre <[email protected]>
Reviewed-by: Michal Simek <[email protected]>
Reviewed-by: Alexandre Courbot <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agogpio: zynq: Fix problem with unbalanced pm_runtime_enable
Michal Simek [Thu, 25 Jun 2015 08:29:19 +0000 (10:29 +0200)]
gpio: zynq: Fix problem with unbalanced pm_runtime_enable

Add missing pm_runtime_disabled to remove().

Error log:
root@zynqmp:~# modprobe gpio_zynq
root@zynqmp:~# lsmod
    Not tainted
gpio_zynq 7086 0 - Live 0xffffffbffc00a000
root@zynqmp:~# rmmod gpio_zynq
root@zynqmp:~# lsmod
    Not tainted
root@zynqmp:~# modprobe gpio_zynq
[  246.924438] zynq-gpio ff0a0000.gpio: Unbalanced pm_runtime_enable!
root@zynqmp:~# rmmod gpio_zynq
root@zynqmp:~# lsmod
    Not tainted

Signed-off-by: Michal Simek <[email protected]>
Reviewed-by: Alexandre Courbot <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agoALSA: hda/realtek: Enable HP amp and mute LED on HP Folio 9480m [v3]
Keith Packard [Wed, 15 Jul 2015 19:14:39 +0000 (12:14 -0700)]
ALSA: hda/realtek: Enable HP amp and mute LED on HP Folio 9480m [v3]

This laptop needs GPIO4 pulled high to enable the headphone amplifier,
and has a mute LED on GPIO3. I modelled the patch on the existing
GPIO4 code which pulls the line low for the same purpose; this time,
the HP amp line is pulled high.

v2: Disable the headphone amplifier when no headphone is connected.
    Don't disable power savings to preserve the LED state.

v3: Remove headset-specific hooks and code; this is just a headphone.

Signed-off-by: Keith Packard <[email protected]>
Signed-off-by: Takashi Iwai <[email protected]>
9 years agogpio: omap: add missed spin_unlock_irqrestore in omap_gpio_irq_type
Grygorii Strashko [Wed, 24 Jun 2015 14:54:17 +0000 (17:54 +0300)]
gpio: omap: add missed spin_unlock_irqrestore in omap_gpio_irq_type

Add missed spin_unlock_irqrestore in omap_gpio_irq_type when
omap_set_gpio_triggering() is failed.

It fixes static checker warning:

drivers/gpio/gpio-omap.c:523 omap_gpio_irq_type()
warn: inconsistent returns 'spin_lock:&bank->lock'.

This fixes commit:
1562e4618ded ('gpio: omap: fix error handling in omap_gpio_irq_type')

Reported-by: Javier Martinez Canillas <[email protected]>
Signed-off-by: Grygorii Strashko <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agocan: mcp251x: get regulators optionally
Stefan Agner [Mon, 18 May 2015 16:33:28 +0000 (18:33 +0200)]
can: mcp251x: get regulators optionally

The regulators power and transceiver are optional. If those are not
present, the pointer (or error pointer) is correctly handled by the
driver, hence we can use devm_regulator_get_optional safely, which
avoids regulators getting created.

Signed-off-by: Stefan Agner <[email protected]>
Signed-off-by: Marc Kleine-Budde <[email protected]>
9 years agocan: mcp251x: fix resume when device is down
Stefan Agner [Mon, 18 May 2015 16:33:27 +0000 (18:33 +0200)]
can: mcp251x: fix resume when device is down

If a valid power regulator or a dummy regulator is used (which
happens to be the case when no regulator is specified), restart_work
is queued no matter whether the device was running or not at suspend
time. Since work queues get initialized in the ndo_open callback,
resuming leads to a NULL pointer exception.

Reverse exactly the steps executed at suspend time:
- Enable the power regulator in any case
- Enable the transceiver regulator if the device was running, even in
  case we have a power regulator
- Queue restart_work only in case the device was running

Fixes: bf66f3736a94 ("can: mcp251x: Move to threaded interrupts instead of workqueues.")
Signed-off-by: Stefan Agner <[email protected]>
Cc: linux-stable <[email protected]>
Signed-off-by: Marc Kleine-Budde <[email protected]>
9 years agoMerge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetoot...
David S. Miller [Thu, 16 Jul 2015 04:59:23 +0000 (21:59 -0700)]
Merge branch 'for-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/bluetooth/bluetooth

Johan Hedberg says:

====================
pull request: bluetooth 2015-07-15

Here's a regression fix for Broadcom Bluetooth adapters found at least
in certain Apple laptops. The issue was introduced in 4.1 so there's the
appropriate "Cc: stable" entry for it.
====================

Signed-off-by: David S. Miller <[email protected]>
9 years agotc: act_bpf: fix memory leak
Alexei Starovoitov [Tue, 14 Jul 2015 19:15:19 +0000 (12:15 -0700)]
tc: act_bpf: fix memory leak

prog->bpf_ops is populated when act_bpf is used with classic BPF and
prog->bpf_name is optionally used with extended BPF.
Fix memory leak when act_bpf is released.

Fixes: d23b8ad8ab23 ("tc: add BPF based action")
Fixes: a8cb5f556b56 ("act_bpf: add initial eBPF support for actions")
Acked-by: Daniel Borkmann <[email protected]>
Signed-off-by: Alexei Starovoitov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agofq_codel: fix return value of fq_codel_drop()
WANG Cong [Tue, 14 Jul 2015 18:21:58 +0000 (11:21 -0700)]
fq_codel: fix return value of fq_codel_drop()

The ->drop() is supposed to return the number of bytes it dropped,
however fq_codel_drop() returns the index of the flow where it drops
a packet from.

Fix this by introducing a helper to wrap fq_codel_drop().

Cc: Eric Dumazet <[email protected]>
Signed-off-by: Cong Wang <[email protected]>
Signed-off-by: Cong Wang <[email protected]>
Acked-by: Eric Dumazet <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agonet_sched: fix a use-after-free in sfq
WANG Cong [Tue, 14 Jul 2015 18:21:57 +0000 (11:21 -0700)]
net_sched: fix a use-after-free in sfq

Fixes: 25331d6ce42b ("net: sched: implement qstat helper routines")
Cc: John Fastabend <[email protected]>
Signed-off-by: Cong Wang <[email protected]>
Signed-off-by: Cong Wang <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoMerge branch 'ipvlan'
David S. Miller [Thu, 16 Jul 2015 04:33:40 +0000 (21:33 -0700)]
Merge branch 'ipvlan'

Konstantin Khlebnikov says:

====================
ipvlan: cleanups and fixes

v1: http://comments.gmane.org/gmane.linux.network/363346
v2: http://comments.gmane.org/gmane.linux.network/369086

v3 has reduced set of patches from "ipvlan: fix ipv6 autoconfiguration".
Here just cleanups and patch which ignores ipv6 notifications from RA.
====================

Signed-off-by: David S. Miller <[email protected]>
9 years agoipvlan: ignore addresses from ipv6 autoconfiguration
Konstantin Khlebnikov [Tue, 14 Jul 2015 13:35:55 +0000 (16:35 +0300)]
ipvlan: ignore addresses from ipv6 autoconfiguration

Inet6addr notifier is atomic and runs in bh context without RTNL when
ipv6 receives router advertisement packet and performs autoconfiguration.

Proper fix still in discussion. Let's at least plug the bug.
v1: http://lkml.kernel.org/r/20150514135618.14062.1969.stgit@buzz
v2: http://lkml.kernel.org/r/20150703125840.24121.91556.stgit@buzz

Signed-off-by: Konstantin Khlebnikov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoipvlan: use rcu_deference_bh() in ipvlan_queue_xmit()
WANG Cong [Tue, 14 Jul 2015 13:35:54 +0000 (16:35 +0300)]
ipvlan: use rcu_deference_bh() in ipvlan_queue_xmit()

In tx path rcu_read_lock_bh() is held, so we need rcu_deference_bh().
This fixes the following warning:

 ===============================
 [ INFO: suspicious RCU usage. ]
 4.1.0-rc1+ #1007 Not tainted
 -------------------------------
 drivers/net/ipvlan/ipvlan.h:106 suspicious rcu_dereference_check() usage!

 other info that might help us debug this:

 rcu_scheduler_active = 1, debug_locks = 0
 1 lock held by dhclient/1076:
  #0:  (rcu_read_lock_bh){......}, at: [<ffffffff817e8d84>] rcu_lock_acquire+0x0/0x26

 stack backtrace:
 CPU: 2 PID: 1076 Comm: dhclient Not tainted 4.1.0-rc1+ #1007
 Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
  0000000000000001 ffff8800d381bac8 ffffffff81a4154f 000000003c1a3c19
  ffff8800d4d0a690 ffff8800d381baf8 ffffffff810b849f ffff880117d41148
  ffff880117d40000 ffff880117d40068 0000000000000156 ffff8800d381bb18
 Call Trace:
  [<ffffffff81a4154f>] dump_stack+0x4c/0x65
  [<ffffffff810b849f>] lockdep_rcu_suspicious+0x107/0x110
  [<ffffffff8165a522>] ipvlan_port_get_rcu+0x47/0x4e
  [<ffffffff8165ad14>] ipvlan_queue_xmit+0x35/0x450
  [<ffffffff817ea45d>] ? rcu_read_unlock+0x3e/0x5f
  [<ffffffff810a20bf>] ? local_clock+0x19/0x22
  [<ffffffff810b4781>] ? __lock_is_held+0x39/0x52
  [<ffffffff8165b64c>] ipvlan_start_xmit+0x1b/0x44
  [<ffffffff817edf7f>] dev_hard_start_xmit+0x2ae/0x467
  [<ffffffff817ee642>] __dev_queue_xmit+0x50a/0x60c
  [<ffffffff817ee7a7>] dev_queue_xmit_sk+0x13/0x15
  [<ffffffff81997596>] dev_queue_xmit+0x10/0x12
  [<ffffffff8199b41c>] packet_sendmsg+0xb6b/0xbdf
  [<ffffffff810b5ea7>] ? mark_lock+0x2e/0x226
  [<ffffffff810a1fcc>] ? sched_clock_cpu+0x9e/0xb7
  [<ffffffff817d56f9>] sock_sendmsg_nosec+0x12/0x1d
  [<ffffffff817d7257>] sock_sendmsg+0x29/0x2e
  [<ffffffff817d72cc>] sock_write_iter+0x70/0x91
  [<ffffffff81199563>] __vfs_write+0x7e/0xa7
  [<ffffffff811996bc>] vfs_write+0x92/0xe8
  [<ffffffff811997d7>] SyS_write+0x47/0x7e
  [<ffffffff81a4d517>] system_call_fastpath+0x12/0x6f

Fixes: 2ad7bf363841 ("ipvlan: Initial check-in of the IPVLAN driver.")
Cc: Mahesh Bandewar <[email protected]>
Signed-off-by: Cong Wang <[email protected]>
Acked-by: Mahesh Bandewar <[email protected]>
Acked-by: Konstantin Khlebnikov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoipvlan: unhash addresses without synchronize_rcu
Konstantin Khlebnikov [Tue, 14 Jul 2015 13:35:53 +0000 (16:35 +0300)]
ipvlan: unhash addresses without synchronize_rcu

All structures used in traffic forwarding are rcu-protected:
ipvl_addr, ipvl_dev and ipvl_port. Thus we can unhash addresses
without synchronization. We'll anyway hash it back into the same
bucket: in worst case lockless lookup will scan hash once again.

Signed-off-by: Konstantin Khlebnikov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoipvlan: plug memory leak in ipvlan_link_delete
Konstantin Khlebnikov [Tue, 14 Jul 2015 13:35:51 +0000 (16:35 +0300)]
ipvlan: plug memory leak in ipvlan_link_delete

Add missing kfree_rcu(addr, rcu);

Signed-off-by: Konstantin Khlebnikov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoipvlan: remove counters of ipv4 and ipv6 addresses
Konstantin Khlebnikov [Tue, 14 Jul 2015 13:35:50 +0000 (16:35 +0300)]
ipvlan: remove counters of ipv4 and ipv6 addresses

They are unused after commit f631c44bbe15 ("ipvlan: Always set broadcast bit in
multicast filter").

Signed-off-by: Konstantin Khlebnikov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris...
Linus Torvalds [Thu, 16 Jul 2015 01:38:24 +0000 (18:38 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security

Pull TPM bugfixes from James Morris.

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security:
  tpm, tpm_crb: fail when TPM2 ACPI table contents look corrupted
  tpm: Fix initialization of the cdev

9 years agoMerge tag 'linux-can-fixes-for-4.2-20150715' of git://git.kernel.org/pub/scm/linux...
David S. Miller [Thu, 16 Jul 2015 00:27:13 +0000 (17:27 -0700)]
Merge tag 'linux-can-fixes-for-4.2-20150715' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can

Marc Kleine-Budde says:

====================
pull-request: can 2015-07-15

this is a pull request of 12 patches by me.

This series fixes the use of the skb after netif_receive_skb() / netif_rx()
which exists in several drivers.
====================

Signed-off-by: David S. Miller <[email protected]>
9 years agoipv6: lock socket in ip6_datagram_connect()
Eric Dumazet [Tue, 14 Jul 2015 06:10:22 +0000 (08:10 +0200)]
ipv6: lock socket in ip6_datagram_connect()

ip6_datagram_connect() is doing a lot of socket changes without
socket being locked.

This looks wrong, at least for udp_lib_rehash() which could corrupt
lists because of concurrent udp_sk(sk)->udp_portaddr_hash accesses.

Signed-off-by: Eric Dumazet <[email protected]>
Acked-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoMerge branch 'isdn-gigaset-fixes'
David S. Miller [Thu, 16 Jul 2015 00:24:45 +0000 (17:24 -0700)]
Merge branch 'isdn-gigaset-fixes'

Tilman Schmidt says:

====================
Fix long-standing regression in ser_gigaset ISDN driver

This series fixes a serious regression in the Gigaset M101 driver
introduced in kernel release 3.10 and removes some unneeded code.

Please also queue up patch 1 of the series for inclusion in the
stable/longterm releases 3.10 and later.
====================

Signed-off-by: David S. Miller <[email protected]>
9 years agoisdn/gigaset: drop unused ldisc methods
Tilman Schmidt [Mon, 13 Jul 2015 22:37:13 +0000 (00:37 +0200)]
isdn/gigaset: drop unused ldisc methods

The line discipline read and write methods are optional so the dummy
methods in ser_gigaset are unnecessary and can be removed.

Signed-off-by: Tilman Schmidt <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoisdn/gigaset: reset tty->receive_room when attaching ser_gigaset
Tilman Schmidt [Mon, 13 Jul 2015 22:37:13 +0000 (00:37 +0200)]
isdn/gigaset: reset tty->receive_room when attaching ser_gigaset

Commit 79901317ce80 ("n_tty: Don't flush buffer when closing ldisc"),
first merged in kernel release 3.10, caused the following regression
in the Gigaset M101 driver:

Before that commit, when closing the N_TTY line discipline in
preparation to switching to N_GIGASET_M101, receive_room would be
reset to a non-zero value by the call to n_tty_flush_buffer() in
n_tty's close method. With the removal of that call, receive_room
might be left at zero, blocking data reception on the serial line.

The present patch fixes that regression by setting receive_room
to an appropriate value in the ldisc open method.

Fixes: 79901317ce80 ("n_tty: Don't flush buffer when closing ldisc")
Signed-off-by: Tilman Schmidt <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agofq_codel: fix a use-after-free
WANG Cong [Mon, 13 Jul 2015 19:30:07 +0000 (12:30 -0700)]
fq_codel: fix a use-after-free

Fixes: 25331d6ce42b ("net: sched: implement qstat helper routines")
Cc: John Fastabend <[email protected]>
Signed-off-by: Cong Wang <[email protected]>
Signed-off-by: Cong Wang <[email protected]>
Acked-by: Eric Dumazet <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agotcp: don't use F-RTO on non-recurring timeouts
Yuchung Cheng [Mon, 13 Jul 2015 19:10:20 +0000 (12:10 -0700)]
tcp: don't use F-RTO on non-recurring timeouts

Currently F-RTO may repeatedly send new data packets on non-recurring
timeouts in CA_Loss mode. This is a bug because F-RTO (RFC5682)
should only be used on either new recovery or recurring timeouts.

This exacerbates the recovery progress during frequent timeout &
repair, because we prioritize sending new data packets instead of
repairing the holes when the bandwidth is already scarce.

Fix it by correcting the test of a new recovery episode.

Signed-off-by: Yuchung Cheng <[email protected]>
Signed-off-by: Neal Cardwell <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agobridge: mdb: fix double add notification
Nikolay Aleksandrov [Mon, 13 Jul 2015 13:36:19 +0000 (06:36 -0700)]
bridge: mdb: fix double add notification

Since the mdb add/del code was introduced there have been 2 br_mdb_notify
calls when doing br_mdb_add() resulting in 2 notifications on each add.

Example:
 Command: bridge mdb add dev br0 port eth1 grp 239.0.0.1 permanent
 Before patch:
 root@debian:~# bridge monitor all
 [MDB]dev br0 port eth1 grp 239.0.0.1 permanent
 [MDB]dev br0 port eth1 grp 239.0.0.1 permanent

 After patch:
 root@debian:~# bridge monitor all
 [MDB]dev br0 port eth1 grp 239.0.0.1 permanent

Signed-off-by: Nikolay Aleksandrov <[email protected]>
Fixes: cfd567543590 ("bridge: add support of adding and deleting mdb entries")
Signed-off-by: David S. Miller <[email protected]>
9 years agobridge: multicast: treat igmpv3 report with INCLUDE and no sources as a leave
Satish Ashok [Mon, 13 Jul 2015 12:28:37 +0000 (05:28 -0700)]
bridge: multicast: treat igmpv3 report with INCLUDE and no sources as a leave

A report with INCLUDE/Change_to_include and empty source list should be
treated as a leave, specified by RFC 3376, section 3.1:
"If the requested filter mode is INCLUDE *and* the requested source
 list is empty, then the entry corresponding to the requested
 interface and multicast address is deleted if present.  If no such
 entry is present, the request is ignored."

Signed-off-by: Satish Ashok <[email protected]>
Signed-off-by: Nikolay Aleksandrov <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
Linus Torvalds [Thu, 16 Jul 2015 00:03:03 +0000 (17:03 -0700)]
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma

Pull rdma fixes from Doug Ledford:
 "Mainly fix-ups for the various 4.2 items"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma: (24 commits)
  IB/core: Destroy ocrdma_dev_id IDR on module exit
  IB/core: Destroy multcast_idr on module exit
  IB/mlx4: Optimize do_slave_init
  IB/mlx4: Fix memory leak in do_slave_init
  IB/mlx4: Optimize freeing of items on error unwind
  IB/mlx4: Fix use of flow-counters for process_mad
  IB/ipath: Convert use of __constant_<foo> to <foo>
  IB/ipoib: Set MTU to max allowed by mode when mode changes
  IB/ipoib: Scatter-Gather support in connected mode
  IB/ucm: Fix bitmap wrap when devnum > IB_UCM_MAX_DEVICES
  IB/ipoib: Prevent lockdep warning in __ipoib_ib_dev_flush
  IB/ucma: Fix lockdep warning in ucma_lock_files
  rds: rds_ib_device.refcount overflow
  RDMA/nes: Fix for incorrect recording of the MAC address
  RDMA/nes: Fix for resolving the neigh
  RDMA/core: Fixes for port mapper client registration
  IB/IPoIB: Fix bad error flow in ipoib_add_port()
  IB/mlx4: Do not attemp to report HCA clock offset on VFs
  IB/cm: Do not queue work to a device that's going away
  IB/srp: Avoid using uninitialized variable
  ...

9 years agonet: Fix skb csum races when peeking
Herbert Xu [Mon, 13 Jul 2015 12:01:42 +0000 (20:01 +0800)]
net: Fix skb csum races when peeking

When we calculate the checksum on the recv path, we store the
result in the skb as an optimisation in case we need the checksum
again down the line.

This is in fact bogus for the MSG_PEEK case as this is done without
any locking.  So multiple threads can peek and then store the result
to the same skb, potentially resulting in bogus skb states.

This patch fixes this by only storing the result if the skb is not
shared.  This preserves the optimisations for the few cases where
it can be done safely due to locking or other reasons, e.g., SIOCINQ.

Signed-off-by: Herbert Xu <[email protected]>
Acked-by: Eric Dumazet <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoRevert "net: fec: Ensure clocks are enabled while using mdio bus"
Fabio Estevam [Mon, 13 Jul 2015 11:13:52 +0000 (08:13 -0300)]
Revert "net: fec: Ensure clocks are enabled while using mdio bus"

This reverts commit 6c3e921b18edca290099adfddde8a50236bf2d80.

commit 6c3e921b18ed ("net: fec: Ensure clocks are enabled while using mdio
 bus") prevents the kernel to boot on mx6 boards, so let's revert it.

Reported-by: Tyler Baker <[email protected]>
Signed-off-by: Fabio Estevam <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agoNET: AX.25: Stop heartbeat timer on disconnect.
Richard Stearn [Mon, 13 Jul 2015 09:38:24 +0000 (11:38 +0200)]
NET: AX.25: Stop heartbeat timer on disconnect.

This may result in a kernel panic.  The bug has always existed but
somehow we've run out of luck now and it bites.

Signed-off-by: Richard Stearn <[email protected]>
Cc: [email protected] # all branches
Signed-off-by: Ralf Baechle <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agonet: Clone skb before setting peeked flag
Herbert Xu [Mon, 13 Jul 2015 08:04:13 +0000 (16:04 +0800)]
net: Clone skb before setting peeked flag

Shared skbs must not be modified and this is crucial for broadcast
and/or multicast paths where we use it as an optimisation to avoid
unnecessary cloning.

The function skb_recv_datagram breaks this rule by setting peeked
without cloning the skb first.  This causes funky races which leads
to double-free.

This patch fixes this by cloning the skb and replacing the skb
in the list when setting skb->peeked.

Fixes: a59322be07c9 ("[UDP]: Only increment counter on first peek/recv")
Reported-by: Konstantin Khlebnikov <[email protected]>
Signed-off-by: Herbert Xu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agortnetlink: reject non-IFLA_VF_PORT attributes inside IFLA_VF_PORTS
Daniel Borkmann [Sun, 12 Jul 2015 22:06:02 +0000 (00:06 +0200)]
rtnetlink: reject non-IFLA_VF_PORT attributes inside IFLA_VF_PORTS

Similarly as in commit 4f7d2cdfdde7 ("rtnetlink: verify IFLA_VF_INFO
attributes before passing them to driver"), we have a double nesting
of netlink attributes, i.e. IFLA_VF_PORTS only contains IFLA_VF_PORT
that is nested itself. While IFLA_VF_PORTS is a verified attribute
from ifla_policy[], we only check if the IFLA_VF_PORTS container has
IFLA_VF_PORT attributes and then pass the attribute's content itself
via nla_parse_nested(). It would be more correct to reject inner types
other than IFLA_VF_PORT instead of continuing parsing and also similarly
as in commit 4f7d2cdfdde7, to check for a minimum of NLA_HDRLEN.

Signed-off-by: Daniel Borkmann <[email protected]>
Cc: Roopa Prabhu <[email protected]>
Cc: Scott Feldman <[email protected]>
Cc: Jason Gunthorpe <[email protected]>
Acked-by: Roopa Prabhu <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
9 years agogpio: brcmstb: fix null ptr dereference in driver remove
Gregory Fong [Thu, 18 Jun 2015 01:00:40 +0000 (18:00 -0700)]
gpio: brcmstb: fix null ptr dereference in driver remove

If a failure occurs during probe, brcmstb_gpio_remove() is called. In
remove, we call platform_get_drvdata(), but at the time of failure in
the probe the driver data hadn't yet been set which leads to a NULL
ptr dereference in the remove's list_for_each.  Call
platform_set_drvdata() and set up list head right after allocating the
priv struct to both avoid the null pointer dereference that could
occur today.  To guard against potential future changes, check for
null pointer in remove.

Reported-by: Tim Ross <[email protected]>
Signed-off-by: Gregory Fong <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agogpio: Remove double "base" in comment
Geert Uytterhoeven [Mon, 15 Jun 2015 11:31:33 +0000 (13:31 +0200)]
gpio: Remove double "base" in comment

Signed-off-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: Linus Walleij <[email protected]>
9 years agoNVMe: Reread partitions on metadata formats
Keith Busch [Tue, 14 Jul 2015 17:57:48 +0000 (11:57 -0600)]
NVMe: Reread partitions on metadata formats

This patch has the driver automatically reread partitions if a namespace
has a separate metadata format. Previously revalidating a disk was
sufficient to get the correct capacity set on such formatted drives,
but partitions that may exist would not have been surfaced.

Reported-by: Paul Grabinar <[email protected]>
Signed-off-by: Keith Busch <[email protected]>
Cc: Matthew Wilcox <[email protected]>
Tested-by: Paul Grabinar <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>
9 years agoMerge tag 'locks-v4.2-1' of git://git.samba.org/jlayton/linux
Linus Torvalds [Wed, 15 Jul 2015 20:35:23 +0000 (13:35 -0700)]
Merge tag 'locks-v4.2-1' of git://git.samba.org/jlayton/linux

Pull file locking updates from Jeff Layton:
 "I had thought that I was going to get away without a pull request this
  cycle.  There was a NFSv4 file locking problem that cropped up that I
  tried to fix in the NFSv4 code alone, but that fix has turned out to
  be problematic.  These patches fix this in the correct way.

  Note that this touches some NFSv4 code as well.  Ordinarily I'd wait
  for Trond to ACK this, but he's on holiday right now and the bug is
  rather nasty.  So I suggest we merge this and if he raises issues with
  it we can sort it out when he gets back"

Acked-by: Bruce Fields <[email protected]>
Acked-by: Dan Williams <[email protected]>
 [ +1 to this series fixing a 100% reproducible slab corruption +
   general protection fault in my nfs-root test environment. - Dan ]
Acked-by: Anna Schumaker <[email protected]>
* tag 'locks-v4.2-1' of git://git.samba.org/jlayton/linux:
  locks: inline posix_lock_file_wait and flock_lock_file_wait
  nfs4: have do_vfs_lock take an inode pointer
  locks: new helpers - flock_lock_inode_wait and posix_lock_inode_wait
  locks: have flock_lock_file take an inode pointer instead of a filp
  Revert "nfs: take extra reference to fl->fl_file when running a LOCKU operation"

This page took 0.152993 seconds and 4 git commands to generate.