OSDN Git Service

android-x86/kernel.git
7 years agodmaengine: cppi41: Fix oops in cppi41_runtime_resume
Tony Lindgren [Thu, 19 Jan 2017 16:49:08 +0000 (08:49 -0800)]
dmaengine: cppi41: Fix oops in cppi41_runtime_resume

commit 362f4562466c3b9490e733e06999025638310d4a upstream.

Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support")
together with recent MUSB changes allowed USB and DMA on BeagleBone to idle
when no cable is connected. But looks like few corner case issues still
remain.

Looks like just by re-plugging USB cable about ten or so times on BeagleBone
when configured in USB peripheral mode we can get warnings and eventually
trigger an oops in cppi41 DMA:

WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:1154 cppi41_runtime_suspend+
x28/0x38 [cppi41]
...

WARNING: CPU: 0 PID: 14 at drivers/dma/cppi41.c:452
push_desc_queue+0x94/0x9c [cppi41]
...

Unable to handle kernel NULL pointer dereference at virtual
address 00000104
pgd = c0004000
[00000104] *pgd=00000000
Internal error: Oops: 805 [#1] SMP ARM
...
[<bf0d92cc>] (cppi41_runtime_resume [cppi41]) from [<c0589838>]
(__rpm_callback+0xc0/0x214)
[<c0589838>] (__rpm_callback) from [<c05899ac>] (rpm_callback+0x20/0x80)
[<c05899ac>] (rpm_callback) from [<c0589460>] (rpm_resume+0x504/0x78c)
[<c0589460>] (rpm_resume) from [<c058a1a0>] (pm_runtime_work+0x60/0xa8)
[<c058a1a0>] (pm_runtime_work) from [<c0156120>] (process_one_work+0x2b4/0x808)

This is because of a race with runtime PM and cppi41_dma_issue_pending()
as reported by Alexandre Bailon <abailon@baylibre.com> in earlier
set of patches. Based on mailing list discussions we however came to the
conclusion that a different fix from Alexandre's fix is needed in order
to guarantee that DMA is really active when we try to use it.

To fix the issue, we need to add a driver specific flag as we otherwise
can have -EINPROGRESS state set by runtime PM and can't rely on
pm_runtime_active() to tell us when we can use the DMA.

And we need to make sure the DMA transfers get triggered in the queued
order. So let's always queue the transfers, then flush the queue
from both cppi41_dma_issue_pending() and cppi41_runtime_resume()
as suggested by Grygorii Strashko <grygorii.strashko@ti.com> in an
earlier example patch.

For reference, this is also documented in Documentation/power/runtime_pm.txt
in the example at the end of the file as pointed out by Grygorii Strashko
<grygorii.strashko@ti.com>.

Based on earlier patches from Alexandre Bailon <abailon@baylibre.com>
and Grygorii Strashko <grygorii.strashko@ti.com> modified based on
testing and what was discussed on the mailing lists.

Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support")
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Bin Liu <b-liu@ti.com>
Cc: Grygorii Strashko <grygorii.strashko@ti.com>
Cc: Kevin Hilman <khilman@baylibre.com>
Cc: Patrick Titiano <ptitiano@baylibre.com>
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Reported-by: Alexandre Bailon <abailon@baylibre.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Tested-by: Bin Liu <b-liu@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodmaengine: cppi41: Fix runtime PM timeouts with USB mass storage
Tony Lindgren [Thu, 19 Jan 2017 16:49:07 +0000 (08:49 -0800)]
dmaengine: cppi41: Fix runtime PM timeouts with USB mass storage

commit ae4a3e028bb8b59e7cfeb0cc9ef03d885182ce8b upstream.

Commit fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support")
added runtime PM support for cppi41, but had corner case issues. Some of
the issues were fixed with commit 098de42ad670 ("dmaengine: cppi41: Fix
unpaired pm runtime when only a USB hub is connected"). That fix however
caused a new regression where we can get error -115 messages with USB on
BeagleBone when connecting a USB mass storage device to a hub.

This is because when connecting a USB mass storage device to a hub, the
initial DMA transfers can take over 200ms to complete and cppi41
autosuspend delay times out.

To fix the issue, we want to implement refcounting for chan_busy array
that contains the active dma transfers. Increasing the autosuspend delay
won't help as that the delay could be potentially seconds, and it's best
to let the USB subsystem to deal with the timeouts on errors.

The earlier attempt for runtime PM was buggy as the pm_runtime_get/put()
calls could get unpaired easily as they did not follow the state of
the chan_busy array as described in commit 098de42ad670 ("dmaengine:
cppi41: Fix unpaired pm runtime when only a USB hub is connected".

Let's fix the issue by adding pm_runtime_get() to where a new transfer
is added to the chan_busy array, and calls to pm_runtime_put() where
chan_busy array entry is cleared. This prevents any autosuspend timeouts
from happening while dma transfers are active.

Fixes: 098de42ad670 ("dmaengine: cppi41: Fix unpaired pm runtime when
only a USB hub is connected")
Fixes: fdea2d09b997 ("dmaengine: cppi41: Add basic PM runtime support")
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Bin Liu <b-liu@ti.com>
Cc: Grygorii Strashko <grygorii.strashko@ti.com>
Cc: Kevin Hilman <khilman@baylibre.com>
Cc: Patrick Titiano <ptitiano@baylibre.com>
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Tested-by: Bin Liu <b-liu@ti.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoperf/x86/intel/uncore: Clean up hotplug conversion fallout
Thomas Gleixner [Tue, 31 Jan 2017 22:58:39 +0000 (23:58 +0100)]
perf/x86/intel/uncore: Clean up hotplug conversion fallout

commit 1aa6cfd33df492939b0be15ebdbcff1f8ae5ddb6 upstream.

The recent conversion to the hotplug state machine kept two mechanisms from
the original code:

 1) The first_init logic which adds the number of online CPUs in a package
    to the refcount. That's wrong because the callbacks are executed for
    all online CPUs.

    Remove it so the refcounting is correct.

 2) The on_each_cpu() call to undo box->init() in the error handling
    path. That's bogus because when the prepare callback fails no box has
    been initialized yet.

    Remove it.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Siewior <bigeasy@linutronix.de>
Cc: Stephane Eranian <eranian@google.com>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Yasuaki Ishimatsu <yasu.isimatu@gmail.com>
Fixes: 1a246b9f58c6 ("perf/x86/intel/uncore: Convert to hotplug state machine")
Link: http://lkml.kernel.org/r/20170131230141.298032324@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoHID: wacom: Fix poor prox handling in 'wacom_pl_irq'
Jason Gerecke [Thu, 26 Jan 2017 17:06:22 +0000 (09:06 -0800)]
HID: wacom: Fix poor prox handling in 'wacom_pl_irq'

commit 282e4637bc1c0b338708bcebd09d31c69abec070 upstream.

Commit 025bcc1 performed cleanup work on the 'wacom_pl_irq' function, making
it follow the standards used in the rest of the codebase. The change
unintiontionally allowed the function to send input events from reports
that are not marked as being in prox. This can cause problems as the
report values for X, Y, etc. are not guaranteed to be correct. In
particular, occasionally the tablet will send a report with these values
set to zero. If such a report is received it can caus an unexpected jump
in the XY position.

This patch surrounds more of the processing code with a proximity check,
preventing these zeroed reports from overwriting the current state. To
be safe, only the tool type and ABS_MISC events should be reported when
the pen is marked as being out of prox.

Fixes: 025bcc1540 ("HID: wacom: Simplify 'wacom_pl_irq'")
Signed-off-by: Jason Gerecke <jason.gerecke@wacom.com>
Reviewed-by: Ping Cheng <pingc@wacom.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoHID: hid-lg: Fix immediate disconnection of Logitech Rumblepad 2
Ardinartsev Nikita [Thu, 26 Jan 2017 13:54:42 +0000 (16:54 +0300)]
HID: hid-lg: Fix immediate disconnection of Logitech Rumblepad 2

commit 877a021e08ccb6434718c0cc781fdf943c884cc0 upstream.

With NOGET quirk Logitech F510 is now fully workable in dinput mode including
rumble effects (according to fftest).

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=117091

[jkosina@suse.cz: fix patch format]
Signed-off-by: Ardinartsev Nikita <ardinar23@gmail.com>
Acked-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoHID: usbhid: Quirk a AMI virtual mouse and keyboard with ALWAYS_POLL
Colin Ian King [Thu, 26 Jan 2017 17:34:40 +0000 (17:34 +0000)]
HID: usbhid: Quirk a AMI virtual mouse and keyboard with ALWAYS_POLL

commit ed9ab4287f96e66340e0390e2c583f2f9110cba0 upstream.

Quirking the following AMI USB device with ALWAYS_POLL fixes an AMI
virtual keyboard and mouse from not responding and timing out when
it is attached to a ppc64el Power 8 system and when we have some
rapid open/closes on the mouse device.

 usb 1-3: new high-speed USB device number 2 using xhci_hcd
 usb 1-3: New USB device found, idVendor=046b, idProduct=ff01
 usb 1-3: New USB device strings: Mfr=1, Product=2, SerialNumber=3
 usb 1-3: Product: Virtual Hub
 usb 1-3: Manufacturer: American Megatrends Inc.
 usb 1-3: SerialNumber: serial
 usb 1-3.3: new high-speed USB device number 3 using xhci_hcd
 usb 1-3.3: New USB device found, idVendor=046b, idProduct=ff31
 usb 1-3.3: New USB device strings: Mfr=1, Product=2, SerialNumber=3
 usb 1-3.3: Product: Virtual HardDisk Device
 usb 1-3.3: Manufacturer: American Megatrends Inc.
 usb 1-3.4: new low-speed USB device number 4 using xhci_hcd
 usb 1-3.4: New USB device found, idVendor=046b, idProduct=ff10
 usb 1-3.4: New USB device strings: Mfr=1, Product=2, SerialNumber=0
 usb 1-3.4: Product: Virtual Keyboard and Mouse
 usb 1-3.4: Manufacturer: American Megatrends Inc.

With the quirk I have not been able to trigger the issue with
half an hour of saturation soak testing.

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoiwlwifi: mvm: avoid crash on restart w/o reserved queues
Johannes Berg [Fri, 2 Dec 2016 11:03:36 +0000 (12:03 +0100)]
iwlwifi: mvm: avoid crash on restart w/o reserved queues

commit 03c902bff524e0cf664737a33f2365f7837040bf upstream.

When the firmware restarts in a situation in which any station
has no queue reserved anymore because that queue was used, the
code will crash trying to access the queue_info array at the
offset 255, which is far too big. Fix this by checking that a
queue is actually reserved before writing its status.

Fixes: 8d98ae6eb0d5 ("iwlwifi: mvm: re-assign old queues after hw restart in dqa mode")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoiwlwifi: fix double hyphen in MODULE_FIRMWARE for 8000
Jürg Billeter [Mon, 10 Oct 2016 16:30:01 +0000 (18:30 +0200)]
iwlwifi: fix double hyphen in MODULE_FIRMWARE for 8000

commit 7941c59e45f3b6d30e07375e9b6713427e0a9f98 upstream.

Mistakenly, the driver is trying to load the 8000C firmware with an
incorrect name (i.e. with two hyphens where there should be only one)
and that fails.  Fix that by removing the hyphen from the format
macro.

Fixes: e1ba684f762b ("iwlwifi: 8000: fix MODULE_FIRMWARE input")
Signed-off-by: Jürg Billeter <j@bitron.ch>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agopinctrl: intel: merrifield: Add missed check in mrfld_config_set()
Andy Shevchenko [Tue, 24 Jan 2017 15:28:22 +0000 (17:28 +0200)]
pinctrl: intel: merrifield: Add missed check in mrfld_config_set()

commit 19b26d92dfb70f56440c187a20c49102ab648b97 upstream.

Not every pin can be configured. Add missed check to prevent access
violation.

Fixes: 4e80c8f50574 ("pinctrl: intel: Add Intel Merrifield pin controller support")
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agopinctrl: baytrail: Debounce register is one per community
Andy Shevchenko [Thu, 26 Jan 2017 17:24:08 +0000 (19:24 +0200)]
pinctrl: baytrail: Debounce register is one per community

commit 1b89970d81bbd52720fc64a3fe9572ee33588363 upstream.

Debounce value is set globally per community. Otherwise user will easily
get a kernel crash when they start using the feature:

BUG: unable to handle kernel paging request at ffffc900003be000
IP: byt_gpio_dbg_show+0xa9/0x430

Make it clear in byt_gpio_reg().

Note that this fix just prevents kernel to crash, but doesn't make any
difference to the existing logic. It means the last caller will win the
trade and debounce value will be configured accordingly. The actual
logic fix needs to be thought about and it's not as important as crash
fix. That's why the latter goes separately and right now.

Fixes: 658b476c742f ("pinctrl: baytrail: Add debounce configuration")
Cc: Cristina Ciocan <cristina.ciocan@intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Jean Delvare <jdelvare@suse.de>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoRevert "vring: Force use of DMA API for ARM-based systems with legacy devices"
Michael S. Tsirkin [Fri, 3 Feb 2017 03:43:52 +0000 (05:43 +0200)]
Revert "vring: Force use of DMA API for ARM-based systems with legacy devices"

commit 0d5415b489f68b58e1983a53793d25d53098ed4b upstream.

This reverts commit c7070619f3408d9a0dffbed9149e6f00479cf43b.

This has been shown to regress on some ARM systems:

by forcing on DMA API usage for ARM systems, we have inadvertently
kicked open a hornets' nest in terms of cache-coherency. Namely that
unless the virtio device is explicitly described as capable of coherent
DMA by firmware, the DMA APIs on ARM and other DT-based platforms will
assume it is non-coherent. This turns out to cause a big problem for the
likes of QEMU and kvmtool, which generate virtio-mmio devices in their
guest DTs but neglect to add the often-overlooked "dma-coherent"
property; as a result, we end up with the guest making non-cacheable
accesses to the vring, the host doing so cacheably, both talking past
each other and things going horribly wrong.

We are working on a safer work-around.

Fixes: c7070619f340 ("vring: Force use of DMA API for ARM-based systems with legacy devices")
Reported-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoRevert "bcma: init serial console directly from ChipCommon code"
Rafał Miłecki [Fri, 13 Jan 2017 11:23:35 +0000 (12:23 +0100)]
Revert "bcma: init serial console directly from ChipCommon code"

commit 7195439d1d71bc4a6c33cfb57bc669a7cd041041 upstream.

This reverts commit 4c81acab3816 ("bcma: init serial console directly
from ChipCommon code") as it broke IRQ assignment. Getting IRQ with
bcma_core_irq helper on SoC requires MIPS core to be set. It happens
*after* ChipCommon initialization so we can't do this so early.

This fixes a user reported regression. It wasn't critical as serial was
still somehow working but lack of IRQs was making in unreliable.

Fixes: 4c81acab3816 ("bcma: init serial console directly from ChipCommon code")
Reported-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agopercpu-refcount: fix reference leak during percpu-atomic transition
Douglas Miller [Sat, 28 Jan 2017 12:42:20 +0000 (06:42 -0600)]
percpu-refcount: fix reference leak during percpu-atomic transition

commit 966d2b04e070bc040319aaebfec09e0144dc3341 upstream.

percpu_ref_tryget() and percpu_ref_tryget_live() should return
"true" IFF they acquire a reference. But the return value from
atomic_long_inc_not_zero() is a long and may have high bits set,
e.g. PERCPU_COUNT_BIAS, and the return value of the tryget routines
is bool so the reference may actually be acquired but the routines
return "false" which results in a reference leak since the caller
assumes it does not need to do a corresponding percpu_ref_put().

This was seen when performing CPU hotplug during I/O, as hangs in
blk_mq_freeze_queue_wait where percpu_ref_kill (blk_mq_freeze_queue_start)
raced with percpu_ref_tryget (blk_mq_timeout_work).
Sample stack trace:

__switch_to+0x2c0/0x450
__schedule+0x2f8/0x970
schedule+0x48/0xc0
blk_mq_freeze_queue_wait+0x94/0x120
blk_mq_queue_reinit_work+0xb8/0x180
blk_mq_queue_reinit_prepare+0x84/0xa0
cpuhp_invoke_callback+0x17c/0x600
cpuhp_up_callbacks+0x58/0x150
_cpu_up+0xf0/0x1c0
do_cpu_up+0x120/0x150
cpu_subsys_online+0x64/0xe0
device_online+0xb4/0x120
online_store+0xb4/0xc0
dev_attr_store+0x68/0xa0
sysfs_kf_write+0x80/0xb0
kernfs_fop_write+0x17c/0x250
__vfs_write+0x6c/0x1e0
vfs_write+0xd0/0x270
SyS_write+0x6c/0x110
system_call+0x38/0xe0

Examination of the queue showed a single reference (no PERCPU_COUNT_BIAS,
and __PERCPU_REF_DEAD, __PERCPU_REF_ATOMIC set) and no requests.
However, conditions at the time of the race are count of PERCPU_COUNT_BIAS + 0
and __PERCPU_REF_DEAD and __PERCPU_REF_ATOMIC set.

The fix is to make the tryget routines use an actual boolean internally instead
of the atomic long result truncated to a int.

Fixes: e625305b3907 percpu-refcount: make percpu_ref based on longs instead of ints
Link: https://bugzilla.kernel.org/show_bug.cgi?id=190751
Signed-off-by: Douglas Miller <dougmill@linux.vnet.ibm.com>
Reviewed-by: Jens Axboe <axboe@fb.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: e625305b3907 ("percpu-refcount: make percpu_ref based on longs instead of ints")
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoregulator: axp20x: AXP806: Fix dcdcb being set instead of dcdce
Rask Ingemann Lambertsen [Sat, 21 Jan 2017 16:11:43 +0000 (17:11 +0100)]
regulator: axp20x: AXP806: Fix dcdcb being set instead of dcdce

commit d0e287a401d9acf67b75180b26e2d62b7d482652 upstream.

A typo or copy-paste bug means that the register access intended for
regulator dcdce goes to dcdcb instead. This patch corrects it.

Fixes: 2ca342d391e3 (regulator: axp20x: Support AXP806 variant)
Signed-off-by: Rask Ingemann Lambertsen <rask@formelder.dk>
Acked-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agovhost: fix initialization for vq->is_le
Halil Pasic [Mon, 30 Jan 2017 10:09:36 +0000 (11:09 +0100)]
vhost: fix initialization for vq->is_le

commit cda8bba0f99d25d2061c531113c14fa41effc3ae upstream.

Currently, under certain circumstances vhost_init_is_le does just a part
of the initialization job, and depends on vhost_reset_is_le being called
too. For this reason vhost_vq_init_access used to call vhost_reset_is_le
when vq->private_data is NULL. This is not only counter intuitive, but
also real a problem because it breaks vhost_net. The bug was introduced to
vhost_net with commit 2751c9882b94 ("vhost: cross-endian support for
legacy devices"). The symptom is corruption of the vq's used.idx field
(virtio) after VHOST_NET_SET_BACKEND was issued as a part of the vhost
shutdown on a vq with pending descriptors.

Let us make sure the outcome of vhost_init_is_le never depend on the state
it is actually supposed to initialize, and fix virtio_net by removing the
reset from vhost_vq_init_access.

With the above, there is no reason for vhost_reset_is_le to do just half
of the job. Let us make vhost_reset_is_le reinitialize is_le.

Signed-off-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Reported-by: Michael A. Tebolt <miket@us.ibm.com>
Reported-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Fixes: commit 2751c9882b94 ("vhost: cross-endian support for legacy devices")
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Tested-by: Michael A. Tebolt <miket@us.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agommc: sdhci: Ignore unexpected CARD_INT interrupts
Gabriel Krisman Bertazi [Mon, 16 Jan 2017 14:23:42 +0000 (12:23 -0200)]
mmc: sdhci: Ignore unexpected CARD_INT interrupts

commit 161e6d44a5e2d3f85365cb717d60e363171b39e6 upstream.

One of our kernelCI boxes hanged at boot because a faulty eSDHC device
was triggering spurious CARD_INT interrupts for SD cards, causing CMD52
reads, which are not allowed for SD devices.  This adds a sanity check
to the interruption path, preventing that illegal command from getting
sent if the CARD_INT interruption should be disabled.

This quirk allows that particular machine to resume boot despite the
faulty hardware, instead of getting hung dealing with thousands of
mishandled interrupts.

Suggested-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocgroup: don't online subsystems before cgroup_name/path() are operational
Tejun Heo [Thu, 26 Jan 2017 21:47:28 +0000 (16:47 -0500)]
cgroup: don't online subsystems before cgroup_name/path() are operational

commit 07cd12945551b63ecb1a349d50a6d69d1d6feb4a upstream.

While refactoring cgroup creation, a5bca2152036 ("cgroup: factor out
cgroup_create() out of cgroup_mkdir()") incorrectly onlined subsystems
before the new cgroup is associated with it kernfs_node.  This is fine
for cgroup proper but cgroup_name/path() depend on the associated
kernfs_node and if a subsystem makes the new cgroup_subsys_state
visible, which they're allowed to after onlining, it can lead to NULL
dereference.

The current code performs cgroup creation and subsystem onlining in
cgroup_create() and cgroup_mkdir() makes the cgroup and subsystems
visible afterwards.  There's no reason to online the subsystems early
and we can simply drop cgroup_apply_control_enable() call from
cgroup_create() so that the subsystems are onlined and made visible at
the same time.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Fixes: a5bca2152036 ("cgroup: factor out cgroup_create() out of cgroup_mkdir()")
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocan: bcm: fix hrtimer/tasklet termination in bcm op removal
Oliver Hartkopp [Wed, 18 Jan 2017 20:30:51 +0000 (21:30 +0100)]
can: bcm: fix hrtimer/tasklet termination in bcm op removal

commit a06393ed03167771246c4c43192d9c264bc48412 upstream.

When removing a bcm tx operation either a hrtimer or a tasklet might run.
As the hrtimer triggers its associated tasklet and vice versa we need to
take care to mutually terminate both handlers.

Reported-by: Michael Josenhans <michael.josenhans@web.de>
Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Tested-by: Michael Josenhans <michael.josenhans@web.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agotracing: Fix hwlat kthread migration
Steven Rostedt (VMware) [Tue, 31 Jan 2017 00:27:10 +0000 (19:27 -0500)]
tracing: Fix hwlat kthread migration

commit 79c6f448c8b79c321e4a1f31f98194e4f6b6cae7 upstream.

The hwlat tracer creates a kernel thread at start of the tracer. It is
pinned to a single CPU and will move to the next CPU after each period of
running. If the user modifies the migration thread's affinity, it will not
change after that happens.

The original code created the thread at the first instance it was called,
but later was changed to destroy the thread after the tracer was finished,
and would not be created until the next instance of the tracer was
established. The code that initialized the affinity was only called on the
initial instantiation of the tracer. After that, it was not initialized, and
the previous affinity did not match the current newly created one, making
it appear that the user modified the thread's affinity when it did not, and
the thread failed to migrate again.

Fixes: 0330f7aa8ee6 ("tracing: Have hwlat trace migrate across tracing_cpumask CPUs")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomm, fs: check for fatal signals in do_generic_file_read()
Michal Hocko [Fri, 3 Feb 2017 21:13:29 +0000 (13:13 -0800)]
mm, fs: check for fatal signals in do_generic_file_read()

commit 5abf186a30a89d5b9c18a6bf93a2c192c9fd52f6 upstream.

do_generic_file_read() can be told to perform a large request from
userspace.  If the system is under OOM and the reading task is the OOM
victim then it has an access to memory reserves and finishing the full
request can lead to the full memory depletion which is dangerous.  Make
sure we rather go with a short read and allow the killed task to
terminate.

Link: http://lkml.kernel.org/r/20170201092706.9966-3-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.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>
7 years agobase/memory, hotplug: fix a kernel oops in show_valid_zones()
Toshi Kani [Fri, 3 Feb 2017 21:13:23 +0000 (13:13 -0800)]
base/memory, hotplug: fix a kernel oops in show_valid_zones()

commit a96dfddbcc04336bbed50dc2b24823e45e09e80c upstream.

Reading a sysfs "memoryN/valid_zones" file leads to the following oops
when the first page of a range is not backed by struct page.
show_valid_zones() assumes that 'start_pfn' is always valid for
page_zone().

 BUG: unable to handle kernel paging request at ffffea017a000000
 IP: show_valid_zones+0x6f/0x160

This issue may happen on x86-64 systems with 64GiB or more memory since
their memory block size is bumped up to 2GiB.  [1] An example of such
systems is desribed below.  0x3240000000 is only aligned by 1GiB and
this memory block starts from 0x3200000000, which is not backed by
struct page.

 BIOS-e820: [mem 0x0000003240000000-0x000000603fffffff] usable

Since test_pages_in_a_zone() already checks holes, fix this issue by
extending this function to return 'valid_start' and 'valid_end' for a
given range.  show_valid_zones() then proceeds with the valid range.

[1] 'Commit bdee237c0343 ("x86: mm: Use 2GB memory block size on
    large-memory x86-64 systems")'

Link: http://lkml.kernel.org/r/20170127222149.30893-3-toshi.kani@hpe.com
Signed-off-by: Toshi Kani <toshi.kani@hpe.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Zhang Zhen <zhenzhang.zhang@huawei.com>
Cc: Reza Arbab <arbab@linux.vnet.ibm.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/memory_hotplug.c: check start_pfn in test_pages_in_a_zone()
Toshi Kani [Fri, 3 Feb 2017 21:13:20 +0000 (13:13 -0800)]
mm/memory_hotplug.c: check start_pfn in test_pages_in_a_zone()

commit deb88a2a19e85842d79ba96b05031739ec327ff4 upstream.

Patch series "fix a kernel oops when reading sysfs valid_zones", v2.

A sysfs memory file is created for each 2GiB memory block on x86-64 when
the system has 64GiB or more memory.  [1] When the start address of a
memory block is not backed by struct page, i.e.  a memory range is not
aligned by 2GiB, reading its 'valid_zones' attribute file leads to a
kernel oops.  This issue was observed on multiple x86-64 systems with
more than 64GiB of memory.  This patch-set fixes this issue.

Patch 1 first fixes an issue in test_pages_in_a_zone(), which does not
test the start section.

Patch 2 then fixes the kernel oops by extending test_pages_in_a_zone()
to return valid [start, end).

Note for stable kernels: The memory block size change was made by commit
bdee237c0343 ("x86: mm: Use 2GB memory block size on large-memory x86-64
systems"), which was accepted to 3.9.  However, this patch-set depends
on (and fixes) the change to test_pages_in_a_zone() made by commit
5f0f2887f4de ("mm/memory_hotplug.c: check for missing sections in
test_pages_in_a_zone()"), which was accepted to 4.4.

So, I recommend that we backport it up to 4.4.

[1] 'Commit bdee237c0343 ("x86: mm: Use 2GB memory block size on
    large-memory x86-64 systems")'

This patch (of 2):

test_pages_in_a_zone() does not check 'start_pfn' when it is aligned by
section since 'sec_end_pfn' is set equal to 'pfn'.  Since this function
is called for testing the range of a sysfs memory file, 'start_pfn' is
always aligned by section.

Fix it by properly setting 'sec_end_pfn' to the next section pfn.

Also make sure that this function returns 1 only when the range belongs
to a zone.

Link: http://lkml.kernel.org/r/20170127222149.30893-2-toshi.kani@hpe.com
Signed-off-by: Toshi Kani <toshi.kani@hpe.com>
Cc: Andrew Banman <abanman@sgi.com>
Cc: Reza Arbab <arbab@linux.vnet.ibm.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocifs: initialize file_info_lock
Rabin Vincent [Fri, 13 Jan 2017 14:00:16 +0000 (15:00 +0100)]
cifs: initialize file_info_lock

commit 81ddd8c0c5e1cb41184d66567140cb48c53eb3d1 upstream.

Reviewed-by: Jeff Layton <jlayton@redhat.com>
file_info_lock is not initalized in initiate_cifs_search(), leading to the
following splat after a simple "mount.cifs ... dir && ls dir/":

 BUG: spinlock bad magic on CPU#0, ls/486
  lock: 0xffff880009301110, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
 CPU: 0 PID: 486 Comm: ls Not tainted 4.9.0 #27
 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996)
  ffffc900042f3db0 ffffffff81327533 0000000000000000 ffff880009301110
  ffffc900042f3dd0 ffffffff810baf75 ffff880009301110 ffffffff817ae077
  ffffc900042f3df0 ffffffff810baff6 ffff880009301110 ffff880008d69900
 Call Trace:
  [<ffffffff81327533>] dump_stack+0x65/0x92
  [<ffffffff810baf75>] spin_dump+0x85/0xe0
  [<ffffffff810baff6>] spin_bug+0x26/0x30
  [<ffffffff810bb159>] do_raw_spin_lock+0xe9/0x130
  [<ffffffff8159ad2f>] _raw_spin_lock+0x1f/0x30
  [<ffffffff8127e50d>] cifs_closedir+0x4d/0x100
  [<ffffffff81181cfd>] __fput+0x5d/0x160
  [<ffffffff81181e3e>] ____fput+0xe/0x10
  [<ffffffff8109410e>] task_work_run+0x7e/0xa0
  [<ffffffff81002512>] exit_to_usermode_loop+0x92/0xa0
  [<ffffffff810026f9>] syscall_return_slowpath+0x49/0x50
  [<ffffffff8159b484>] entry_SYSCALL_64_fastpath+0xa7/0xa9

Fixes: 3afca265b5f53a0 ("Clarify locking of cifs file and tcon structures and make more granular")
Signed-off-by: Rabin Vincent <rabinv@axis.com>
Signed-off-by: Steve French <smfrench@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agozswap: disable changing params if init fails
Dan Streetman [Fri, 3 Feb 2017 21:13:09 +0000 (13:13 -0800)]
zswap: disable changing params if init fails

commit d7b028f56a971a2e4d8d7887540a144eeefcd4ab upstream.

Add zswap_init_failed bool that prevents changing any of the module
params, if init_zswap() fails, and set zswap_enabled to false.  Change
'enabled' param to a callback, and check zswap_init_failed before
allowing any change to 'enabled', 'zpool', or 'compressor' params.

Any driver that is built-in to the kernel will not be unloaded if its
init function returns error, and its module params remain accessible for
users to change via sysfs.  Since zswap uses param callbacks, which
assume that zswap has been initialized, changing the zswap params after
a failed initialization will result in WARNING due to the param
callbacks expecting a pool to already exist.  This prevents that by
immediately exiting any of the param callbacks if initialization failed.

This was reported here:
  https://marc.info/?l=linux-mm&m=147004228125528&w=4

And fixes this WARNING:
  [  429.723476] WARNING: CPU: 0 PID: 5140 at mm/zswap.c:503 __zswap_pool_current+0x56/0x60

The warning is just noise, and not serious.  However, when init fails,
zswap frees all its percpu dstmem pages and its kmem cache.  The kmem
cache might be serious, if kmem_cache_alloc(NULL, gfp) has problems; but
the percpu dstmem pages are definitely a problem, as they're used as
temporary buffer for compressed pages before copying into place in the
zpool.

If the user does get zswap enabled after an init failure, then zswap
will likely Oops on the first page it tries to compress (or worse, start
corrupting memory).

Fixes: 90b0fc26d5db ("zswap: change zpool/compressor at runtime")
Link: http://lkml.kernel.org/r/20170124200259.16191-2-ddstreet@ieee.org
Signed-off-by: Dan Streetman <dan.streetman@canonical.com>
Reported-by: Marcin Miroslaw <marcin@mejor.pl>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Minchan Kim <minchan@kernel.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>
7 years agosvcrpc: fix oops in absence of krb5 module
J. Bruce Fields [Tue, 31 Jan 2017 16:37:50 +0000 (11:37 -0500)]
svcrpc: fix oops in absence of krb5 module

commit 034dd34ff4916ec1f8f74e39ca3efb04eab2f791 upstream.

Olga Kornievskaia says: "I ran into this oops in the nfsd (below)
(4.10-rc3 kernel). To trigger this I had a client (unsuccessfully) try
to mount the server with krb5 where the server doesn't have the
rpcsec_gss_krb5 module built."

The problem is that rsci.cred is copied from a svc_cred structure that
gss_proxy didn't properly initialize.  Fix that.

[120408.542387] general protection fault: 0000 [#1] SMP
...
[120408.565724] CPU: 0 PID: 3601 Comm: nfsd Not tainted 4.10.0-rc3+ #16
[120408.567037] Hardware name: VMware, Inc. VMware Virtual =
Platform/440BX Desktop Reference Platform, BIOS 6.00 07/02/2015
[120408.569225] task: ffff8800776f95c0 task.stack: ffffc90003d58000
[120408.570483] RIP: 0010:gss_mech_put+0xb/0x20 [auth_rpcgss]
...
[120408.584946]  ? rsc_free+0x55/0x90 [auth_rpcgss]
[120408.585901]  gss_proxy_save_rsc+0xb2/0x2a0 [auth_rpcgss]
[120408.587017]  svcauth_gss_proxy_init+0x3cc/0x520 [auth_rpcgss]
[120408.588257]  ? __enqueue_entity+0x6c/0x70
[120408.589101]  svcauth_gss_accept+0x391/0xb90 [auth_rpcgss]
[120408.590212]  ? try_to_wake_up+0x4a/0x360
[120408.591036]  ? wake_up_process+0x15/0x20
[120408.592093]  ? svc_xprt_do_enqueue+0x12e/0x2d0 [sunrpc]
[120408.593177]  svc_authenticate+0xe1/0x100 [sunrpc]
[120408.594168]  svc_process_common+0x203/0x710 [sunrpc]
[120408.595220]  svc_process+0x105/0x1c0 [sunrpc]
[120408.596278]  nfsd+0xe9/0x160 [nfsd]
[120408.597060]  kthread+0x101/0x140
[120408.597734]  ? nfsd_destroy+0x60/0x60 [nfsd]
[120408.598626]  ? kthread_park+0x90/0x90
[120408.599448]  ret_from_fork+0x22/0x30

Fixes: 1d658336b05f "SUNRPC: Add RPC based upcall mechanism for RPCGSS auth"
Cc: Simo Sorce <simo@redhat.com>
Reported-by: Olga Kornievskaia <kolga@netapp.com>
Tested-by: Olga Kornievskaia <kolga@netapp.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoNFSD: Fix a null reference case in find_or_create_lock_stateid()
Kinglong Mee [Wed, 18 Jan 2017 11:04:42 +0000 (19:04 +0800)]
NFSD: Fix a null reference case in find_or_create_lock_stateid()

commit d19fb70dd68c4e960e2ac09b0b9c79dfdeefa726 upstream.

nfsd assigns the nfs4_free_lock_stateid to .sc_free in init_lock_stateid().

If nfsd doesn't go through init_lock_stateid() and put stateid at end,
there is a NULL reference to .sc_free when calling nfs4_put_stid(ns).

This patch let the nfs4_stid.sc_free assignment to nfs4_alloc_stid().

Fixes: 356a95ece7aa "nfsd: clean up races in lock stateid searching..."
Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agopowerpc/mm: Use the correct pointer when setting a 2MB pte
Reza Arbab [Wed, 25 Jan 2017 15:54:33 +0000 (09:54 -0600)]
powerpc/mm: Use the correct pointer when setting a 2MB pte

commit a0615a16f7d0ceb5804d295203c302d496d8ee91 upstream.

When setting a 2MB pte, radix__map_kernel_page() is using the address

ptep = (pte_t *)pudp;

Fix this conversion to use pmdp instead. Use pmdp_ptep() to do this
instead of casting the pointer.

Fixes: 2bfd65e45e87 ("powerpc/mm/radix: Add radix callbacks for early init routines")
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Reza Arbab <arbab@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agopowerpc: Fix build failure with clang due to BUILD_BUG_ON()
Michael Ellerman [Tue, 24 Jan 2017 05:36:57 +0000 (16:36 +1100)]
powerpc: Fix build failure with clang due to BUILD_BUG_ON()

commit b5fa0f7f88edcde37df1807fdf9ff10ec787a60e upstream.

Anton says: In commit 4db7327194db ("powerpc: Add option to use jump
label for cpu_has_feature()") and commit c12e6f24d413 ("powerpc: Add
option to use jump label for mmu_has_feature()") we added:

  BUILD_BUG_ON(!__builtin_constant_p(feature))

to cpu_has_feature() and mmu_has_feature() in order to catch usage
issues (such as cpu_has_feature(cpu_has_feature(X), which has happened
once in the past). Unfortunately LLVM isn't smart enough to resolve
this, and it errors out.

I work around it in my clang/LLVM builds of the kernel, but I have just
discovered that it causes a lot of issues for the bcc (eBPF) trace tool
(which uses LLVM).

For now just #ifdef it away for clang builds.

Fixes: 4db7327194db ("powerpc: Add option to use jump label for cpu_has_feature()")
Fixes: c12e6f24d413 ("powerpc: Add option to use jump label for mmu_has_feature()")
Reported-by: Anton Blanchard <anton@samba.org>
Tested-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agopowerpc: Add missing error check to prom_find_boot_cpu()
Darren Stevens [Mon, 23 Jan 2017 19:42:54 +0000 (19:42 +0000)]
powerpc: Add missing error check to prom_find_boot_cpu()

commit af2b7fa17eb92e52b65f96604448ff7a2a89ee99 upstream.

prom_init.c calls 'instance-to-package' twice, but the return
is not checked during prom_find_boot_cpu(). The result is then
passed to prom_getprop(), which could be PROM_ERROR. Add a return check
to prevent this.

This was found on a pasemi system, where CFE doesn't have a working
'instance-to package' prom call.

Before Commit 5c0484e25ec0 ('powerpc: Endian safe trampoline') the area
around addr 0 was mostly 0's and this doesn't cause a problem. Once the
macro 'FIXUP_ENDIAN' has been added to head_64.S, the low memory area
now has non-zero values, which cause the prom_getprop() call
to hang.

mpe: Also confirmed that under SLOF if 'instance-to-package' did fail
with PROM_ERROR we would crash in SLOF. So the bug is not specific to
CFE, it's just that other open firmwares don't trigger it because they
have a working 'instance-to-package'.

Fixes: 5c0484e25ec0 ("powerpc: Endian safe trampoline")
Signed-off-by: Darren Stevens <darren@stevens-zone.net>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agopowerpc/eeh: Fix wrong flag passed to eeh_unfreeze_pe()
Gavin Shan [Wed, 18 Jan 2017 23:10:16 +0000 (10:10 +1100)]
powerpc/eeh: Fix wrong flag passed to eeh_unfreeze_pe()

commit f05fea5b3574a5926c53865eea27139bb40b2f2b upstream.

In __eeh_clear_pe_frozen_state(), we should pass the flag's value
instead of its address to eeh_unfreeze_pe(). The isolated flag is
cleared if no error returned from __eeh_clear_pe_frozen_state(). We
never observed the error from the function. So the isolated flag should
have been always cleared, no real issue is caused because of the misused
@flag.

This fixes the code by passing the value of @flag to eeh_unfreeze_pe().

Fixes: 5cfb20b96f6 ("powerpc/eeh: Emulate EEH recovery for VFIO devices")
Signed-off-by: Gavin Shan <gwshan@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agolibata: Fix ATA request sense
Damien Le Moal [Mon, 19 Dec 2016 01:17:40 +0000 (10:17 +0900)]
libata: Fix ATA request sense

commit 2dae99558e86894e9e5dbf097477baaa5eb70134 upstream.

For an ATA device supporting the sense data reporting feature set, a
failed command will trigger the execution of ata_eh_request_sense if
the result task file of the failed command has the ATA_SENSE bit set
(sense data available bit). ata_eh_request_sense executes the REQUEST
SENSE DATA EXT command to retrieve the sense data of the failed
command. On success of REQUEST SENSE DATA EXT, the ATA_SENSE bit will
NOT be set (the command succeeded) but ata_eh_request_sense
nevertheless tests the availability of sense data by testing that bit
presence in the result tf of the REQUEST SENSE DATA EXT command.  This
leads us to falsely assume that request sense data failed and to the
warning message:

atax.xx: request sense failed stat 50 emask 0

Upon success of REQUEST SENSE DATA EXT, set the ATA_SENSE bit in the
result task file command so that sense data can be returned by
ata_eh_request_sense.

Signed-off-by: Damien Le Moal <damien.lemoal@wdc.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agolibata: apply MAX_SEC_1024 to all CX1-JB*-HP devices
Tejun Heo [Fri, 6 Jan 2017 16:48:50 +0000 (11:48 -0500)]
libata: apply MAX_SEC_1024 to all CX1-JB*-HP devices

commit e0edc8c546463f268d41d064d855bcff994c52fa upstream.

Marko reports that CX1-JB512-HP shows the same timeout issues as
CX1-JB256-HP.  Let's apply MAX_SEC_128 to all devices in the series.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Marko Koski-Vähälä <marko@koski-vahala.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoata: sata_mv:- Handle return value of devm_ioremap.
Arvind Yadav [Mon, 12 Dec 2016 17:43:27 +0000 (23:13 +0530)]
ata: sata_mv:- Handle return value of devm_ioremap.

commit 064c3db9c564cc5be514ac21fb4aa26cc33db746 upstream.

Here, If devm_ioremap will fail. It will return NULL.
Then hpriv->base = NULL - 0x20000; Kernel can run into
a NULL-pointer dereference. This error check will avoid
NULL pointer dereference.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoperf/core: Fix PERF_RECORD_MMAP2 prot/flags for anonymous memory
Peter Zijlstra [Thu, 26 Jan 2017 22:15:08 +0000 (23:15 +0100)]
perf/core: Fix PERF_RECORD_MMAP2 prot/flags for anonymous memory

commit 0b3589be9b98994ce3d5aeca52445d1f5627c4ba upstream.

Andres reported that MMAP2 records for anonymous memory always have
their protection field 0.

Turns out, someone daft put the prot/flags generation code in the file
branch, leaving them unset for anonymous memory.

Reported-by: Andres Freund <andres@anarazel.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Don Zickus <dzickus@redhat.com
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@gmail.com>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@kernel.org
Cc: anton@ozlabs.org
Cc: namhyung@kernel.org
Fixes: f972eb63b100 ("perf: Pass protection and flags bits through mmap2 interface")
Link: http://lkml.kernel.org/r/20170126221508.GF6536@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoperf/core: Fix use-after-free bug
Peter Zijlstra [Thu, 26 Jan 2017 15:39:55 +0000 (16:39 +0100)]
perf/core: Fix use-after-free bug

commit a76a82a3e38c8d3fb6499e3dfaeb0949241ab588 upstream.

Dmitry reported a KASAN use-after-free on event->group_leader.

It turns out there's a hole in perf_remove_from_context() due to
event_function_call() not calling its function when the task
associated with the event is already dead.

In this case the event will have been detached from the task, but the
grouping will have been retained, such that group operations might
still work properly while there are live child events etc.

This does however mean that we can miss a perf_group_detach() call
when the group decomposes, this in turn can then lead to
use-after-free.

Fix it by explicitly doing the group detach if its still required.

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Tested-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: syzkaller <syzkaller@googlegroups.com>
Fixes: 63b6da39bb38 ("perf: Fix perf_event_exit_task() race")
Link: http://lkml.kernel.org/r/20170126153955.GD6515@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocrypto: arm64/aes-blk - honour iv_out requirement in CBC and CTR modes
Ard Biesheuvel [Tue, 17 Jan 2017 13:46:29 +0000 (13:46 +0000)]
crypto: arm64/aes-blk - honour iv_out requirement in CBC and CTR modes

commit 11e3b725cfc282efe9d4a354153e99d86a16af08 upstream.

Update the ARMv8 Crypto Extensions and the plain NEON AES implementations
in CBC and CTR modes to return the next IV back to the skcipher API client.
This is necessary for chaining to work correctly.

Note that for CTR, this is only done if the request is a round multiple of
the block size, since otherwise, chaining is impossible anyway.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agocrypto: api - Clear CRYPTO_ALG_DEAD bit before registering an alg
Salvatore Benedetto [Fri, 13 Jan 2017 11:54:08 +0000 (11:54 +0000)]
crypto: api - Clear CRYPTO_ALG_DEAD bit before registering an alg

commit d6040764adcb5cb6de1489422411d701c158bb69 upstream.

Make sure CRYPTO_ALG_DEAD bit is cleared before proceeding with
the algorithm registration. This fixes qat-dh registration when
driver is restarted

Signed-off-by: Salvatore Benedetto <salvatore.benedetto@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodrm/nouveau/nv1a,nv1f/disp: fix memory clock rate retrieval
Ilia Mirkin [Fri, 20 Jan 2017 03:56:30 +0000 (22:56 -0500)]
drm/nouveau/nv1a,nv1f/disp: fix memory clock rate retrieval

commit 24bf7ae359b8cca165bb30742d2b1c03a1eb23af upstream.

Based on the xf86-video-nv code, NFORCE (NV1A) and NFORCE2 (NV1F) have a
different way of retrieving clocks. See the
nv_hw.c:nForceUpdateArbitrationSettings function in the original code
for how these clocks were accessed.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54587
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodrm/nouveau/disp/gt215: Fix HDA ELD handling (thus, HDMI audio) on gt215
Alastair Bridgewater [Wed, 11 Jan 2017 20:47:18 +0000 (15:47 -0500)]
drm/nouveau/disp/gt215: Fix HDA ELD handling (thus, HDMI audio) on gt215

commit d347583a39e2df609a9e40c835f72d3614665b53 upstream.

Store the ELD correctly, not just enough copies of the first byte
to pad out the given ELD size.

Signed-off-by: Alastair Bridgewater <alastair.bridgewater@gmail.com>
Fixes: 120b0c39c756 ("drm/nv50-/disp: audit and version SOR_HDA_ELD method")
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agodrm/amdgpu/si: fix crash on headless asics
Alex Deucher [Fri, 27 Jan 2017 15:31:52 +0000 (10:31 -0500)]
drm/amdgpu/si: fix crash on headless asics

commit 57bcd0a6364cd4eaa362d7ff1777e88ddf501602 upstream.

Missing check for crtcs present.

Fixes:
https://bugzilla.kernel.org/show_bug.cgi?id=193341
https://bugs.freedesktop.org/show_bug.cgi?id=99387

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agopinctrl: baytrail: Add missing spinlock usage in byt_gpio_irq_handler
Alexander Stein [Mon, 30 Jan 2017 11:35:28 +0000 (12:35 +0100)]
pinctrl: baytrail: Add missing spinlock usage in byt_gpio_irq_handler

commit cdca06e4e85974d8a3503ab15709dbbaf90d3dd1 upstream.

According to VLI64 Intel Atom E3800 Specification Update (#329901)
concurrent read accesses may result in returning 0xffffffff and write
accesses may be dropped silently.
To workaround all accesses must be protected by locks.

Signed-off-by: Alexander Stein <alexander.stein@systec-electronic.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoHID: cp2112: fix gpio-callback error handling
Johan Hovold [Mon, 30 Jan 2017 10:26:39 +0000 (11:26 +0100)]
HID: cp2112: fix gpio-callback error handling

commit 8e9faa15469ed7c7467423db4c62aeed3ff4cae3 upstream.

In case of a zero-length report, the gpio direction_input callback would
currently return success instead of an errno.

Fixes: 1ffb3c40ffb5 ("HID: cp2112: make transfer buffers DMA capable")
Signed-off-by: Johan Hovold <johan@kernel.org>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoHID: cp2112: fix sleep-while-atomic
Johan Hovold [Mon, 30 Jan 2017 10:26:38 +0000 (11:26 +0100)]
HID: cp2112: fix sleep-while-atomic

commit 7a7b5df84b6b4e5d599c7289526eed96541a0654 upstream.

A recent commit fixing DMA-buffers on stack added a shared transfer
buffer protected by a spinlock. This is broken as the USB HID request
callbacks can sleep. Fix this up by replacing the spinlock with a mutex.

Fixes: 1ffb3c40ffb5 ("HID: cp2112: make transfer buffers DMA capable")
Signed-off-by: Johan Hovold <johan@kernel.org>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@redhat.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxtensa: fix noMMU build on cores with MMU
Max Filippov [Wed, 1 Feb 2017 02:35:37 +0000 (18:35 -0800)]
xtensa: fix noMMU build on cores with MMU

commit 4b3e6f2ef3722f1a6a97b6034ed492c1a21fd4ae upstream.

Commit bf15f86b343ed8 ("xtensa: initialize MMU before jumping to reset
vector") calls MMU management functions even when CONFIG_MMU is not
selected. That breaks noMMU build on cores with MMU.

Don't manage MMU when CONFIG_MMU is not selected.

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoefi/fdt: Avoid FDT manipulation after ExitBootServices()
Ard Biesheuvel [Wed, 1 Feb 2017 17:45:02 +0000 (17:45 +0000)]
efi/fdt: Avoid FDT manipulation after ExitBootServices()

commit c8f325a59cfc718d13a50fbc746ed9b415c25e92 upstream.

Some AArch64 UEFI implementations disable the MMU in ExitBootServices(),
after which unaligned accesses to RAM are no longer supported.

Commit:

  abfb7b686a3e ("efi/libstub/arm*: Pass latest memory map to the kernel")

fixed an issue in the memory map handling of the stub FDT code, but
inadvertently created an issue with such firmware, by moving some
of the FDT manipulation to after the invocation of ExitBootServices().

Given that the stub's libfdt implementation uses the ordinary, accelerated
string functions, which rely on hardware handling of unaligned accesses,
manipulating the FDT with the MMU off may result in alignment faults.

So fix the situation by moving the update_fdt_memmap() call into the
callback function invoked by efi_exit_boot_services() right before it
calls the ExitBootServices() UEFI service (which is arguably a better
place for it anyway)

Note that disabling the MMU in ExitBootServices() is not compliant with
the UEFI spec, and carries great risk due to the fact that switching from
cached to uncached memory accesses halfway through compiler generated code
(i.e., involving a stack) can never be done in a way that is architecturally
safe.

Fixes: abfb7b686a3e ("efi/libstub/arm*: Pass latest memory map to the kernel")
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Riku Voipio <riku.voipio@linaro.org>
Cc: mark.rutland@arm.com
Cc: linux-efi@vger.kernel.org
Cc: matt@codeblueprint.co.uk
Cc: leif.lindholm@linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1485971102-23330-2-git-send-email-ard.biesheuvel@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agox86/efi: Always map the first physical page into the EFI pagetables
Jiri Kosina [Fri, 27 Jan 2017 22:25:52 +0000 (22:25 +0000)]
x86/efi: Always map the first physical page into the EFI pagetables

commit bf29bddf0417a4783da3b24e8c9e017ac649326f upstream.

Commit:

  129766708 ("x86/efi: Only map RAM into EFI page tables if in mixed-mode")

stopped creating 1:1 mappings for all RAM, when running in native 64-bit mode.

It turns out though that there are 64-bit EFI implementations in the wild
(this particular problem has been reported on a Lenovo Yoga 710-11IKB),
which still make use of the first physical page for their own private use,
even though they explicitly mark it EFI_CONVENTIONAL_MEMORY in the memory
map.

In case there is no mapping for this particular frame in the EFI pagetables,
as soon as firmware tries to make use of it, a triple fault occurs and the
system reboots (in case of the Yoga 710-11IKB this is very early during bootup).

Fix that by always mapping the first page of physical memory into the EFI
pagetables. We're free to hand this page to the BIOS, as trim_bios_range()
will reserve the first page and isolate it away from memory allocators anyway.

Note that just reverting 129766708 alone is not enough on v4.9-rc1+ to fix the
regression on affected hardware, as this commit:

   ab72a27da ("x86/efi: Consolidate region mapping logic")

later made the first physical frame not to be mapped anyway.

Reported-by: Hanka Pavlikova <hanka@ucw.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bp@suse.de>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vojtech Pavlik <vojtech@ucw.cz>
Cc: Waiman Long <waiman.long@hpe.com>
Cc: linux-efi@vger.kernel.org
Fixes: 129766708 ("x86/efi: Only map RAM into EFI page tables if in mixed-mode")
Link: http://lkml.kernel.org/r/20170127222552.22336-1-matt@codeblueprint.co.uk
[ Tidied up the changelog and the comment. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoext4: validate s_first_meta_bg at mount time
Eryu Guan [Thu, 1 Dec 2016 20:08:37 +0000 (15:08 -0500)]
ext4: validate s_first_meta_bg at mount time

commit 3a4b77cd47bb837b8557595ec7425f281f2ca1fe upstream.

Ralf Spenneberg reported that he hit a kernel crash when mounting a
modified ext4 image. And it turns out that kernel crashed when
calculating fs overhead (ext4_calculate_overhead()), this is because
the image has very large s_first_meta_bg (debug code shows it's
842150400), and ext4 overruns the memory in count_overhead() when
setting bitmap buffer, which is PAGE_SIZE.

ext4_calculate_overhead():
  buf = get_zeroed_page(GFP_NOFS);  <=== PAGE_SIZE buffer
  blks = count_overhead(sb, i, buf);

count_overhead():
  for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) { <=== j = 842150400
          ext4_set_bit(EXT4_B2C(sbi, s++), buf);   <=== buffer overrun
          count++;
  }

This can be reproduced easily for me by this script:

  #!/bin/bash
  rm -f fs.img
  mkdir -p /mnt/ext4
  fallocate -l 16M fs.img
  mke2fs -t ext4 -O bigalloc,meta_bg,^resize_inode -F fs.img
  debugfs -w -R "ssv first_meta_bg 842150400" fs.img
  mount -o loop fs.img /mnt/ext4

Fix it by validating s_first_meta_bg first at mount time, and
refusing to mount if its value exceeds the largest possible meta_bg
number.

Reported-by: Ralf Spenneberg <ralf@os-t.de>
Signed-off-by: Eryu Guan <guaneryu@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoPCI/ASPM: Handle PCI-to-PCIe bridges as roots of PCIe hierarchies
Bjorn Helgaas [Fri, 27 Jan 2017 21:00:45 +0000 (15:00 -0600)]
PCI/ASPM: Handle PCI-to-PCIe bridges as roots of PCIe hierarchies

commit 030305d69fc6963c16003f50d7e8d74b02d0a143 upstream.

In a struct pcie_link_state, link->root points to the pcie_link_state of
the root of the PCIe hierarchy.  For the topmost link, this points to
itself (link->root = link).  For others, we copy the pointer from the
parent (link->root = link->parent->root).

Previously we recognized that Root Ports originated PCIe hierarchies, but
we treated PCI/PCI-X to PCIe Bridges as being in the middle of the
hierarchy, and when we tried to copy the pointer from link->parent->root,
there was no parent, and we dereferenced a NULL pointer:

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000090
  IP: [<ffffffff9e424350>] pcie_aspm_init_link_state+0x170/0x820

Recognize that PCI/PCI-X to PCIe Bridges originate PCIe hierarchies just
like Root Ports do, so link->root for these devices should also point to
itself.

Fixes: 51ebfc92b72b ("PCI: Enumerate switches below PCI-to-PCIe bridges")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=193411
Link: https://bugzilla.opensuse.org/show_bug.cgi?id=1022181
Tested-by: lists@ssl-mail.com
Tested-by: Jayachandran C. <jnair@caviumnetworks.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoLinux 4.9.8
Greg Kroah-Hartman [Sat, 4 Feb 2017 08:47:29 +0000 (09:47 +0100)]
Linux 4.9.8

7 years agoxfs: fix bmv_count confusion w/ shared extents
Darrick J. Wong [Thu, 2 Feb 2017 07:56:11 +0000 (08:56 +0100)]
xfs: fix bmv_count confusion w/ shared extents

commit c364b6d0b6cda1cd5d9ab689489adda3e82529aa upstream.

In a bmapx call, bmv_count is the total size of the array, including the
zeroth element that userspace uses to supply the search key.  The output
array starts at offset 1 so that we can set up the user for the next
invocation.  Since we now can split an extent into multiple bmap records
due to shared/unshared status, we have to be careful that we don't
overflow the output array.

In the original patch f86f403794b ("xfs: teach get_bmapx about shared
extents and the CoW fork") I used cur_ext (the output index) to check
for overflows, albeit with an off-by-one error.  Since nexleft no longer
describes the number of unfilled slots in the output, we can rip all
that out and use cur_ext for the overflow check directly.

Failure to do this causes heap corruption in bmapx callers such as
xfs_io and xfs_scrub.  xfs/328 can reproduce this problem.

Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: clear _XBF_PAGES from buffers when readahead page
Darrick J. Wong [Thu, 2 Feb 2017 07:56:10 +0000 (08:56 +0100)]
xfs: clear _XBF_PAGES from buffers when readahead page

commit 2aa6ba7b5ad3189cc27f14540aa2f57f0ed8df4b upstream.

If we try to allocate memory pages to back an xfs_buf that we're trying
to read, it's possible that we'll be so short on memory that the page
allocation fails.  For a blocking read we'll just wait, but for
readahead we simply dump all the pages we've collected so far.

Unfortunately, after dumping the pages we neglect to clear the
_XBF_PAGES state, which means that the subsequent call to xfs_buf_free
thinks that b_pages still points to pages we own.  It then double-frees
the b_pages pages.

This results in screaming about negative page refcounts from the memory
manager, which xfs oughtn't be triggering.  To reproduce this case,
mount a filesystem where the size of the inodes far outweighs the
availalble memory (a ~500M inode filesystem on a VM with 300MB memory
did the trick here) and run bulkstat in parallel with other memory
eating processes to put a huge load on the system.  The "check summary"
phase of xfs_scrub also works for this purpose.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: extsize hints are not unlikely in xfs_bmap_btalloc
Christoph Hellwig [Thu, 2 Feb 2017 07:56:09 +0000 (08:56 +0100)]
xfs: extsize hints are not unlikely in xfs_bmap_btalloc

commit 493611ebd62673f39e2f52c2561182c558a21cb6 upstream.

With COW files they are the hotpath, just like for files with the
extent size hint attribute.  We really shouldn't micro-manage anything
but failure cases with unlikely.

Additionally Arnd Bergmann recently reported that one of these two
unlikely annotations causes link failures together with an upcoming
kernel instrumentation patch, so let's get rid of it ASAP.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: remove racy hasattr check from attr ops
Brian Foster [Thu, 2 Feb 2017 07:56:08 +0000 (08:56 +0100)]
xfs: remove racy hasattr check from attr ops

commit 5a93790d4e2df73e30c965ec6e49be82fc3ccfce upstream.

xfs_attr_[get|remove]() have unlocked attribute fork checks to optimize
away a lock cycle in cases where the fork does not exist or is otherwise
empty. This check is not safe, however, because an attribute fork short
form to extent format conversion includes a transient state that causes
the xfs_inode_hasattr() check to fail. Specifically,
xfs_attr_shortform_to_leaf() creates an empty extent format attribute
fork and then adds the existing shortform attributes to it.

This means that lookup of an existing xattr can spuriously return
-ENOATTR when racing against a setxattr that causes the associated
format conversion. This was originally reproduced by an untar on a
particularly configured glusterfs volume, but can also be reproduced on
demand with properly crafted xattr requests.

The format conversion occurs under the exclusive ilock. xfs_attr_get()
and xfs_attr_remove() already have the proper locking and checks further
down in the functions to handle this situation correctly. Drop the
unlocked checks to avoid the spurious failure and rely on the existing
logic.

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: verify dirblocklog correctly
Darrick J. Wong [Thu, 2 Feb 2017 07:56:07 +0000 (08:56 +0100)]
xfs: verify dirblocklog correctly

commit 83d230eb5c638949350f4761acdfc0af5cb1bc00 upstream.

sb_dirblklog is added to sb_blocklog to compute the directory block size
in bytes.  Therefore, we must compare the sum of both those values
against XFS_MAX_BLOCKSIZE_LOG, not just dirblklog.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: fix COW writeback race
Christoph Hellwig [Thu, 2 Feb 2017 07:56:06 +0000 (08:56 +0100)]
xfs: fix COW writeback race

commit d2b3964a0780d2d2994eba57f950d6c9fe489ed8 upstream.

Due to the way how xfs_iomap_write_allocate tries to convert the whole
found extents from delalloc to real space we can run into a race
condition with multiple threads doing writes to this same extent.
For the non-COW case that is harmless as the only thing that can happen
is that we call xfs_bmapi_write on an extent that has already been
converted to a real allocation.  For COW writes where we move the extent
from the COW to the data fork after I/O completion the race is, however,
not quite as harmless.  In the worst case we are now calling
xfs_bmapi_write on a region that contains hole in the COW work, which
will trip up an assert in debug builds or lead to file system corruption
in non-debug builds.  This seems to be reproducible with workloads of
small O_DSYNC write, although so far I've not managed to come up with
a with an isolated reproducer.

The fix for the issue is relatively simple:  tell xfs_bmapi_write
that we are only asked to convert delayed allocations and skip holes
in that case.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: fix xfs_mode_to_ftype() prototype
Arnd Bergmann [Thu, 2 Feb 2017 07:56:05 +0000 (08:56 +0100)]
xfs: fix xfs_mode_to_ftype() prototype

commit fd29f7af75b7adf250beccffa63746c6a88e2b74 upstream.

A harmless warning just got introduced:

fs/xfs/libxfs/xfs_dir2.h:40:8: error: type qualifiers ignored on function return type [-Werror=ignored-qualifiers]

Removing the 'const' modifier avoids the warning and has no
other effect.

Fixes: 1fc4d33fed12 ("xfs: replace xfs_mode_to_ftype table with switch statement")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: don't wrap ID in xfs_dq_get_next_id
Eric Sandeen [Thu, 2 Feb 2017 07:56:04 +0000 (08:56 +0100)]
xfs: don't wrap ID in xfs_dq_get_next_id

commit 657bdfb7f5e68ca5e2ed009ab473c429b0d6af85 upstream.

The GETNEXTQOTA ioctl takes whatever ID is sent in,
and looks for the next active quota for an user
equal or higher to that ID.

But if we are at the maximum ID and then ask for the "next"
one, we may wrap back to zero.  In this case, userspace
may loop forever, because it will start querying again
at zero.

We'll fix this in userspace as well, but for the kernel,
return -ENOENT if we ask for the next quota ID
past UINT_MAX so the caller knows to stop.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: sanity check inode di_mode
Amir Goldstein [Thu, 2 Feb 2017 07:56:03 +0000 (08:56 +0100)]
xfs: sanity check inode di_mode

commit a324cbf10a3c67aaa10c9f47f7b5801562925bc2 upstream.

Check for invalid file type in xfs_dinode_verify()
and fail to load the inode structure from disk.

Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: sanity check inode mode when creating new dentry
Amir Goldstein [Thu, 2 Feb 2017 07:56:02 +0000 (08:56 +0100)]
xfs: sanity check inode mode when creating new dentry

commit fab8eef86c814c3dd46bc5d760b6e4a53d5fc5a6 upstream.

The helper xfs_dentry_to_name() is used by 2 different
classes of callers: Callers that pass zero mode and don't care
about the returned name.type field and Callers that pass
non zero mode and do care about the name.type field.

Change xfs_dentry_to_name() to not take the mode argument and
change the call sites of the first class to not pass the mode
argument.

Create a new helper xfs_dentry_mode_to_name() which does pass
the mode argument and returns -EFSCORRUPTED if mode is invalid.
Callers that translate non zero mode to on-disk file type now
check the return value and will export the error to user instead
of staging an invalid file type to be written to directory entry.

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: replace xfs_mode_to_ftype table with switch statement
Amir Goldstein [Thu, 2 Feb 2017 07:56:01 +0000 (08:56 +0100)]
xfs: replace xfs_mode_to_ftype table with switch statement

commit 1fc4d33fed124fb182e8e6c214e973a29389ae83.

The size of the xfs_mode_to_ftype[] conversion table
was too small to handle an invalid value of mode=S_IFMT.

Instead of fixing the table size, replace the conversion table
with a conversion helper that uses a switch statement.

Suggested-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: add missing include dependencies to xfs_dir2.h
Amir Goldstein [Thu, 2 Feb 2017 07:56:00 +0000 (08:56 +0100)]
xfs: add missing include dependencies to xfs_dir2.h

commit b597dd5373a1ccc08218665dc8417433b1c09550 upstream.

xfs_dir2.h dereferences some data types in inline functions
and fails to include those type definitions, e.g.:
xfs_dir2_data_aoff_t, struct xfs_da_geometry.

Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: sanity check directory inode di_size
Amir Goldstein [Thu, 2 Feb 2017 07:55:59 +0000 (08:55 +0100)]
xfs: sanity check directory inode di_size

commit 3c6f46eacd876bd723a9bad3c6882714c052fd8e upstream.

This changes fixes an assertion hit when fuzzing on-disk
i_mode values.

The easy case to fix is when changing an empty file
i_mode to S_IFDIR. In this case, xfs_dinode_verify()
detects an illegal zero size for directory and fails
to load the inode structure from disk.

For the case of non empty file whose i_mode is changed
to S_IFDIR, the ASSERT() statement in xfs_dir2_isblock()
is replaced with return -EFSCORRUPTED, to avoid interacting
with corrupted jusk also when XFS_DEBUG is disabled.

Suggested-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: make the ASSERT() condition likely
Amir Goldstein [Thu, 2 Feb 2017 07:55:58 +0000 (08:55 +0100)]
xfs: make the ASSERT() condition likely

commit bf46ecc3d8cca05f2907cf482755c42c2b11a79d upstream.

The ASSERT() condition is the normal case, not the exception,
so testing the condition should be likely(), not unlikely().

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: don't print warnings when xfs_log_force fails
Christoph Hellwig [Thu, 2 Feb 2017 07:55:57 +0000 (08:55 +0100)]
xfs: don't print warnings when xfs_log_force fails

commit 84a4620cfe97c9d57e39b2369bfb77faff55063d upstream.

There are only two reasons for xfs_log_force / xfs_log_force_lsn to fail:
one is an I/O error, for which xlog_bdstrat already logs a warning, and
the second is an already shutdown log due to a previous I/O errors.  In
the latter case we'll already have a previous indication for the actual
error, but the large stream of misleading warnings from xfs_log_force
will probably scroll it out of the message buffer.

Simply removing the warnings thus makes the XFS log reporting significantly
better.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Carlos Maiolino <cmaiolino@redhat.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: don't rely on ->total in xfs_alloc_space_available
Christoph Hellwig [Thu, 2 Feb 2017 07:55:56 +0000 (08:55 +0100)]
xfs: don't rely on ->total in xfs_alloc_space_available

commit 12ef830198b0d71668eb9b59f9ba69d32951a48a upstream.

->total is a bit of an odd parameter passed down to the low-level
allocator all the way from the high-level callers.  It's supposed to
contain the maximum number of blocks to be allocated for the whole
transaction [1].

But in xfs_iomap_write_allocate we only convert existing delayed
allocations and thus only have a minimal block reservation for the
current transaction, so xfs_alloc_space_available can't use it for
the allocation decisions.  Use the maximum of args->total and the
calculated block requirement to make a decision.  We probably should
get rid of args->total eventually and instead apply ->minleft more
broadly, but that will require some extensive changes all over.

[1] which creates lots of confusion as most callers don't decrement it
once doing a first allocation.  But that's for a separate series.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: adjust allocation length in xfs_alloc_space_available
Christoph Hellwig [Thu, 2 Feb 2017 07:55:55 +0000 (08:55 +0100)]
xfs: adjust allocation length in xfs_alloc_space_available

commit 54fee133ad59c87ab01dd84ab3e9397134b32acb upstream.

We must decide in xfs_alloc_fix_freelist if we can perform an
allocation from a given AG is possible or not based on the available
space, and should not fail the allocation past that point on a
healthy file system.

But currently we have two additional places that second-guess
xfs_alloc_fix_freelist: xfs_alloc_ag_vextent tries to adjust the
maxlen parameter to remove the reservation before doing the
allocation (but ignores the various minium freespace requirements),
and xfs_alloc_fix_minleft tries to fix up the allocated length
after we've found an extent, but ignores the reservations and also
doesn't take the AGFL into account (and thus fails allocations
for not matching minlen in some cases).

Remove all these later fixups and just correct the maxlen argument
inside xfs_alloc_fix_freelist once we have the AGF buffer locked.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: fix bogus minleft manipulations
Christoph Hellwig [Thu, 2 Feb 2017 07:55:54 +0000 (08:55 +0100)]
xfs: fix bogus minleft manipulations

commit 255c516278175a6dc7037d1406307f35237d8688 upstream.

We can't just set minleft to 0 when we're low on space - that's exactly
what we need minleft for: to protect space in the AG for btree block
allocations when we are low on free space.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoxfs: bump up reserved blocks in xfs_alloc_set_aside
Christoph Hellwig [Thu, 2 Feb 2017 07:55:53 +0000 (08:55 +0100)]
xfs: bump up reserved blocks in xfs_alloc_set_aside

commit 5149fd327f16e393c1d04fa5325ab072c32472bf upstream.

Setting aside 4 blocks globally for bmbt splits isn't all that useful,
as different threads can allocate space in parallel.  Bump it to 4
blocks per AG to allow each thread that is currently doing an
allocation to dip into it separately.  Without that we may no have
enough reserved blocks if there are enough parallel transactions
in an almost out space file system that all run into bmap btree
splits.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet: dsa: Bring back device detaching in dsa_slave_suspend()
Florian Fainelli [Wed, 25 Jan 2017 17:10:41 +0000 (09:10 -0800)]
net: dsa: Bring back device detaching in dsa_slave_suspend()

[ Upstream commit f154be241d22298d2b63c9b613f619fa1086ea75 ]

Commit 448b4482c671 ("net: dsa: Add lockdep class to tx queues to avoid
lockdep splat") removed the netif_device_detach() call done in
dsa_slave_suspend() which is necessary, and paired with a corresponding
netif_device_attach(), bring it back.

Fixes: 448b4482c671 ("net: dsa: Add lockdep class to tx queues to avoid lockdep splat")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agolwtunnel: Fix oops on state free after encap module unload
Robert Shearman [Tue, 24 Jan 2017 16:26:48 +0000 (16:26 +0000)]
lwtunnel: Fix oops on state free after encap module unload

[ Upstream commit 85c814016ce3b371016c2c054a905fa2492f5a65 ]

When attempting to free lwtunnel state after the module for the encap
has been unloaded an oops occurs:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
IP: lwtstate_free+0x18/0x40
[..]
task: ffff88003e372380 task.stack: ffffc900001fc000
RIP: 0010:lwtstate_free+0x18/0x40
RSP: 0018:ffff88003fd83e88 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff88002bbb3380 RCX: ffff88000c91a300
[..]
Call Trace:
 <IRQ>
 free_fib_info_rcu+0x195/0x1a0
 ? rt_fibinfo_free+0x50/0x50
 rcu_process_callbacks+0x2d3/0x850
 ? rcu_process_callbacks+0x296/0x850
 __do_softirq+0xe4/0x4cb
 irq_exit+0xb0/0xc0
 smp_apic_timer_interrupt+0x3d/0x50
 apic_timer_interrupt+0x93/0xa0
[..]
Code: e8 6e c6 fc ff 89 d8 5b 5d c3 bb de ff ff ff eb f4 66 90 66 66 66 66 90 55 48 89 e5 53 0f b7 07 48 89 fb 48 8b 04 c5 00 81 d5 81 <48> 8b 40 08 48 85 c0 74 13 ff d0 48 8d 7b 20 be 20 00 00 00 e8

The problem is after the module for the encap can be unloaded the
corresponding ops is removed and is thus NULL here.

Modules implementing lwtunnel ops should not be allowed to unload
while there is state alive using those ops, so grab the module
reference for the ops on creating lwtunnel state and of course release
the reference when freeing the state.

Fixes: 1104d9ba443a ("lwtunnel: Add destroy state operation")
Signed-off-by: Robert Shearman <rshearma@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet: Specify the owning module for lwtunnel ops
Robert Shearman [Tue, 24 Jan 2017 16:26:47 +0000 (16:26 +0000)]
net: Specify the owning module for lwtunnel ops

[ Upstream commit 88ff7334f25909802140e690c0e16433e485b0a0 ]

Modules implementing lwtunnel ops should not be allowed to unload
while there is state alive using those ops, so specify the owning
module for all lwtunnel ops.

Signed-off-by: Robert Shearman <rshearma@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoqmi_wwan/cdc_ether: add device ID for HP lt2523 (Novatel E371) WWAN card
Bjørn Mork [Tue, 24 Jan 2017 09:45:38 +0000 (10:45 +0100)]
qmi_wwan/cdc_ether: add device ID for HP lt2523 (Novatel E371) WWAN card

[ Upstream commit 5b9f57516337b523f7466a53939aaaea7b78141b ]

Another rebranded Novatel E371.  qmi_wwan should drive this device, while
cdc_ether should ignore it.  Even though the USB descriptors are plain
CDC-ETHER that USB interface is a QMI interface.  Ref commit 7fdb7846c9ca
("qmi_wwan/cdc_ether: add device IDs for Dell 5804 (Novatel E371) WWAN
card")

Cc: Dan Williams <dcbw@redhat.com>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoaf_unix: move unix_mknod() out of bindlock
WANG Cong [Mon, 23 Jan 2017 19:17:35 +0000 (11:17 -0800)]
af_unix: move unix_mknod() out of bindlock

[ Upstream commit 0fb44559ffd67de8517098b81f675fa0210f13f0 ]

Dmitry reported a deadlock scenario:

unix_bind() path:
u->bindlock ==> sb_writer

do_splice() path:
sb_writer ==> pipe->mutex ==> u->bindlock

In the unix_bind() code path, unix_mknod() does not have to
be done with u->bindlock held, since it is a pure fs operation,
so we can just move unix_mknod() out.

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Tested-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Rainer Weikusat <rweikusat@mobileactivedefense.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agor8152: don't execute runtime suspend if the tx is not empty
hayeswang [Mon, 23 Jan 2017 06:18:43 +0000 (14:18 +0800)]
r8152: don't execute runtime suspend if the tx is not empty

[ Upstream commit 6a0b76c04ec157c88ca943debf78a8ee58469f2d ]

Runtime suspend shouldn't be executed if the tx queue is not empty,
because the device is not idle.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet: mpls: Fix multipath selection for LSR use case
David Ahern [Fri, 20 Jan 2017 20:58:34 +0000 (12:58 -0800)]
net: mpls: Fix multipath selection for LSR use case

[ Upstream commit 9f427a0e474a67b454420c131709600d44850486 ]

MPLS multipath for LSR is broken -- always selecting the first nexthop
in the one label case. For example:

    $ ip -f mpls ro ls
    100
            nexthop as to 200 via inet 172.16.2.2  dev virt12
            nexthop as to 300 via inet 172.16.3.2  dev virt13
    101
            nexthop as to 201 via inet6 2000:2::2  dev virt12
            nexthop as to 301 via inet6 2000:3::2  dev virt13

In this example incoming packets have a single MPLS labels which means
BOS bit is set. The BOS bit is passed from mpls_forward down to
mpls_multipath_hash which never processes the hash loop because BOS is 1.

Update mpls_multipath_hash to process the entire label stack. mpls_hdr_len
tracks the total mpls header length on each pass (on pass N mpls_hdr_len
is N * sizeof(mpls_shim_hdr)). When the label is found with the BOS set
it verifies the skb has sufficient header for ipv4 or ipv6, and find the
IPv4 and IPv6 header by using the last mpls_hdr pointer and adding 1 to
advance past it.

With these changes I have verified the code correctly sees the label,
BOS, IPv4 and IPv6 addresses in the network header and icmp/tcp/udp
traffic for ipv4 and ipv6 are distributed across the nexthops.

Fixes: 1c78efa8319ca ("mpls: flow-based multipath selection")
Acked-by: Robert Shearman <rshearma@brocade.com>
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agobridge: netlink: call br_changelink() during br_dev_newlink()
Ivan Vecera [Fri, 20 Jan 2017 17:12:17 +0000 (18:12 +0100)]
bridge: netlink: call br_changelink() during br_dev_newlink()

[ Upstream commit b6677449dff674cf5b81429b11d5c7f358852ef9 ]

Any bridge options specified during link creation (e.g. ip link add)
are ignored as br_dev_newlink() does not process them.
Use br_changelink() to do it.

Fixes: 133235161721 ("bridge: implement rtnl_link_ops->changelink")
Signed-off-by: Ivan Vecera <cera@cera.cz>
Reviewed-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet/mlx5e: Do not recycle pages from emergency reserve
Eric Dumazet [Thu, 19 Jan 2017 07:03:08 +0000 (23:03 -0800)]
net/mlx5e: Do not recycle pages from emergency reserve

[ Upstream commit e048fc50d7bde23136e098e04a324d7e3404408d ]

A driver using dev_alloc_page() must not reuse a page allocated from
emergency memory reserve.

Otherwise all packets using this page will be immediately dropped,
unless for very specific sockets having SOCK_MEMALLOC bit set.

This issue might be hard to debug, because only a fraction of received
packets would be dropped.

Fixes: 4415a0319f92 ("net/mlx5e: Implement RX mapped page cache for page recycle")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Cc: Saeed Mahameed <saeedm@mellanox.com>
Acked-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agotcp: initialize max window for a new fastopen socket
Alexey Kodanev [Thu, 19 Jan 2017 13:36:39 +0000 (16:36 +0300)]
tcp: initialize max window for a new fastopen socket

[ Upstream commit 0dbd7ff3ac5017a46033a9d0a87a8267d69119d9 ]

Found that if we run LTP netstress test with large MSS (65K),
the first attempt from server to send data comparable to this
MSS on fastopen connection will be delayed by the probe timer.

Here is an example:

     < S  seq 0:0 win 43690 options [mss 65495 wscale 7 tfo cookie] length 32
     > S. seq 0:0 ack 1 win 43690 options [mss 65495 wscale 7] length 0
     < .  ack 1 win 342 length 0

Inside tcp_sendmsg(), tcp_send_mss() returns max MSS in 'mss_now',
as well as in 'size_goal'. This results the segment not queued for
transmition until all the data copied from user buffer. Then, inside
__tcp_push_pending_frames(), it breaks on send window test and
continues with the check probe timer.

Fragmentation occurs in tcp_write_wakeup()...

+0.2 > P. seq 1:43777 ack 1 win 342 length 43776
     < .  ack 43777, win 1365 length 0
     > P. seq 43777:65001 ack 1 win 342 options [...] length 21224
     ...

This also contradicts with the fact that we should bound to the half
of the window if it is large.

Fix this flaw by correctly initializing max_window. Before that, it
could have large values that affect further calculations of 'size_goal'.

Fixes: 168a8f58059a ("tcp: TCP Fast Open Server - main code path")
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoipv6: addrconf: Avoid addrconf_disable_change() using RCU read-side lock
Kefeng Wang [Thu, 19 Jan 2017 08:26:21 +0000 (16:26 +0800)]
ipv6: addrconf: Avoid addrconf_disable_change() using RCU read-side lock

[ Upstream commit 03e4deff4987f79c34112c5ba4eb195d4f9382b0 ]

Just like commit 4acd4945cd1e ("ipv6: addrconf: Avoid calling
netdevice notifiers with RCU read-side lock"), it is unnecessary
to make addrconf_disable_change() use RCU iteration over the
netdev list, since it already holds the RTNL lock, or we may meet
Illegal context switch in RCU read-side critical section.

Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agolwtunnel: fix autoload of lwt modules
David Ahern [Tue, 17 Jan 2017 22:57:36 +0000 (14:57 -0800)]
lwtunnel: fix autoload of lwt modules

[ Upstream commit 9ed59592e3e379b2e9557dc1d9e9ec8fcbb33f16]

Trying to add an mpls encap route when the MPLS modules are not loaded
hangs. For example:

    CONFIG_MPLS=y
    CONFIG_NET_MPLS_GSO=m
    CONFIG_MPLS_ROUTING=m
    CONFIG_MPLS_IPTUNNEL=m

    $ ip route add 10.10.10.10/32 encap mpls 100 via inet 10.100.1.2

The ip command hangs:
root       880   826  0 21:25 pts/0    00:00:00 ip route add 10.10.10.10/32 encap mpls 100 via inet 10.100.1.2

    $ cat /proc/880/stack
    [<ffffffff81065a9b>] call_usermodehelper_exec+0xd6/0x134
    [<ffffffff81065efc>] __request_module+0x27b/0x30a
    [<ffffffff814542f6>] lwtunnel_build_state+0xe4/0x178
    [<ffffffff814aa1e4>] fib_create_info+0x47f/0xdd4
    [<ffffffff814ae451>] fib_table_insert+0x90/0x41f
    [<ffffffff814a8010>] inet_rtm_newroute+0x4b/0x52
    ...

modprobe is trying to load rtnl-lwt-MPLS:

root       881     5  0 21:25 ?        00:00:00 /sbin/modprobe -q -- rtnl-lwt-MPLS

and it hangs after loading mpls_router:

    $ cat /proc/881/stack
    [<ffffffff81441537>] rtnl_lock+0x12/0x14
    [<ffffffff8142ca2a>] register_netdevice_notifier+0x16/0x179
    [<ffffffffa0033025>] mpls_init+0x25/0x1000 [mpls_router]
    [<ffffffff81000471>] do_one_initcall+0x8e/0x13f
    [<ffffffff81119961>] do_init_module+0x5a/0x1e5
    [<ffffffff810bd070>] load_module+0x13bd/0x17d6
    ...

The problem is that lwtunnel_build_state is called with rtnl lock
held preventing mpls_init from registering.

Given the potential references held by the time lwtunnel_build_state it
can not drop the rtnl lock to the load module. So, extract the module
loading code from lwtunnel_build_state into a new function to validate
the encap type. The new function is called while converting the user
request into a fib_config which is well before any table, device or
fib entries are examined.

Fixes: 745041e2aaf1 ("lwtunnel: autoload of lwt modules")
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet: phy: bcm63xx: Utilize correct config_intr function
Daniel Gonzalez Cabanelas [Wed, 18 Jan 2017 00:26:55 +0000 (16:26 -0800)]
net: phy: bcm63xx: Utilize correct config_intr function

[ Upstream commit cd33b3e0da43522ff8e8f2b2b71d3d08298512b0 ]

Commit a1cba5613edf ("net: phy: Add Broadcom phy library for common
interfaces") make the BCM63xx PHY driver utilize bcm_phy_config_intr()
which would appear to do the right thing, except that it does not write
to the MII_BCM63XX_IR register but to MII_BCM54XX_ECR which is
different.

This would be causing invalid link parameters and events from being
generated by the PHY interrupt.

Fixes: a1cba5613edf ("net: phy: Add Broadcom phy library for common interfaces")
Signed-off-by: Daniel Gonzalez Cabanelas <dgcbueu@gmail.com>
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet: fix harmonize_features() vs NETIF_F_HIGHDMA
Eric Dumazet [Wed, 18 Jan 2017 20:12:17 +0000 (12:12 -0800)]
net: fix harmonize_features() vs NETIF_F_HIGHDMA

[ Upstream commit 7be2c82cfd5d28d7adb66821a992604eb6dd112e ]

Ashizuka reported a highmem oddity and sent a patch for freescale
fec driver.

But the problem root cause is that core networking stack
must ensure no skb with highmem fragment is ever sent through
a device that does not assert NETIF_F_HIGHDMA in its features.

We need to call illegal_highdma() from harmonize_features()
regardless of CSUM checks.

Fixes: ec5f06156423 ("net: Kill link between CSUM and SG features.")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Pravin Shelar <pshelar@ovn.org>
Reported-by: "Ashizuka, Yuusuke" <ashiduka@jp.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agovxlan: fix byte order of vxlan-gpe port number
Lance Richardson [Mon, 16 Jan 2017 23:37:58 +0000 (18:37 -0500)]
vxlan: fix byte order of vxlan-gpe port number

[ Upstream commit d5ff72d9af73bc3cbaa3edb541333a851f8c7295 ]

vxlan->cfg.dst_port is in network byte order, so an htons()
is needed here. Also reduced comment length to stay closer
to 80 column width (still slightly over, however).

Fixes: e1e5314de08b ("vxlan: implement GPE")
Signed-off-by: Lance Richardson <lrichard@redhat.com>
Acked-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agovirtio-net: restore VIRTIO_HDR_F_DATA_VALID on receiving
Jason Wang [Fri, 20 Jan 2017 06:32:42 +0000 (14:32 +0800)]
virtio-net: restore VIRTIO_HDR_F_DATA_VALID on receiving

[ Upstream commit 6391a4481ba0796805d6581e42f9f0418c099e34 ]

Commit 501db511397f ("virtio: don't set VIRTIO_NET_HDR_F_DATA_VALID on
xmit") in fact disables VIRTIO_HDR_F_DATA_VALID on receiving path too,
fixing this by adding a hint (has_data_valid) and set it only on the
receiving path.

Cc: Rolf Neugebauer <rolf.neugebauer@docker.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Rolf Neugebauer <rolf.neugebauer@docker.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agovirtio: don't set VIRTIO_NET_HDR_F_DATA_VALID on xmit
Rolf Neugebauer [Tue, 17 Jan 2017 18:13:51 +0000 (18:13 +0000)]
virtio: don't set VIRTIO_NET_HDR_F_DATA_VALID on xmit

[ Upstream commit 501db511397fd6efff3aa5b4e8de415b55559550 ]

This patch part reverts fd2a0437dc33 and e858fae2b0b8 which introduced a
subtle change in how the virtio_net flags are derived from the SKBs
ip_summed field.

With the above commits, the flags are set to VIRTIO_NET_HDR_F_DATA_VALID
when ip_summed == CHECKSUM_UNNECESSARY, thus treating it differently to
ip_summed == CHECKSUM_NONE, which should be the same.

Further, the virtio spec 1.0 / CS04 explicitly says that
VIRTIO_NET_HDR_F_DATA_VALID must not be set by the driver.

Fixes: fd2a0437dc33 ("virtio_net: introduce virtio_net_hdr_{from,to}_skb")
Fixes: e858fae2b0b8 (" virtio_net: use common code for virtio_net_hdr and skb GSO conversion")
Signed-off-by: Rolf Neugebauer <rolf.neugebauer@docker.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet sched actions: fix refcnt when GETing of action after bind
Jamal Hadi Salim [Sun, 15 Jan 2017 15:14:06 +0000 (10:14 -0500)]
net sched actions: fix refcnt when GETing of action after bind

[ Upstream commit 0faa9cb5b3836a979864a6357e01d2046884ad52 ]

Demonstrating the issue:

.. add a drop action
$sudo $TC actions add action drop index 10

.. retrieve it
$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 2 bind 0 installed 29 sec used 29 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

... bug 1 above: reference is two.
    Reference is actually 1 but we forget to subtract 1.

... do a GET again and we see the same issue
    try a few times and nothing changes
~$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 2 bind 0 installed 31 sec used 31 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

... lets try to bind the action to a filter..
$ sudo $TC qdisc add dev lo ingress
$ sudo $TC filter add dev lo parent ffff: protocol ip prio 1 \
  u32 match ip dst 127.0.0.1/32 flowid 1:1 action gact index 10

... and now a few GETs:
$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 3 bind 1 installed 204 sec used 204 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 4 bind 1 installed 206 sec used 206 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 5 bind 1 installed 235 sec used 235 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

.... as can be observed the reference count keeps going up.

After the fix

$ sudo $TC actions add action drop index 10
$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 1 bind 0 installed 4 sec used 4 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 1 bind 0 installed 6 sec used 6 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

$ sudo $TC qdisc add dev lo ingress
$ sudo $TC filter add dev lo parent ffff: protocol ip prio 1 \
  u32 match ip dst 127.0.0.1/32 flowid 1:1 action gact index 10

$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 2 bind 1 installed 32 sec used 32 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

$ sudo $TC -s actions get action gact index 10

action order 1: gact action drop
 random type none pass val 0
 index 10 ref 2 bind 1 installed 33 sec used 33 sec
  Action statistics:
Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
backlog 0b 0p requeues 0

Fixes: aecc5cefc389 ("net sched actions: fix GETing actions")
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoax25: Fix segfault after sock connection timeout
Basil Gunn [Sat, 14 Jan 2017 20:18:55 +0000 (12:18 -0800)]
ax25: Fix segfault after sock connection timeout

[ Upstream commit 8a367e74c0120ef68c8c70d5a025648c96626dff ]

The ax.25 socket connection timed out & the sock struct has been
previously taken down ie. sock struct is now a NULL pointer. Checking
the sock_flag causes the segfault.  Check if the socket struct pointer
is NULL before checking sock_flag. This segfault is seen in
timed out netrom connections.

Please submit to -stable.

Signed-off-by: Basil Gunn <basil@pacabunga.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoip6_tunnel: Account for tunnel header in tunnel MTU
Jakub Sitnicki [Fri, 13 Jan 2017 09:12:20 +0000 (10:12 +0100)]
ip6_tunnel: Account for tunnel header in tunnel MTU

[ Upstream commit 02ca0423fd65a0a9c4d70da0dbb8f4b8503f08c7 ]

With ip6gre we have a tunnel header which also makes the tunnel MTU
smaller. We need to reserve room for it. Previously we were using up
space reserved for the Tunnel Encapsulation Limit option
header (RFC 2473).

Also, after commit b05229f44228 ("gre6: Cleanup GREv6 transmit path,
call common GRE functions") our contract with the caller has
changed. Now we check if the packet length exceeds the tunnel MTU after
the tunnel header has been pushed, unlike before.

This is reflected in the check where we look at the packet length minus
the size of the tunnel header, which is already accounted for in tunnel
MTU.

Fixes: b05229f44228 ("gre6: Cleanup GREv6 transmit path, call common GRE functions")
Signed-off-by: Jakub Sitnicki <jkbs@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoravb: do not use zero-length alignment DMA descriptor
Masaru Nagai [Mon, 16 Jan 2017 10:45:21 +0000 (11:45 +0100)]
ravb: do not use zero-length alignment DMA descriptor

[ Upstream commit 8ec3e8a192ba6f13be4522ee81227c792c86fb1a ]

Due to alignment requirements of the hardware transmissions are split into
two DMA descriptors, a small padding descriptor of 0 - 3 bytes in length
followed by a descriptor for rest of the packet.

In the case of IP packets the first descriptor will never be zero due to
the way that the stack aligns buffers for IP packets. However, for non-IP
packets it may be zero.

In that case it has been reported that timeouts occur, presumably because
transmission stops at the first zero-length DMA descriptor and thus the
packet is not transmitted. However, in my environment a BUG is triggered as
follows:

[   20.381417] ------------[ cut here ]------------
[   20.386054] kernel BUG at lib/swiotlb.c:495!
[   20.390324] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   20.395805] Modules linked in:
[   20.398862] CPU: 0 PID: 2089 Comm: mz Not tainted 4.10.0-rc3-00001-gf13ad2db193f #162
[   20.406689] Hardware name: Renesas Salvator-X board based on r8a7796 (DT)
[   20.413474] task: ffff80063b1f1900 task.stack: ffff80063a71c000
[   20.419404] PC is at swiotlb_tbl_map_single+0x178/0x2ec
[   20.424625] LR is at map_single+0x4c/0x98
[   20.428629] pc : [<ffff00000839c4c0>] lr : [<ffff00000839c680>] pstate: 800001c5
[   20.436019] sp : ffff80063a71f9b0
[   20.439327] x29: ffff80063a71f9b0 x28: ffff80063a20d500
[   20.444636] x27: ffff000008ed5000 x26: 0000000000000000
[   20.449944] x25: 000000067abe2adc x24: 0000000000000000
[   20.455252] x23: 0000000000200000 x22: 0000000000000001
[   20.460559] x21: 0000000000175ffe x20: ffff80063b2a0010
[   20.465866] x19: 0000000000000000 x18: 0000ffffcae6fb20
[   20.471173] x17: 0000ffffa09ba018 x16: ffff0000087c8b70
[   20.476480] x15: 0000ffffa084f588 x14: 0000ffffa09cfa14
[   20.481787] x13: 0000ffffcae87ff0 x12: 000000000063abe2
[   20.487098] x11: ffff000008096360 x10: ffff80063abe2adc
[   20.492407] x9 : 0000000000000000 x8 : 0000000000000000
[   20.497718] x7 : 0000000000000000 x6 : ffff000008ed50d0
[   20.503028] x5 : 0000000000000000 x4 : 0000000000000001
[   20.508338] x3 : 0000000000000000 x2 : 000000067abe2adc
[   20.513648] x1 : 00000000bafff000 x0 : 0000000000000000
[   20.518958]
[   20.520446] Process mz (pid: 2089, stack limit = 0xffff80063a71c000)
[   20.526798] Stack: (0xffff80063a71f9b0 to 0xffff80063a720000)
[   20.532543] f9a0:                                   ffff80063a71fa30 ffff00000839c680
[   20.540374] f9c0: ffff80063b2a0010 ffff80063b2a0010 0000000000000001 0000000000000000
[   20.548204] f9e0: 000000000000006e ffff80063b23c000 ffff80063b23c000 0000000000000000
[   20.556034] fa00: ffff80063b23c000 ffff80063a20d500 000000013b1f1900 0000000000000000
[   20.563864] fa20: ffff80063ffd18e0 ffff80063b2a0010 ffff80063a71fa60 ffff00000839cd10
[   20.571694] fa40: ffff80063b2a0010 0000000000000000 ffff80063ffd18e0 000000067abe2adc
[   20.579524] fa60: ffff80063a71fa90 ffff000008096380 ffff80063b2a0010 0000000000000000
[   20.587353] fa80: 0000000000000000 0000000000000001 ffff80063a71fac0 ffff00000864f770
[   20.595184] faa0: ffff80063b23caf0 0000000000000000 0000000000000000 0000000000000140
[   20.603014] fac0: ffff80063a71fb60 ffff0000087e6498 ffff80063a20d500 ffff80063b23c000
[   20.610843] fae0: 0000000000000000 ffff000008daeaf0 0000000000000000 ffff000008daeb00
[   20.618673] fb00: ffff80063a71fc0c ffff000008da7000 ffff80063b23c090 ffff80063a44f000
[   20.626503] fb20: 0000000000000000 ffff000008daeb00 ffff80063a71fc0c ffff000008da7000
[   20.634333] fb40: ffff80063b23c090 0000000000000000 ffff800600000037 ffff0000087e63d8
[   20.642163] fb60: ffff80063a71fbc0 ffff000008807510 ffff80063a692400 ffff80063a20d500
[   20.649993] fb80: ffff80063a44f000 ffff80063b23c000 ffff80063a69249c 0000000000000000
[   20.657823] fba0: 0000000000000000 ffff80063a087800 ffff80063b23c000 ffff80063a20d500
[   20.665653] fbc0: ffff80063a71fc10 ffff0000087e67dc ffff80063a20d500 ffff80063a692400
[   20.673483] fbe0: ffff80063b23c000 0000000000000000 ffff80063a44f000 ffff80063a69249c
[   20.681312] fc00: ffff80063a5f1a10 000000103a087800 ffff80063a71fc70 ffff0000087e6b24
[   20.689142] fc20: ffff80063a5f1a80 ffff80063a71fde8 000000000000000f 00000000000005ea
[   20.696972] fc40: ffff80063a5f1a10 0000000000000000 000000000000000f ffff00000887fbd0
[   20.704802] fc60: fffffff43a5f1a80 0000000000000000 ffff80063a71fc80 ffff000008880240
[   20.712632] fc80: ffff80063a71fd90 ffff0000087c7a34 ffff80063afc7180 0000000000000000
[   20.720462] fca0: 0000ffffcae6fe18 0000000000000014 0000000060000000 0000000000000015
[   20.728292] fcc0: 0000000000000123 00000000000000ce ffff0000088d2000 ffff80063b1f1900
[   20.736122] fce0: 0000000000008933 ffff000008e7cb80 ffff80063a71fd80 ffff0000087c50a4
[   20.743951] fd00: 0000000000008933 ffff000008e7cb80 ffff000008e7cb80 000000100000000e
[   20.751781] fd20: ffff80063a71fe4c 0000ffff00000300 0000000000000123 0000000000000000
[   20.759611] fd40: 0000000000000000 ffff80063b1f0000 000000000000000e 0000000000000300
[   20.767441] fd60: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
[   20.775271] fd80: 0000000000000000 0000000000000000 ffff80063a71fda0 ffff0000087c8c20
[   20.783100] fda0: 0000000000000000 ffff000008082f30 0000000000000000 0000800637260000
[   20.790930] fdc0: ffffffffffffffff 0000ffffa0903078 0000000000000000 000000001ea87232
[   20.798760] fde0: 000000000000000f ffff80063a71fe40 ffff800600000014 ffff000000000001
[   20.806590] fe00: 0000000000000000 0000000000000000 ffff80063a71fde8 0000000000000000
[   20.814420] fe20: 0000000000000000 0000000000000000 0000000000000000 0000000000000001
[   20.822249] fe40: 0000000203000011 0000000000000000 0000000000000000 ffff80063a68aa00
[   20.830079] fe60: ffff80063a68aa00 0000000000000003 0000000000008933 ffff0000081f1b9c
[   20.837909] fe80: 0000000000000000 ffff000008082f30 0000000000000000 0000800637260000
[   20.845739] fea0: ffffffffffffffff 0000ffffa07ca81c 0000000060000000 0000000000000015
[   20.853569] fec0: 0000000000000003 000000001ea87232 000000000000000f 0000000000000000
[   20.861399] fee0: 0000ffffcae6fe18 0000000000000014 0000000000000300 0000000000000000
[   20.869228] ff00: 00000000000000ce 0000000000000000 00000000ffffffff 0000000000000000
[   20.877059] ff20: 0000000000000002 0000ffffcae87ff0 0000ffffa09cfa14 0000ffffa084f588
[   20.884888] ff40: 0000000000000000 0000ffffa09ba018 0000ffffcae6fb20 000000001ea87010
[   20.892718] ff60: 0000ffffa09b9000 0000ffffcae6fe30 0000ffffcae6fe18 000000000000000f
[   20.900548] ff80: 0000000000000003 000000001ea87232 0000000000000000 0000000000000000
[   20.908378] ffa0: 0000000000000000 0000ffffcae6fdc0 0000ffffa09a7824 0000ffffcae6fdc0
[   20.916208] ffc0: 0000ffffa0903078 0000000060000000 0000000000000003 00000000000000ce
[   20.924038] ffe0: 0000000000000000 0000000000000000 ffffffffffffffff ffffffffffffffff
[   20.931867] Call trace:
[   20.934312] Exception stack(0xffff80063a71f7e0 to 0xffff80063a71f910)
[   20.940750] f7e0: 0000000000000000 0001000000000000 ffff80063a71f9b0 ffff00000839c4c0
[   20.948580] f800: ffff80063a71f840 ffff00000888a6e4 ffff80063a24c418 ffff80063a24c448
[   20.956410] f820: 0000000000000000 ffff00000811cd54 ffff80063a71f860 ffff80063a24c458
[   20.964240] f840: ffff80063a71f870 ffff00000888b258 ffff80063a24c418 0000000000000001
[   20.972070] f860: ffff80063a71f910 ffff80063a7b7028 ffff80063a71f890 ffff0000088825e4
[   20.979899] f880: 0000000000000000 00000000bafff000 000000067abe2adc 0000000000000000
[   20.987729] f8a0: 0000000000000001 0000000000000000 ffff000008ed50d0 0000000000000000
[   20.995560] f8c0: 0000000000000000 0000000000000000 ffff80063abe2adc ffff000008096360
[   21.003390] f8e0: 000000000063abe2 0000ffffcae87ff0 0000ffffa09cfa14 0000ffffa084f588
[   21.011219] f900: ffff0000087c8b70 0000ffffa09ba018
[   21.016097] [<ffff00000839c4c0>] swiotlb_tbl_map_single+0x178/0x2ec
[   21.022362] [<ffff00000839c680>] map_single+0x4c/0x98
[   21.027411] [<ffff00000839cd10>] swiotlb_map_page+0xa4/0x138
[   21.033072] [<ffff000008096380>] __swiotlb_map_page+0x20/0x7c
[   21.038821] [<ffff00000864f770>] ravb_start_xmit+0x174/0x668
[   21.044484] [<ffff0000087e6498>] dev_hard_start_xmit+0x8c/0x120
[   21.050407] [<ffff000008807510>] sch_direct_xmit+0x108/0x1a0
[   21.056064] [<ffff0000087e67dc>] __dev_queue_xmit+0x194/0x4cc
[   21.061807] [<ffff0000087e6b24>] dev_queue_xmit+0x10/0x18
[   21.067214] [<ffff000008880240>] packet_sendmsg+0xf40/0x1220
[   21.072873] [<ffff0000087c7a34>] sock_sendmsg+0x18/0x2c
[   21.078097] [<ffff0000087c8c20>] SyS_sendto+0xb0/0xf0
[   21.083150] [<ffff000008082f30>] el0_svc_naked+0x24/0x28
[   21.088462] Code: d34bfef7 2a1803f3 1a9f86d6 35fff878 (d4210000)
[   21.094611] ---[ end trace 5bc544ad491f3814 ]---
[   21.099234] Kernel panic - not syncing: Fatal exception in interrupt
[   21.105587] Kernel Offset: disabled
[   21.109073] Memory Limit: none
[   21.112126] ---[ end Kernel panic - not syncing: Fatal exception in interrupt

Fixes: 2f45d1902acf ("ravb: minimize TX data copying")
Signed-off-by: Masaru Nagai <masaru.nagai.vx@renesas.com
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomlx4: do not call napi_schedule() without care
Eric Dumazet [Fri, 13 Jan 2017 16:39:24 +0000 (08:39 -0800)]
mlx4: do not call napi_schedule() without care

[ Upstream commit 8cf699ec849f4ca1413cea01289bd7d37dbcc626 ]

Disable BH around the call to napi_schedule() to avoid following warning

[   52.095499] NOHZ: local_softirq_pending 08
[   52.421291] NOHZ: local_softirq_pending 08
[   52.608313] NOHZ: local_softirq_pending 08

Fixes: 8d59de8f7bb3 ("net/mlx4_en: Process all completions in RX rings after port goes up")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Erez Shitrit <erezsh@mellanox.com>
Cc: Eugenia Emantayev <eugenia@mellanox.com>
Cc: Tariq Toukan <tariqt@mellanox.com>
Acked-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agoopenvswitch: maintain correct checksum state in conntrack actions
Lance Richardson [Fri, 13 Jan 2017 00:33:18 +0000 (19:33 -0500)]
openvswitch: maintain correct checksum state in conntrack actions

[ Upstream commit 75f01a4c9cc291ff5cb28ca1216adb163b7a20ee ]

When executing conntrack actions on skbuffs with checksum mode
CHECKSUM_COMPLETE, the checksum must be updated to account for
header pushes and pulls. Otherwise we get "hw csum failure"
logs similar to this (ICMP packet received on geneve tunnel
via ixgbe NIC):

[  405.740065] genev_sys_6081: hw csum failure
[  405.740106] CPU: 3 PID: 0 Comm: swapper/3 Tainted: G          I     4.10.0-rc3+ #1
[  405.740108] Call Trace:
[  405.740110]  <IRQ>
[  405.740113]  dump_stack+0x63/0x87
[  405.740116]  netdev_rx_csum_fault+0x3a/0x40
[  405.740118]  __skb_checksum_complete+0xcf/0xe0
[  405.740120]  nf_ip_checksum+0xc8/0xf0
[  405.740124]  icmp_error+0x1de/0x351 [nf_conntrack_ipv4]
[  405.740132]  nf_conntrack_in+0xe1/0x550 [nf_conntrack]
[  405.740137]  ? find_bucket.isra.2+0x62/0x70 [openvswitch]
[  405.740143]  __ovs_ct_lookup+0x95/0x980 [openvswitch]
[  405.740145]  ? netif_rx_internal+0x44/0x110
[  405.740149]  ovs_ct_execute+0x147/0x4b0 [openvswitch]
[  405.740153]  do_execute_actions+0x22e/0xa70 [openvswitch]
[  405.740157]  ovs_execute_actions+0x40/0x120 [openvswitch]
[  405.740161]  ovs_dp_process_packet+0x84/0x120 [openvswitch]
[  405.740166]  ovs_vport_receive+0x73/0xd0 [openvswitch]
[  405.740168]  ? udp_rcv+0x1a/0x20
[  405.740170]  ? ip_local_deliver_finish+0x93/0x1e0
[  405.740172]  ? ip_local_deliver+0x6f/0xe0
[  405.740174]  ? ip_rcv_finish+0x3a0/0x3a0
[  405.740176]  ? ip_rcv_finish+0xdb/0x3a0
[  405.740177]  ? ip_rcv+0x2a7/0x400
[  405.740180]  ? __netif_receive_skb_core+0x970/0xa00
[  405.740185]  netdev_frame_hook+0xd3/0x160 [openvswitch]
[  405.740187]  __netif_receive_skb_core+0x1dc/0xa00
[  405.740194]  ? ixgbe_clean_rx_irq+0x46d/0xa20 [ixgbe]
[  405.740197]  __netif_receive_skb+0x18/0x60
[  405.740199]  netif_receive_skb_internal+0x40/0xb0
[  405.740201]  napi_gro_receive+0xcd/0x120
[  405.740204]  gro_cell_poll+0x57/0x80 [geneve]
[  405.740206]  net_rx_action+0x260/0x3c0
[  405.740209]  __do_softirq+0xc9/0x28c
[  405.740211]  irq_exit+0xd9/0xf0
[  405.740213]  do_IRQ+0x51/0xd0
[  405.740215]  common_interrupt+0x93/0x93

Fixes: 7f8a436eaa2c ("openvswitch: Add conntrack action")
Signed-off-by: Lance Richardson <lrichard@redhat.com>
Acked-by: Pravin B Shelar <pshelar@ovn.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agotcp: fix tcp_fastopen unaligned access complaints on sparc
Shannon Nelson [Thu, 12 Jan 2017 22:24:58 +0000 (14:24 -0800)]
tcp: fix tcp_fastopen unaligned access complaints on sparc

[ Upstream commit 003c941057eaa868ca6fedd29a274c863167230d ]

Fix up a data alignment issue on sparc by swapping the order
of the cookie byte array field with the length field in
struct tcp_fastopen_cookie, and making it a proper union
to clean up the typecasting.

This addresses log complaints like these:
    log_unaligned: 113 callbacks suppressed
    Kernel unaligned access at TPC[976490] tcp_try_fastopen+0x2d0/0x360
    Kernel unaligned access at TPC[9764ac] tcp_try_fastopen+0x2ec/0x360
    Kernel unaligned access at TPC[9764c8] tcp_try_fastopen+0x308/0x360
    Kernel unaligned access at TPC[9764e4] tcp_try_fastopen+0x324/0x360
    Kernel unaligned access at TPC[976490] tcp_try_fastopen+0x2d0/0x360

Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Shannon Nelson <shannon.nelson@oracle.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet: systemport: Decouple flow control from __bcm_sysport_tx_reclaim
Florian Fainelli [Thu, 12 Jan 2017 20:09:09 +0000 (12:09 -0800)]
net: systemport: Decouple flow control from __bcm_sysport_tx_reclaim

[ Upstream commit 148d3d021cf9724fcf189ce4e525a094bbf5ce89 ]

The __bcm_sysport_tx_reclaim() function is used to reclaim transmit
resources in different places within the driver. Most of them should
not affect the state of the transit flow control.

Introduce bcm_sysport_tx_clean() which cleans the ring, but does not
re-enable flow control towards the networking stack, and make
bcm_sysport_tx_reclaim() do the actual transmit queue flow control.

Fixes: 80105befdb4b ("net: systemport: add Broadcom SYSTEMPORT Ethernet MAC driver")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet: ipv4: fix table id in getroute response
David Ahern [Wed, 11 Jan 2017 23:42:17 +0000 (15:42 -0800)]
net: ipv4: fix table id in getroute response

[ Upstream commit 8a430ed50bb1b19ca14a46661f3b1b35f2fb5c39 ]

rtm_table is an 8-bit field while table ids are allowed up to u32. Commit
709772e6e065 ("net: Fix routing tables with id > 255 for legacy software")
added the preference to set rtm_table in dumps to RT_TABLE_COMPAT if the
table id is > 255. The table id returned on get route requests should do
the same.

Fixes: c36ba6603a11 ("net: Allow user to get table id from route lookup")
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonet: lwtunnel: Handle lwtunnel_fill_encap failure
David Ahern [Wed, 11 Jan 2017 22:29:54 +0000 (14:29 -0800)]
net: lwtunnel: Handle lwtunnel_fill_encap failure

[ Upstream commit ea7a80858f57d8878b1499ea0f1b8a635cc48de7 ]

Handle failure in lwtunnel_fill_encap adding attributes to skb.

Fixes: 571e722676fe ("ipv4: support for fib route lwtunnel encap attributes")
Fixes: 19e42e451506 ("ipv6: support for fib route lwtunnel encap attributes")
Signed-off-by: David Ahern <dsa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomlxsw: pci: Fix EQE structure definition
Elad Raz [Thu, 12 Jan 2017 08:10:39 +0000 (09:10 +0100)]
mlxsw: pci: Fix EQE structure definition

[ Upstream commit 28e46a0f2e03ab4ed0e23cace1ea89a68c8c115b ]

The event_data starts from address 0x00-0x0C and not from 0x08-0x014. This
leads to duplication with other fields in the Event Queue Element such as
sub-type, cqn and owner.

Fixes: eda6500a987a0 ("mlxsw: Add PCI bus implementation")
Signed-off-by: Elad Raz <eladr@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomlxsw: switchx2: Fix memory leak at skb reallocation
Arkadi Sharshevsky [Thu, 12 Jan 2017 08:10:38 +0000 (09:10 +0100)]
mlxsw: switchx2: Fix memory leak at skb reallocation

[ Upstream commit 400fc0106dd8c27ed84781c929c1a184785b9c79 ]

During transmission the skb is checked for headroom in order to
add vendor specific header. In case the skb needs to be re-allocated,
skb_realloc_headroom() is called to make a private copy of the original,
but doesn't release it. Current code assumes that the original skb is
released during reallocation and only releases it at the error path
which causes a memory leak.

Fix this by adding the original skb release to the main path.

Fixes: d003462a50de ("mlxsw: Simplify mlxsw_sx_port_xmit function")
Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agomlxsw: spectrum: Fix memory leak at skb reallocation
Arkadi Sharshevsky [Thu, 12 Jan 2017 08:10:37 +0000 (09:10 +0100)]
mlxsw: spectrum: Fix memory leak at skb reallocation

[ Upstream commit 36bf38d158d3482119b3e159c0619b3c1539b508 ]

During transmission the skb is checked for headroom in order to
add vendor specific header. In case the skb needs to be re-allocated,
skb_realloc_headroom() is called to make a private copy of the original,
but doesn't release it. Current code assumes that the original skb is
released during reallocation and only releases it at the error path
which causes a memory leak.

Fix this by adding the original skb release to the main path.

Fixes: 56ade8fe3fe1 ("mlxsw: spectrum: Add initial support for Spectrum ASIC")
Signed-off-by: Arkadi Sharshevsky <arkadis@mellanox.com>
Reviewed-by: Ido Schimmel <idosch@mellanox.com>
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agonetvsc: add rcu_read locking to netvsc callback
stephen hemminger [Wed, 11 Jan 2017 17:16:32 +0000 (09:16 -0800)]
netvsc: add rcu_read locking to netvsc callback

[ Upstream commit 0719e72ccb801829a3d735d187ca8417f0930459 ]

The receive callback (in tasklet context) is using RCU to get reference
to associated VF network device but this is not safe. RCU read lock
needs to be held. Found by running with full lockdep debugging
enabled.

Fixes: f207c10d9823 ("hv_netvsc: use RCU to protect vf_netdev")
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agor8152: fix the sw rx checksum is unavailable
hayeswang [Wed, 11 Jan 2017 08:25:34 +0000 (16:25 +0800)]
r8152: fix the sw rx checksum is unavailable

[ Upstream commit 19c0f40d4fca3a47b8f784a627f0467f0138ccc8 ]

Fix the hw rx checksum is always enabled, and the user couldn't switch
it to sw rx checksum.

Note that the RTL_VER_01 only support sw rx checksum only. Besides,
the hw rx checksum for RTL_VER_02 is disabled after
commit b9a321b48af4 ("r8152: Fix broken RX checksums."). Re-enable it.

Signed-off-by: Hayes Wang <hayeswang@realtek.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>