5 years agoMerge remote-tracking branches 'asoc/topic/cs35l35', 'asoc/topic/cs53l30', 'asoc...
Mark Brown [Sun, 30 Apr 2017 13:15:50 +0000 (22:15 +0900)]
Merge remote-tracking branches 'asoc/topic/cs35l35', 'asoc/topic/cs53l30', 'asoc/topic/da7213', 'asoc/topic/dio2125' and 'asoc/topic/dwc' into asoc-next

5 years agoMerge remote-tracking branches 'asoc/topic/adsp', 'asoc/topic/ak4613', 'asoc/topic...
Mark Brown [Sun, 30 Apr 2017 13:15:45 +0000 (22:15 +0900)]
Merge remote-tracking branches 'asoc/topic/adsp', 'asoc/topic/ak4613', 'asoc/topic/au1761' and 'asoc/topic/blackfin' into asoc-next

5 years agoMerge remote-tracking branch 'asoc/topic/rt5665' into asoc-next
Mark Brown [Sun, 30 Apr 2017 13:15:43 +0000 (22:15 +0900)]
Merge remote-tracking branch 'asoc/topic/rt5665' into asoc-next

5 years agoMerge remote-tracking branch 'asoc/topic/rcar' into asoc-next
Mark Brown [Sun, 30 Apr 2017 13:15:42 +0000 (22:15 +0900)]
Merge remote-tracking branch 'asoc/topic/rcar' into asoc-next

5 years agoMerge remote-tracking branch 'asoc/topic/intel' into asoc-next
Mark Brown [Sun, 30 Apr 2017 13:15:41 +0000 (22:15 +0900)]
Merge remote-tracking branch 'asoc/topic/intel' into asoc-next

5 years agoMerge remote-tracking branch 'asoc/topic/core' into asoc-next
Mark Brown [Sun, 30 Apr 2017 13:15:40 +0000 (22:15 +0900)]
Merge remote-tracking branch 'asoc/topic/core' into asoc-next

5 years agoMerge remote-tracking branch 'asoc/fix/cs4271' into asoc-linus
Mark Brown [Sun, 30 Apr 2017 13:15:37 +0000 (22:15 +0900)]
Merge remote-tracking branch 'asoc/fix/cs4271' into asoc-linus

5 years agoMerge remote-tracking branch 'asoc/fix/rcar' into asoc-linus
Mark Brown [Sun, 30 Apr 2017 13:15:36 +0000 (22:15 +0900)]
Merge remote-tracking branch 'asoc/fix/rcar' into asoc-linus

5 years agoMerge tag 'asoc-fix-v4.11-rc7' into asoc-linus
Mark Brown [Sun, 30 Apr 2017 13:15:35 +0000 (22:15 +0900)]
Merge tag 'asoc-fix-v4.11-rc7' into asoc-linus

ASoC: Fixes for v4.11

A few last minute fixes for v4.11, the STI fix is relatively large but
driver specific and has been cooking in -next for a little while now:

 - A fix from Takashi for some suspend/resume related crashes in the
   Intel drivers.
 - A fix from Mousumi Jana for issues with incorrectly created
   enumeration controls generated from topology files which could cause
   problems for userspace.
 - Fixes from Arnaud Pouliquen for some crashes due to races with the
   interrupt handler in the STI driver.

# gpg: Signature made Wed 26 Apr 2017 00:27:33 KST
# gpg:                using RSA key ADE668AA675718B59FE29FEA24D68B725D5487D0
# gpg:                issuer ""
# gpg: key 0D9EACE2CD7BEEBC: no public key for trusted key - skipped
# gpg: key 0D9EACE2CD7BEEBC marked as ultimately trusted
# gpg: key CCB0A420AF88CD16: no public key for trusted key - skipped
# gpg: key CCB0A420AF88CD16 marked as ultimately trusted
# gpg: key 162614E316005C11: no public key for trusted key - skipped
# gpg: key 162614E316005C11 marked as ultimately trusted
# gpg: key A730C53A5621E907: no public key for trusted key - skipped
# gpg: key A730C53A5621E907 marked as ultimately trusted
# gpg: key 276568D75C6153AD: no public key for trusted key - skipped
# gpg: key 276568D75C6153AD marked as ultimately trusted
# gpg: Good signature from "Mark Brown <>" [ultimate]
# gpg:                 aka "Mark Brown <>" [ultimate]
# gpg:                 aka "Mark Brown <>" [ultimate]
# gpg:                 aka "Mark Brown <>" [ultimate]
# gpg:                 aka "Mark Brown <>" [ultimate]
# gpg:                 aka "Mark Brown <>" [ultimate]

5 years agoASoC: Intel: Skylake: Return negative error code
Guneshwor Singh [Thu, 27 Apr 2017 06:51:22 +0000 (12:21 +0530)]
ASoC: Intel: Skylake: Return negative error code

skl_tplg_add_pipe() returned EEXIST instead of negative EEXIST, so fix that
and handle the return value as well.

Signed-off-by: Guneshwor Singh <>
Signed-off-by: Vinod Koul <>
Reviewed-by: Takashi Sakamoto <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Fix unused variable warning
Vinod Koul [Thu, 27 Apr 2017 06:51:21 +0000 (12:21 +0530)]
ASoC: Intel: Skylake: Fix unused variable warning

With compiler option W=1, we have one more warning in the driver for
'set but unused variable', so remove the unused variable to fix it.

sound/soc/intel/skylake/skl-pcm.c: In function ‘skl_platform_open’:
sound/soc/intel/skylake/skl-pcm.c:954:26: warning: variable ‘runtime’ set but not used [-Wunused-but-set-variable]
  struct snd_pcm_runtime *runtime;

Signed-off-by: Vinod Koul <>
Reviewed-by: Takashi Sakamoto <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: fix uninitialized pointer use
Arnd Bergmann [Thu, 27 Apr 2017 10:59:58 +0000 (12:59 +0200)]
ASoC: Intel: Skylake: fix uninitialized pointer use

The error handling in bxt_sst_dsp_init() got changed in a way that
it now derefences an uninitialized pointer when printing a warning
about the device not being found:

sound/soc/intel/skylake/bxt-sst.c: In function 'bxt_sst_dsp_init':
sound/soc/intel/skylake/bxt-sst.c:567:14: error: 'skl' may be used uninitialized in this function [-Werror=maybe-uninitialized]

As we do have a valid device pointer available at the call site,
let's use that instead.

Fixes: 9fe9c7119283 ("ASoC: Intel: Skylake: Move sst common initialization to a helper function")
Signed-off-by: Arnd Bergmann <>
Reviewed-by: Takashi Sakamoto <>
Signed-off-by: Mark Brown <>
5 years agoASoC: cs4271: configure reset GPIO as output
Alexander Sverdlin [Sat, 29 Apr 2017 10:19:33 +0000 (12:19 +0200)]
ASoC: cs4271: configure reset GPIO as output

During reset "refactoring" the output configuration was lost.
This commit repairs sound on EDB93XX boards.

Fixes: 9a397f4 ("ASoC: cs4271: add regulator consumer support")
Signed-off-by: Alexander Sverdlin <>
Signed-off-by: Mark Brown <>
Cc: <> # 4.6+
5 years agoASoC: dwc: Disallow building designware_pcm as a module
Jose Abreu [Fri, 28 Apr 2017 09:55:25 +0000 (10:55 +0100)]
ASoC: dwc: Disallow building designware_pcm as a module

Designware PCM is an extension to Designware I2S and they are dependent
on each other. For this reason, make Designware PCM a boolean which will
compile with Desigwnare I2S module. The name of the module is not changed
but the name of the files need to be changed.

Also, without this commit we get errors when probbing designware_i2s module
because of unspecified license:

designware_pcm: module license 'unspecified' taints kernel.
Disabling lock debugging due to kernel taint
designware_pcm: Unknown symbol __rcu_read_lock (err 0)
designware_pcm: Unknown symbol devm_snd_soc_register_platform (err 0)
designware_pcm: Unknown symbol synchronize_rcu (err 0)
designware_pcm: Unknown symbol __rcu_read_unlock (err 0)
designware_pcm: Unknown symbol snd_soc_set_runtime_hwparams (err 0)

So, this is really needed as a fix.

Fixes: 79361b2b98b7 ("ASoC: dwc: Add PIO PCM extension")
Signed-off-by: Lubomir Rintel <>
Signed-off-by: Jose Abreu <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Add loadable module support on KBL platform
Sodhi, VunnyX [Tue, 25 Apr 2017 06:48:24 +0000 (12:18 +0530)]
ASoC: Intel: Skylake: Add loadable module support on KBL platform

Kabylake platform expects modules in a library manifest. After loading
base firmware library manifest is loaded using load library IPC. This is
followed by module load using load multiple modules IPC.

Signed-off-by: Sodhi, VunnyX <>
Signed-off-by: G Kranthi <>
Signed-off-by: Subhransu S. Prusty <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Modify load_lib_ipc arguments for a nowait version
Subhransu S. Prusty [Tue, 25 Apr 2017 06:48:23 +0000 (12:18 +0530)]
ASoC: Intel: Skylake: Modify load_lib_ipc arguments for a nowait version

Kabylake uses code loader dma and wait on notification instead of ipc
reply for load library ipc status. So modify the argument of
skl_sst_ipc_load_library to check on flag to wait for ipc reply.

Signed-off-by: Subhransu S. Prusty <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Register dsp_fw_ops for kabylake
Subhransu S. Prusty [Tue, 25 Apr 2017 06:48:22 +0000 (12:18 +0530)]
ASoC: Intel: Skylake: Register dsp_fw_ops for kabylake

For audio kabylake is same as skylake except the module load approach.
This patch registers different dsp_fw_ops for kabylake and next patch
adds the module load support for kabylake.

Signed-off-by: Subhransu S. Prusty <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Modify arguments to reuse module transfer function
G Kranthi [Tue, 25 Apr 2017 06:48:21 +0000 (12:18 +0530)]
ASoC: Intel: Skylake: Modify arguments to reuse module transfer function

Kabylake also uses code loader dma for module load and library load.
skl_transfer_module can be reused. Modify the arguments to include
library index to be passed to lib load ipc and module/lib check to use
correct ipc for lib/module load.

Signed-off-by: G Kranthi <>
Signed-off-by: Subhransu S. Prusty <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Commonize library load
Subhransu S. Prusty [Tue, 25 Apr 2017 06:48:20 +0000 (12:18 +0530)]
ASoC: Intel: Skylake: Commonize library load

request firmware, strip extended manifest and release library changes
are common to kabylake and APL.

So move these common code to utils to be reused in later patches for
kabylake library load.

Signed-off-by: Subhransu S. Prusty <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Move sst common initialization to a helper function
G Kranthi [Tue, 25 Apr 2017 06:48:19 +0000 (12:18 +0530)]
ASoC: Intel: Skylake: Move sst common initialization to a helper function

Some skl sst context are not dependent of platform and initializing them
independently for each platform can lead to errors. So optimize by
moving them to a helper function and platform specific init code can
call this.

Signed-off-by: G Kranthi <>
Signed-off-by: Subhransu S. Prusty <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
5 years agoASoC: bytcr_rt5640: log quirk configuration errors
Pierre-Louis Bossart [Mon, 24 Apr 2017 21:34:33 +0000 (16:34 -0500)]
ASoC: bytcr_rt5640: log quirk configuration errors

Now that quirks can be overridden with a module parameter,
log errors so that non-sensical quirks introduced by mistake
are identified.

Signed-off-by: Pierre-Louis Bossart <>
Signed-off-by: Mark Brown <>
5 years agoASoC: bytcr_rt5640: Fix a typo and quirk parameter type
Takashi Iwai [Mon, 24 Apr 2017 21:34:30 +0000 (23:34 +0200)]
ASoC: bytcr_rt5640: Fix a typo and quirk parameter type

The previous patch for adding the quirk module option had a typo in
its info print, which results in a weird output.  Also, the parameter
type should be rather unsigned int instead of signed int.

Fixes: 9f2cf73ed65b ("ASoC: bytcr_rt5640: Allow quirk set via module option")
Reported-by: Pierre-Louis Bossart <>
Signed-off-by: Takashi Iwai <>
Signed-off-by: Mark Brown <>
5 years agoMerge remote-tracking branches 'asoc/fix/intel', 'asoc/fix/topology' and 'asoc/fix...
Mark Brown [Tue, 25 Apr 2017 15:25:07 +0000 (16:25 +0100)]
Merge remote-tracking branches 'asoc/fix/intel', 'asoc/fix/topology' and 'asoc/fix/sti' into asoc-linus

5 years agoASoC: intel: Fix PM and non-atomic crash in bytcr drivers
Takashi Iwai [Mon, 24 Apr 2017 12:09:55 +0000 (14:09 +0200)]
ASoC: intel: Fix PM and non-atomic crash in bytcr drivers

The FE setups of Intel SST bytcr_rt5640 and bytcr_rt5651 drivers carry
the ignore_suspend flag, and this prevents the suspend/resume working
properly while the stream is running, since SST core code has the
check of the running streams and returns -EBUSY.  Drop these
superfluous flags for fixing the behavior.

Also, the bytcr_rt5640 driver lacks of nonatomic flag in some FE
definitions, which leads to the kernel Oops at suspend/resume like:

  BUG: scheduling while atomic: systemd-sleep/3144/0x00000003
  Call Trace:
   ? del_timer_sync+0x50/0x50
   ? sst_wait_timeout+0xa9/0x170 [snd_intel_sst_core]
   ? sst_wait_timeout+0xa9/0x170 [snd_intel_sst_core]
   ? remove_wait_queue+0x60/0x60
   ? sst_prepare_and_post_msg+0x275/0x960 [snd_intel_sst_core]
   ? sst_pause_stream+0x9b/0x110 [snd_intel_sst_core]

This patch addresses these appropriately, too.

Signed-off-by: Takashi Iwai <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
Cc: <> # v4.1+
5 years agoASoC: bytcr_rt5640: Allow quirk set via module option
Takashi Iwai [Sun, 23 Apr 2017 07:22:56 +0000 (09:22 +0200)]
ASoC: bytcr_rt5640: Allow quirk set via module option

The bytcr-rt5640 driver has a few quirk setups depending on the board,
where the quirk value is set by DMI matching.  When you have a new
device to add the support, you often experience to try the different
quirk by trial-and-error.  Or, you may have a development model that
still has no proper DMI string.  In either case, you'd need to compile
the driver at each time.

This patch introduces a module option to override the quirk value on
the fly.  User can boot like snd-soc-sst-bytcr-rt5640.quirk=0x4004 to
override the default value without recompilation.  It's a raw value,
so user needs to check the source code for the meaning of each bit.

Signed-off-by: Takashi Iwai <>
Acked-by: Pierre-Louis Bossart <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Call snd_soc_set_dmi_name() unconditionally
Takashi Iwai [Mon, 24 Apr 2017 06:54:42 +0000 (08:54 +0200)]
ASoC: Call snd_soc_set_dmi_name() unconditionally

Since recently UCM can pick up a configuration specific to the board
via card longname field, and we introduced a helper function
snd_soc_set_dmi_name() for that.  So far, it was used only in one
place (sound/soc/intel/boards/broadwell.c), but it should be more
widely applied.

This patch puts a big hammer for that: it lets snd_soc_register_card()
calling snd_soc_set_dmi_name() unconditionally, so that all x86
devices get the better longname string.  This would have no impact for
other systems without DMI support, as snd_soc_set_dmi_name() is no-op
on them.

Signed-off-by: Takashi Iwai <>
Acked-by: Liam Girdwood <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Provide a dummy wrapper of snd_soc_set_dmi_name()
Takashi Iwai [Mon, 24 Apr 2017 06:54:41 +0000 (08:54 +0200)]
ASoC: Provide a dummy wrapper of snd_soc_set_dmi_name()

For systems without DMI, it makes no sense to have the code.

Signed-off-by: Takashi Iwai <>
Acked-by: Liam Girdwood <>
Acked-by: Vinod Koul <>
Signed-off-by: Mark Brown <>
5 years agoLinux 4.11-rc8 v4.11-rc8
Linus Torvalds [Sun, 23 Apr 2017 23:53:00 +0000 (16:53 -0700)]
Linux 4.11-rc8

5 years agoMerge tag 'upstream-4.11-rc7' of git://
Linus Torvalds [Sun, 23 Apr 2017 23:49:16 +0000 (16:49 -0700)]
Merge tag 'upstream-4.11-rc7' of git://

Pull UBI/UBIFS fixes from Richard Weinberger:
 "This contains fixes for issues in both UBI and UBIFS:

   - more O_TMPFILE fallout

   - RENAME_WHITEOUT regression due to a mis-merge

   - memory leak in ubifs_mknod()

   - power-cut problem in UBI's update volume feature"

* tag 'upstream-4.11-rc7' of git://
  ubifs: Fix O_TMPFILE corner case in ubifs_link()
  ubifs: Fix RENAME_WHITEOUT support
  ubifs: Fix debug messages for an invalid filename in ubifs_dump_inode
  ubifs: Fix debug messages for an invalid filename in ubifs_dump_node
  ubifs: Remove filename from debug messages in ubifs_readdir
  ubifs: Fix memory leak in error path in ubifs_mknod
  ubi/upd: Always flush after prepared for an update

5 years agoMerge branch 'ras-urgent-for-linus' of git://
Linus Torvalds [Sun, 23 Apr 2017 19:51:42 +0000 (12:51 -0700)]
Merge branch 'ras-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull RAS fix from Thomas Gleixner:
 "The MCE atomic notifier callchain invokes callbacks which might sleep.

  Convert it to a blocking notifier and prevent calls from atomic

* 'ras-urgent-for-linus' of git://
  x86/mce: Make the MCE notifier a blocking one

5 years agoMerge branch 'irq-urgent-for-linus' of git://
Linus Torvalds [Sun, 23 Apr 2017 19:48:05 +0000 (12:48 -0700)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fix from Thomas Gleixner:
 "The (hopefully) final fix for the irq affinity spreading logic"

* 'irq-urgent-for-linus' of git://
  genirq/affinity: Fix calculating vectors to assign

5 years agoMerge tag 'nfsd-4.11-2' of git://
Linus Torvalds [Fri, 21 Apr 2017 23:37:48 +0000 (16:37 -0700)]
Merge tag 'nfsd-4.11-2' of git://

Pull nfsd bugfix from Bruce Fields:
 "Fix a 4.11 regression that triggers a BUG() on an attempt to use an
  unsupported NFSv4 compound op"

* tag 'nfsd-4.11-2' of git://
  nfsd: fix oops on unsupported operation

5 years agoMerge git://
Linus Torvalds [Fri, 21 Apr 2017 20:20:47 +0000 (13:20 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Don't race in IPSEC dumps, from Yuejie Shi.

 2) Verify lengths properly in IPSEC reqeusts, from Herbert Xu.

 3) Fix out of bounds access in ipv6 segment routing code, from David

 4) Don't write into the header of cloned SKBs in smsc95xx driver, from
    James Hughes.

 5) Several other drivers have this bug too, fix them. From Eric

 6) Fix access to uninitialized data in TC action cookie code, from
    Wolfgang Bumiller.

 7) Fix double free in IPV6 segment routing, again from David Lebrun.

 8) Don't let userspace set the RTF_PCPU flag, oops. From David Ahern.

 9) Fix use after free in qrtr code, from Dan Carpenter.

10) Don't double-destroy devices in ip6mr code, from Nikolay

11) Don't pass out-of-range TX queue indices into drivers, from Tushar

* git:// (30 commits)
  netpoll: Check for skb->queue_mapping
  ip6mr: fix notification device destruction
  bpf, doc: update bpf maintainers entry
  net: qrtr: potential use after free in qrtr_sendmsg()
  bpf: Fix values type used in test_maps
  net: ipv6: RTF_PCPU should not be settable from userspace
  gso: Validate assumption of frag_list segementation
  kaweth: use skb_cow_head() to deal with cloned skbs
  ch9200: use skb_cow_head() to deal with cloned skbs
  lan78xx: use skb_cow_head() to deal with cloned skbs
  sr9700: use skb_cow_head() to deal with cloned skbs
  cx82310_eth: use skb_cow_head() to deal with cloned skbs
  smsc75xx: use skb_cow_head() to deal with cloned skbs
  ipv6: sr: fix double free of skb after handling invalid SRH
  MAINTAINERS: Add "B:" field for networking.
  net sched actions: allocate act cookie early
  qed: Fix issue in populating the PFC config paramters.
  qed: Fix possible system hang in the dcbnl-getdcbx() path.
  qed: Fix sending an invalid PFC error mask to MFW.
  qed: Fix possible error in populating max_tc field.

5 years agonetpoll: Check for skb->queue_mapping
Tushar Dave [Thu, 20 Apr 2017 22:57:31 +0000 (15:57 -0700)]
netpoll: Check for skb->queue_mapping

Reducing real_num_tx_queues needs to be in sync with skb queue_mapping
otherwise skbs with queue_mapping greater than real_num_tx_queues
can be sent to the underlying driver and can result in kernel panic.

One such event is running netconsole and enabling VF on the same
device. Or running netconsole and changing number of tx queues via
ethtool on same device.

Unable to handle kernel NULL pointer dereference
tsk->{mm,active_mm}->context = 0000000000001525
tsk->{mm,active_mm}->pgd = fff800130ff9a000
              \|/ ____ \|/
              "@'/ .. \`@"
              /_| \__/ |_\
kworker/48:1(475): Oops [#1]
CPU: 48 PID: 475 Comm: kworker/48:1 Tainted: G           OE
4.11.0-rc3-davem-net+ #7
Workqueue: events queue_process
task: fff80013113299c0 task.stack: fff800131132c000
TSTATE: 0000004480e01600 TPC: 00000000103f9e3c TNPC: 00000000103f9e40 Y:
00000000    Tainted: G           OE
TPC: <ixgbe_xmit_frame_ring+0x7c/0x6c0 [ixgbe]>
g0: 0000000000000000 g1: 0000000000003fff g2: 0000000000000000 g3:
g4: fff80013113299c0 g5: fff8001fa6808000 g6: fff800131132c000 g7:
o0: fff8001fa760c460 o1: fff8001311329a50 o2: fff8001fa7607504 o3:
o4: fff8001f96e63a40 o5: fff8001311d77ec0 sp: fff800131132f0e1 ret_pc:
RPC: <set_next_entity+0x34/0xb80>
l0: 0000000000000000 l1: 0000000000000800 l2: 0000000000000000 l3:
l4: 000b2aa30e34b10d l5: 0000000000000000 l6: 0000000000000000 l7:
i0: fff80013111a8a00 i1: fff80013155a0780 i2: 0000000000000000 i3:
i4: 0000000000000000 i5: 0000000000100000 i6: fff800131132f1a1 i7:
I7: <ixgbe_xmit_frame+0x30/0xa0 [ixgbe]>
Call Trace:
 [00000000103fa4b0] ixgbe_xmit_frame+0x30/0xa0 [ixgbe]
 [0000000000998c74] netpoll_start_xmit+0xf4/0x200
 [0000000000998e10] queue_process+0x90/0x160
 [0000000000485fa8] process_one_work+0x188/0x480
 [0000000000486410] worker_thread+0x170/0x4c0
 [000000000048c6b8] kthread+0xd8/0x120
 [0000000000406064] ret_from_fork+0x1c/0x2c
 [0000000000000000]           (null)
Disabling lock debugging due to kernel taint
Caller[00000000103fa4b0]: ixgbe_xmit_frame+0x30/0xa0 [ixgbe]
Caller[0000000000998c74]: netpoll_start_xmit+0xf4/0x200
Caller[0000000000998e10]: queue_process+0x90/0x160
Caller[0000000000485fa8]: process_one_work+0x188/0x480
Caller[0000000000486410]: worker_thread+0x170/0x4c0
Caller[000000000048c6b8]: kthread+0xd8/0x120
Caller[0000000000406064]: ret_from_fork+0x1c/0x2c
Caller[0000000000000000]:           (null)

Signed-off-by: Tushar Dave <>
Signed-off-by: David S. Miller <>
5 years agoip6mr: fix notification device destruction
Nikolay Aleksandrov [Fri, 21 Apr 2017 17:42:16 +0000 (20:42 +0300)]
ip6mr: fix notification device destruction

Andrey Konovalov reported a BUG caused by the ip6mr code which is caused
because we call unregister_netdevice_many for a device that is already
being destroyed. In IPv4's ipmr that has been resolved by two commits
long time ago by introducing the "notify" parameter to the delete
function and avoiding the unregister when called from a notifier, so
let's do the same for ip6mr.

The trace from Andrey:
------------[ cut here ]------------
kernel BUG at net/core/dev.c:6813!
invalid opcode: 0000 [#1] SMP KASAN
Modules linked in:
CPU: 1 PID: 1165 Comm: kworker/u4:3 Not tainted 4.11.0-rc7+ #251
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs
Workqueue: netns cleanup_net
task: ffff880069208000 task.stack: ffff8800692d8000
RIP: 0010:rollback_registered_many+0x348/0xeb0 net/core/dev.c:6813
RSP: 0018:ffff8800692de7f0 EFLAGS: 00010297
RAX: ffff880069208000 RBX: 0000000000000002 RCX: 0000000000000001
RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff88006af90569
RBP: ffff8800692de9f0 R08: ffff8800692dec60 R09: 0000000000000000
R10: 0000000000000006 R11: 0000000000000000 R12: ffff88006af90070
R13: ffff8800692debf0 R14: dffffc0000000000 R15: ffff88006af90000
FS:  0000000000000000(0000) GS:ffff88006cb00000(0000)
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007fe7e897d870 CR3: 00000000657e7000 CR4: 00000000000006e0
Call Trace:
 unregister_netdevice_many.part.105+0x87/0x440 net/core/dev.c:7881
 unregister_netdevice_many+0xc8/0x120 net/core/dev.c:7880
 ip6mr_device_event+0x362/0x3f0 net/ipv6/ip6mr.c:1346
 notifier_call_chain+0x145/0x2f0 kernel/notifier.c:93
 __raw_notifier_call_chain kernel/notifier.c:394
 raw_notifier_call_chain+0x2d/0x40 kernel/notifier.c:401
 call_netdevice_notifiers_info+0x51/0x90 net/core/dev.c:1647
 call_netdevice_notifiers net/core/dev.c:1663
 rollback_registered_many+0x919/0xeb0 net/core/dev.c:6841
 unregister_netdevice_many.part.105+0x87/0x440 net/core/dev.c:7881
 unregister_netdevice_many net/core/dev.c:7880
 default_device_exit_batch+0x4fa/0x640 net/core/dev.c:8333
 ops_exit_list.isra.4+0x100/0x150 net/core/net_namespace.c:144
 cleanup_net+0x5a8/0xb40 net/core/net_namespace.c:463
 process_one_work+0xc04/0x1c10 kernel/workqueue.c:2097
 worker_thread+0x223/0x19c0 kernel/workqueue.c:2231
 kthread+0x35e/0x430 kernel/kthread.c:231
 ret_from_fork+0x31/0x40 arch/x86/entry/entry_64.S:430
Code: 3c 32 00 0f 85 70 0b 00 00 48 b8 00 02 00 00 00 00 ad de 49 89
47 78 e9 93 fe ff ff 49 8d 57 70 49 8d 5f 78 eb 9e e8 88 7a 14 fe <0f>
0b 48 8b 9d 28 fe ff ff e8 7a 7a 14 fe 48 b8 00 00 00 00 00
RIP: rollback_registered_many+0x348/0xeb0 RSP: ffff8800692de7f0
---[ end trace e0b29c57e9b3292c ]---

Reported-by: Andrey Konovalov <>
Signed-off-by: Nikolay Aleksandrov <>
Tested-by: Andrey Konovalov <>
Signed-off-by: David S. Miller <>
5 years agobpf, doc: update bpf maintainers entry
Daniel Borkmann [Thu, 20 Apr 2017 15:27:58 +0000 (17:27 +0200)]
bpf, doc: update bpf maintainers entry

Add various related files that have been missing under
BPF entry covering essential parts of its infrastructure
and also add myself as co-maintainer.

Signed-off-by: Daniel Borkmann <>
Acked-by: Alexei Starovoitov <>
Signed-off-by: David S. Miller <>
5 years agonet: qrtr: potential use after free in qrtr_sendmsg()
Dan Carpenter [Thu, 20 Apr 2017 10:21:30 +0000 (13:21 +0300)]
net: qrtr: potential use after free in qrtr_sendmsg()

If skb_pad() fails then it frees the skb so we should check for errors.

Fixes: bdabad3e363d ("net: Add Qualcomm IPC router")
Signed-off-by: Dan Carpenter <>
Signed-off-by: David S. Miller <>
5 years agobpf: Fix values type used in test_maps
David Miller [Thu, 20 Apr 2017 19:20:16 +0000 (15:20 -0400)]
bpf: Fix values type used in test_maps

Maps of per-cpu type have their value element size adjusted to 8 if it
is specified smaller during various map operations.

This makes test_maps as a 32-bit binary fail, in fact the kernel
writes past the end of the value's array on the user's stack.

To be quite honest, I think the kernel should reject creation of a
per-cpu map that doesn't have a value size of at least 8 if that's
what the kernel is going to silently adjust to later.

If the user passed something smaller, it is a sizeof() calcualtion
based upon the type they will actually use (just like in this testcase
code) in later calls to the map operations.

Fixes: df570f577231 ("samples/bpf: unit test for BPF_MAP_TYPE_PERCPU_ARRAY")
Signed-off-by: David S. Miller <>
Acked-by: Daniel Borkmann <>
Acked-by: Alexei Starovoitov <>
5 years agonet: ipv6: RTF_PCPU should not be settable from userspace
David Ahern [Wed, 19 Apr 2017 21:19:43 +0000 (14:19 -0700)]
net: ipv6: RTF_PCPU should not be settable from userspace

Andrey reported a fault in the IPv6 route code:

kasan: GPF could be caused by NULL-ptr deref or user memory access
general protection fault: 0000 [#1] SMP KASAN
Modules linked in:
CPU: 1 PID: 4035 Comm: a.out Not tainted 4.11.0-rc7+ #250
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
task: ffff880069809600 task.stack: ffff880062dc8000
RIP: 0010:ip6_rt_cache_alloc+0xa6/0x560 net/ipv6/route.c:975
RSP: 0018:ffff880062dced30 EFLAGS: 00010206
RAX: dffffc0000000000 RBX: ffff8800670561c0 RCX: 0000000000000006
RDX: 0000000000000003 RSI: ffff880062dcfb28 RDI: 0000000000000018
RBP: ffff880062dced68 R08: 0000000000000001 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
R13: ffff880062dcfb28 R14: dffffc0000000000 R15: 0000000000000000
FS:  00007feebe37e7c0(0000) GS:ffff88006cb00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000205a0fe4 CR3: 000000006b5c9000 CR4: 00000000000006e0
Call Trace:
 ip6_pol_route+0x1512/0x1f20 net/ipv6/route.c:1128
 ip6_pol_route_output+0x4c/0x60 net/ipv6/route.c:1212

Andrey's syzkaller program passes rtmsg.rtmsg_flags with the RTF_PCPU bit
set. Flags passed to the kernel are blindly copied to the allocated
rt6_info by ip6_route_info_create making a newly inserted route appear
as though it is a per-cpu route. ip6_rt_cache_alloc sees the flag set
and expects rt->dst.from to be set - which it is not since it is not
really a per-cpu copy. The subsequent call to __ip6_dst_alloc then
generates the fault.

Fix by checking for the flag and failing with EINVAL.

Fixes: d52d3997f843f ("ipv6: Create percpu rt6_info")
Reported-by: Andrey Konovalov <>
Signed-off-by: David Ahern <>
Acked-by: Martin KaFai Lau <>
Tested-by: Andrey Konovalov <>
Signed-off-by: David S. Miller <>
5 years agogso: Validate assumption of frag_list segementation
Ilan Tayari [Wed, 19 Apr 2017 18:26:07 +0000 (21:26 +0300)]
gso: Validate assumption of frag_list segementation

Commit 07b26c9454a2 ("gso: Support partial splitting at the frag_list
pointer") assumes that all SKBs in a frag_list (except maybe the last
one) contain the same amount of GSO payload.

This assumption is not always correct, resulting in the following
warning message in the log:
    skb_segment: too many frags

For example, mlx5 driver in Striding RQ mode creates some RX SKBs with
one frag, and some with 2 frags.
After GRO, the frag_list SKBs end up having different amounts of payload.
If this frag_list SKB is then forwarded, the aforementioned assumption
is violated.

Validate the assumption, and fall back to software GSO if it not true.

Change-Id: Ia03983f4a47b6534dd987d7a2aad96d54d46d212
Fixes: 07b26c9454a2 ("gso: Support partial splitting at the frag_list pointer")
Signed-off-by: Ilan Tayari <>
Signed-off-by: Ilya Lesokhin <>
Signed-off-by: David S. Miller <>
5 years agoMerge branch 'skb_cow_head'
David S. Miller [Fri, 21 Apr 2017 17:24:07 +0000 (13:24 -0400)]
Merge branch 'skb_cow_head'

Eric Dumazet says:

net: use skb_cow_head() to deal with cloned skbs

James Hughes found an issue with smsc95xx driver. Same problematic code
is found in other drivers.

Signed-off-by: David S. Miller <>
5 years agokaweth: use skb_cow_head() to deal with cloned skbs
Eric Dumazet [Wed, 19 Apr 2017 16:59:26 +0000 (09:59 -0700)]
kaweth: use skb_cow_head() to deal with cloned skbs

We can use skb_cow_head() to properly deal with clones,
especially the ones coming from TCP stack that allow their head being
modified. This avoids a copy.

Signed-off-by: Eric Dumazet <>
Cc: James Hughes <>
Signed-off-by: David S. Miller <>
5 years agoch9200: use skb_cow_head() to deal with cloned skbs
Eric Dumazet [Wed, 19 Apr 2017 16:59:25 +0000 (09:59 -0700)]
ch9200: use skb_cow_head() to deal with cloned skbs

We need to ensure there is enough headroom to push extra header,
but we also need to check if we are allowed to change headers.

skb_cow_head() is the proper helper to deal with this.

Fixes: 4a476bd6d1d9 ("usbnet: New driver for QinHeng CH9200 devices")
Signed-off-by: Eric Dumazet <>
Cc: James Hughes <>
Cc: Matthew Garrett <>
Signed-off-by: David S. Miller <>
5 years agolan78xx: use skb_cow_head() to deal with cloned skbs
Eric Dumazet [Wed, 19 Apr 2017 16:59:24 +0000 (09:59 -0700)]
lan78xx: use skb_cow_head() to deal with cloned skbs

We need to ensure there is enough headroom to push extra header,
but we also need to check if we are allowed to change headers.

skb_cow_head() is the proper helper to deal with this.

Fixes: 55d7de9de6c3 ("Microchip's LAN7800 family USB 2/3 to 10/100/1000 Ethernet device driver")
Signed-off-by: Eric Dumazet <>
Cc: James Hughes <>
Cc: Woojung Huh <>
Signed-off-by: David S. Miller <>
5 years agosr9700: use skb_cow_head() to deal with cloned skbs
Eric Dumazet [Wed, 19 Apr 2017 16:59:23 +0000 (09:59 -0700)]
sr9700: use skb_cow_head() to deal with cloned skbs

We need to ensure there is enough headroom to push extra header,
but we also need to check if we are allowed to change headers.

skb_cow_head() is the proper helper to deal with this.

Fixes: c9b37458e956 ("USB2NET : SR9700 : One chip USB 1.1 USB2NET SR9700Device Driver Support")
Signed-off-by: Eric Dumazet <>
Cc: James Hughes <>
Signed-off-by: David S. Miller <>
5 years agocx82310_eth: use skb_cow_head() to deal with cloned skbs
Eric Dumazet [Wed, 19 Apr 2017 16:59:22 +0000 (09:59 -0700)]
cx82310_eth: use skb_cow_head() to deal with cloned skbs

We need to ensure there is enough headroom to push extra header,
but we also need to check if we are allowed to change headers.

skb_cow_head() is the proper helper to deal with this.

Fixes: cc28a20e77b2 ("introduce cx82310_eth: Conexant CX82310-based ADSL router USB ethernet driver")
Signed-off-by: Eric Dumazet <>
Cc: James Hughes <>
Signed-off-by: David S. Miller <>
5 years agosmsc75xx: use skb_cow_head() to deal with cloned skbs
Eric Dumazet [Wed, 19 Apr 2017 16:59:21 +0000 (09:59 -0700)]
smsc75xx: use skb_cow_head() to deal with cloned skbs

We need to ensure there is enough headroom to push extra header,
but we also need to check if we are allowed to change headers.

skb_cow_head() is the proper helper to deal with this.

Fixes: d0cad871703b ("smsc75xx: SMSC LAN75xx USB gigabit ethernet adapter driver")
Signed-off-by: Eric Dumazet <>
Cc: James Hughes <>
Signed-off-by: David S. Miller <>
5 years agoipv6: sr: fix double free of skb after handling invalid SRH
David Lebrun [Wed, 19 Apr 2017 14:10:19 +0000 (16:10 +0200)]
ipv6: sr: fix double free of skb after handling invalid SRH

The icmpv6_param_prob() function already does a kfree_skb(),
this patch removes the duplicate one.

Fixes: 1ababeba4a21f3dba3da3523c670b207fb2feb62 ("ipv6: implement dataplane support for rthdr type 4 (Segment Routing Header)")
Reported-by: Dan Carpenter <>
Cc: Dan Carpenter <>
Signed-off-by: David Lebrun <>
Signed-off-by: David S. Miller <>
5 years agoASoC: cs35l35: Allow user to configure IMON SCALE
Charles Keepax [Thu, 13 Apr 2017 15:52:09 +0000 (16:52 +0100)]
ASoC: cs35l35: Allow user to configure IMON SCALE

On the chip the IMON signal is a full 24-bits however normally only
some of the bits will be sent over the bus. The chip provides a field
to select which bits of the IMON will be sent back, this is the only
feedback signal that has this feature.

Add an additional entry to the cirrus,imon device tree property to
allow the IMON scale parameter to be passed.

Signed-off-by: Charles Keepax <>
Acked-by: Brian Austin <>
Acked-by: Rob Herring <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Fix a couple user after free bugs
Dan Carpenter [Fri, 14 Apr 2017 19:11:20 +0000 (22:11 +0300)]
ASoC: Intel: Skylake: Fix a couple user after free bugs

We need to use the _safe() version of list_for_each_entry() here because
of the kfree(modules).

Fixes: b8c722ddd548 ("ASoC: Intel: Skylake: Add support for deferred DSP module bind")
Signed-off-by: Dan Carpenter <>
Signed-off-by: Mark Brown <>
5 years agoASoC: Intel: Skylake: Uninitialized variable in probe_codec()
Dan Carpenter [Thu, 20 Apr 2017 10:17:02 +0000 (13:17 +0300)]
ASoC: Intel: Skylake: Uninitialized variable in probe_codec()

My static checker complains that if snd_hdac_bus_get_response() returns
-EIO then "res" is uninitialized.  Fix this by initializing it to -1 so
that the error is handled correctly.

Fixes: d8c2dab8381d ("ASoC: Intel: Add Skylake HDA audio driver")
Signed-off-by: Dan Carpenter <>
Signed-off-by: Mark Brown <>
5 years agoASoC: rsnd: Fix possible NULL pointer dereference
Marek Vasut [Fri, 21 Apr 2017 00:41:20 +0000 (00:41 +0000)]
ASoC: rsnd: Fix possible NULL pointer dereference

("ASoC: rsnd: enable clock-frequency for both 44.1kHz/48kHz")
supports both 44.1kHz/48kHz clock-frequency settings for ADG
which will be used for AUDIO_OLKOUTn.
But some board doesn't need it, thus, it is not mandatory.

But, above patch didn't care about the case of "clock-frequency" DT
property was not present.
This patch ignores ADG settings if AUDIO_OLKOUTn was not used.

Signed-off-by: Marek Vasut <>
[Kuninori: tidyup not to break non AUDIO_OLKOUTn case]
Signed-off-by: Kuninori Morimoto <>
Signed-off-by: Mark Brown <>
5 years agoMerge tag 'powerpc-4.11-8' of git://
Linus Torvalds [Fri, 21 Apr 2017 16:34:45 +0000 (09:34 -0700)]
Merge tag 'powerpc-4.11-8' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "Just two fixes.

  The first fixes kprobing a stdu, and is marked for stable as it's been
  broken for ~ever. In hindsight this could have gone in next.

  The other is a fix for a change we merged this cycle, where if we take
  a certain exception when the kernel is running relocated (currently
  only used for kdump), we checkstop the box.

  Thanks to Ravi Bangoria"

* tag 'powerpc-4.11-8' of git://
  powerpc/64: Fix HMI exception on LE with CONFIG_RELOCATABLE=y
  powerpc/kprobe: Fix oops when kprobed on 'stdu' instruction

5 years agoMerge tag 'pci-v4.11-fixes-5' of git://
Linus Torvalds [Fri, 21 Apr 2017 16:26:34 +0000 (09:26 -0700)]
Merge tag 'pci-v4.11-fixes-5' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fix from Bjorn Helgaas:
 "Sorry this is so late. It's been in -next for over a week, but I
  forgot to send it on until now.

  A single fix to the DT binding of the HiSilicon PCIe host support"

* tag 'pci-v4.11-fixes-5' of git://
  PCI: hisi: Fix DT binding (hisi-pcie-almost-ecam)

5 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Fri, 21 Apr 2017 16:20:35 +0000 (09:20 -0700)]
Merge branch 'for-linus' of git://

Pull block layer fixes from Jens Axboe:
 "A couple of last minute fixes for regressions in this cycle. More

   - Two patches from Andy, adjusting the NVMe APST quirks to avoid some
     issues specific to one Toshiba drive, and some variant of Samsung
     on two specific Dell laptops.

   - A fix for mtip32xx, turning off mq scheduling on that device. We
     have a real fix for this, but it's too late in the cycle.
     Thankfully we already have a NO_SCHED flag we can apply here. A
     prep patch for this is ensuring that we honor the NO_SCHED flag
     when attempting to online switch schedulers, previsouly we only did
     so for drive load time. From Ming.

   - Fixing an oops in blk-mq polling with scheduling attached. This one
     is easily reproducible, it would be a shame to release 4.11 with
     that issue. From me.

  I'd prefer not having to send in patches at this point in time, but
  the above are all things that have regressed in this cycle and the
  fixes are relatively straight forward"

* 'for-linus' of git://
  blk-mq: fix potential oops with polling and blk-mq scheduler
  nvme: Quirk APST off on "THNSF5256GPUK TOSHIBA"
  nvme: Adjust the Samsung APST quirk
  mtip32xx: pass BLK_MQ_F_NO_SCHED
  block: respect BLK_MQ_F_NO_SCHED

5 years agoMerge tag 'acpi-4.11-final' of git://
Linus Torvalds [Fri, 21 Apr 2017 16:18:17 +0000 (09:18 -0700)]
Merge tag 'acpi-4.11-final' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI build fix from Rafael Wysocki:
 "This avoids a false-positive build warning from the compiler.


   - Avoid a false-positive warning regarding a variable that may not be
     initialized that started to trigger after a previous general build
     fix (Arnd Bergmann)"

* tag 'acpi-4.11-final' of git://
  ACPI / power: Avoid maybe-uninitialized warning

5 years agoMerge tag 'mmc-v4.11-rc7' of git://
Linus Torvalds [Fri, 21 Apr 2017 16:15:43 +0000 (09:15 -0700)]
Merge tag 'mmc-v4.11-rc7' of git://git./linux/kernel/git/ulfh/mmc

Pull MMC fixes from Ulf Hansson:
 "MMC core:

   - kmalloc sdio scratch buffer to make it DMA-friendly

  MMC host:

   - dw_mmc: Fix behaviour for SDIO IRQs when runtime PM is used

   - sdhci-esdhc-imx: Correct pad I/O drive strength for UHS-DDR50

* tag 'mmc-v4.11-rc7' of git://
  mmc: sdhci-esdhc-imx: increase the pad I/O drive strength for DDR50 card
  mmc: dw_mmc: Don't allow Runtime PM for SDIO cards
  mmc: sdio: fix alignment issue in struct sdio_func

5 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Fri, 21 Apr 2017 16:13:43 +0000 (09:13 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixlet from Dmitry Torokhov:
 "An update to Elan PS/2 driver to allow working on yet another

* 'for-linus' of git://
  Input: elantech - add Fujitsu Lifebook E547 to force crc_enabled

5 years agoASoC: rsnd: fix semicolon.cocci warnings
kbuild test robot [Fri, 21 Apr 2017 05:02:57 +0000 (13:02 +0800)]
ASoC: rsnd: fix semicolon.cocci warnings

sound/soc/sh/rcar/adg.c:462:54-55: Unneeded semicolon

 Remove unneeded semicolon.

Generated by: scripts/coccinelle/misc/semicolon.cocci

Signed-off-by: Fengguang Wu <>
Acked-by: Kuninori Morimoto <>
Signed-off-by: Mark Brown <>
5 years agoMAINTAINERS: Add "B:" field for networking.
David S. Miller [Fri, 21 Apr 2017 14:44:47 +0000 (10:44 -0400)]
MAINTAINERS: Add "B:" field for networking.

We want people to report bugs to the netdev list.

Signed-off-by: David S. Miller <>
5 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Thu, 20 Apr 2017 22:31:08 +0000 (15:31 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge two mm fixes from Andrew Morton.

* emailed patches from Andrew Morton <>:
  mm: prevent NR_ISOLATE_* stats from going negative
  Revert "mm, page_alloc: only use per-cpu allocator for irq-safe requests"

5 years agomm: prevent NR_ISOLATE_* stats from going negative
Rabin Vincent [Thu, 20 Apr 2017 21:37:46 +0000 (14:37 -0700)]
mm: prevent NR_ISOLATE_* stats from going negative

Commit 6afcf8ef0ca0 ("mm, compaction: fix NR_ISOLATED_* stats for pfn
based migration") moved the dec_node_page_state() call (along with the
page_is_file_cache() call) to after putback_lru_page().

But page_is_file_cache() can change after putback_lru_page() is called,
so it should be called before putback_lru_page(), as it was before that
patch, to prevent NR_ISOLATE_* stats from going negative.

Without this fix, non-CONFIG_SMP kernels end up hanging in the
while(too_many_isolated()) { congestion_wait() } loop in
shrink_active_list() due to the negative stats.

  active_anon:32567 inactive_anon:121 isolated_anon:1
  active_file:6066 inactive_file:6639 isolated_file:4294967295
  unevictable:0 dirty:115 writeback:0 unstable:0
  slab_reclaimable:2086 slab_unreclaimable:3167
  mapped:3398 shmem:18366 pagetables:1145 bounce:0
  free:1798 free_pcp:13 free_cma:0

Fixes: 6afcf8ef0ca0 ("mm, compaction: fix NR_ISOLATED_* stats for pfn based migration")
Signed-off-by: Rabin Vincent <>
Acked-by: Michal Hocko <>
Cc: Ming Ling <>
Cc: Minchan Kim <>
Cc: Vlastimil Babka <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
5 years agoRevert "mm, page_alloc: only use per-cpu allocator for irq-safe requests"
Mel Gorman [Thu, 20 Apr 2017 21:37:43 +0000 (14:37 -0700)]
Revert "mm, page_alloc: only use per-cpu allocator for irq-safe requests"

This reverts commit 374ad05ab64.

While the patch worked great for userspace allocations, the fact that
softirq loses the per-cpu allocator caused problems.  It needs to be
redone taking into account that a separate list is needed for hard/soft
IRQs or alternatively find a cheap way of detecting reentry due to an
interrupt.  Both are possible but sufficiently tricky that it shouldn't
be rushed.

Jesper had one method for allowing softirqs but reported that the cost
was high enough that it performed similarly to a plain revert.  His
figures for netperf TCP_STREAM were as follows

  Baseline v4.10.0  : 60316 Mbit/s
  Current 4.11.0-rc6: 47491 Mbit/s
  Jesper's patch    : 60662 Mbit/s
  This patch        : 60106 Mbit/s

As this is a regression, I wish to revert to noirq allocator for now and
go back to the drawing board.

Signed-off-by: Mel Gorman <>
Reported-by: Tariq Toukan <>
Acked-by: Jesper Dangaard Brouer <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
5 years agoblk-mq: fix potential oops with polling and blk-mq scheduler
Jens Axboe [Thu, 20 Apr 2017 20:53:28 +0000 (14:53 -0600)]
blk-mq: fix potential oops with polling and blk-mq scheduler

If we have a scheduler attached, blk_mq_tag_to_rq() on the
scheduled tags will return NULL if a request is no longer
in flight. This is different than using the normal tags,
where it will always return the fixed request. Check for
this condition for polling, in case we happen to enter
polling for a completed request.

The request address remains valid, so this check and return
should be perfectly safe.

Fixes: bd166ef183c2 ("blk-mq-sched: add framework for MQ capable IO schedulers")
Tested-by: Stephen Bates <>
Signed-off-by: Jens Axboe <>
5 years agonvme: Quirk APST off on "THNSF5256GPUK TOSHIBA"
Andy Lutomirski [Thu, 20 Apr 2017 20:37:56 +0000 (13:37 -0700)]
nvme: Quirk APST off on "THNSF5256GPUK TOSHIBA"

There's a report that it malfunctions with APST on.


Cc: Kai-Heng Feng <>
Signed-off-by: Andy Lutomirski <>
Signed-off-by: Jens Axboe <>
5 years agonvme: Adjust the Samsung APST quirk
Andy Lutomirski [Thu, 20 Apr 2017 20:37:55 +0000 (13:37 -0700)]
nvme: Adjust the Samsung APST quirk

I got a couple more reports: the Samsung APST issues appears to
affect multiple 950-series devices in Dell XPS 15 9550 and Precision
5510 laptops.  Change the quirk: rather than blacklisting the
firmware on the first problematic SSD that was reported, disable
APST on all 144d:a802 devices if they're installed in the two
affected Dell models.  While we're at it, disable only the deepest
sleep state instead of all of them -- the reporters say that this is
sufficient to fix the problem.

(I have a device that appears to be entirely identical to one of the
affected devices, but I have a different Dell laptop, so it's not
the case that all Samsung devices with firmware BXW75D0Q are broken
under all circumstances.)

Samsung engineers have an affected system, and hopefully they'll
give us a better workaround some time soon.  In the mean time, this
should minimize regressions.


Cc: Kai-Heng Feng <>
Signed-off-by: Andy Lutomirski <>
Signed-off-by: Jens Axboe <>
5 years agonet sched actions: allocate act cookie early
Wolfgang Bumiller [Thu, 20 Apr 2017 12:08:26 +0000 (14:08 +0200)]
net sched actions: allocate act cookie early

Policing filters do not use the TCA_ACT_* enum and the tb[]
nlattr array in tcf_action_init_1() doesn't get filled for
them so we should not try to look for a TCA_ACT_COOKIE
attribute in the then uninitialized array.
The error handling in cookie allocation then calls
tcf_hash_release() leading to invalid memory access later
Additionally, if cookie allocation fails after an already
existing non-policing filter has successfully been changed,
tcf_action_release() should not be called, also we would
have to roll back the changes in the error handling, so
instead we now allocate the cookie early and assign it on
success at the end.

Fixes: 1045ba77a596 ("net sched actions: Add support for user cookies")
Signed-off-by: Wolfgang Bumiller <>
Acked-by: Jamal Hadi Salim <>
Signed-off-by: David S. Miller <>
5 years agoMerge branch 'qed-dcbx-fixes'
David S. Miller [Thu, 20 Apr 2017 20:29:41 +0000 (16:29 -0400)]
Merge branch 'qed-dcbx-fixes'

Sudarsana Reddy Kalluru says:

qed: Dcbx bug fixes

The series has set of bug fixes for dcbx implementation of qed driver.
Please consider applying this to 'net' branch.

Signed-off-by: David S. Miller <>
5 years agoqed: Fix issue in populating the PFC config paramters. [Wed, 19 Apr 2017 10:19:55 +0000 (03:19 -0700)]
qed: Fix issue in populating the PFC config paramters.

Change ieee_setpfc() callback implementation to populate traffic class
count with the user provided value.

Signed-off-by: Sudarsana Reddy Kalluru <>
Signed-off-by: Yuval Mintz <>
Signed-off-by: David S. Miller <>
5 years agoqed: Fix possible system hang in the dcbnl-getdcbx() path. [Wed, 19 Apr 2017 10:19:54 +0000 (03:19 -0700)]
qed: Fix possible system hang in the dcbnl-getdcbx() path.

qed_dcbnl_get_dcbx() API uses kmalloc in GFT_KERNEL mode. The API gets
invoked in the interrupt context by qed_dcbnl_getdcbx callback. Need
to invoke this kmalloc in atomic mode.

Signed-off-by: Sudarsana Reddy Kalluru <>
Signed-off-by: Yuval Mintz <>
Signed-off-by: David S. Miller <>
5 years agoqed: Fix sending an invalid PFC error mask to MFW. [Wed, 19 Apr 2017 10:19:53 +0000 (03:19 -0700)]
qed: Fix sending an invalid PFC error mask to MFW.

PFC error-mask value is not supported by MFW, but this bit could be
set in the pfc bit-map of the operational parameters if remote device
supports it. These operational parameters are used as basis for
populating the dcbx config parameters. User provided configs will be
applied on top of these parameters and then send them to MFW when
requested. Driver need to clear the error-mask bit before sending the
config parameters to MFW.

Signed-off-by: Sudarsana Reddy Kalluru <>
Signed-off-by: Yuval Mintz <>
Signed-off-by: David S. Miller <>
5 years agoqed: Fix possible error in populating max_tc field. [Wed, 19 Apr 2017 10:19:52 +0000 (03:19 -0700)]
qed: Fix possible error in populating max_tc field.

Some adapters may not publish the max_tc value. Populate the default
value for max_tc field in case the mfw didn't provide one.

Signed-off-by: Sudarsana Reddy Kalluru <>
Signed-off-by: Yuval Mintz <>
Signed-off-by: David S. Miller <>
5 years agosmsc95xx: Use skb_cow_head to deal with cloned skbs
James Hughes [Wed, 19 Apr 2017 10:13:40 +0000 (11:13 +0100)]
smsc95xx: Use skb_cow_head to deal with cloned skbs

The driver was failing to check that the SKB wasn't cloned
before adding checksum data.
Replace existing handling to extend/copy the header buffer
with skb_cow_head.

Signed-off-by: James Hughes <>
Acked-by: Eric Dumazet <>
Acked-by: Woojung Huh <>
Signed-off-by: David S. Miller <>
5 years agoMAINTAINERS: update entry for TI's CPSW driver
Sekhar Nori [Wed, 19 Apr 2017 08:38:24 +0000 (14:08 +0530)]
MAINTAINERS: update entry for TI's CPSW driver

Mugunthan V N, who was reviewing TI's CPSW driver patches is
not working for TI anymore and wont be reviewing patches for
that driver.

Drop Mugunthan as the maintiainer for this driver.

Grygorii continues to be a reviewer. Dave Miller applies the
patches directly and adding a maintainer is actually
misleading since script stops suggesting
that Dave Miller be copied.

Signed-off-by: Sekhar Nori <>
Signed-off-by: David S. Miller <>
5 years agoMerge branch 'master' of git://
David S. Miller [Thu, 20 Apr 2017 20:19:46 +0000 (16:19 -0400)]
Merge branch 'master' of git://git./linux/kernel/git/klassert/ipsec

Steffen Klassert says:

pull request (net): ipsec 2017-04-19

Two fixes for af_key:

1) Add a lock to key dump to prevent a NULL pointer dereference.
   From Yuejie Shi.

2) Fix slab-out-of-bounds in parse_ipsecrequests.
   From Herbert Xu.

Please pull or let me know if there are problems.

Signed-off-by: David S. Miller <>
5 years agodp83640: don't recieve time stamps twice
Dan Carpenter [Tue, 18 Apr 2017 19:14:26 +0000 (22:14 +0300)]
dp83640: don't recieve time stamps twice

This patch is prompted by a static checker warning about a potential
use after free.  The concern is that netif_rx_ni() can free "skb" and we
call it twice.

When I look at the commit that added this, it looks like some stray
lines were added accidentally.  It doesn't make sense to me that we
would recieve the same data two times.  I asked the author but never
recieved a response.

I can't test this code, but I'm pretty sure my patch is correct.

Fixes: 4b063258ab93 ("dp83640: Delay scheduled work.")
Signed-off-by: Dan Carpenter <>
Acked-by: Stefan Sørensen <>
Signed-off-by: David S. Miller <>
5 years agoipv6: sr: fix out-of-bounds access in SRH validation
David Lebrun [Tue, 18 Apr 2017 15:59:49 +0000 (17:59 +0200)]
ipv6: sr: fix out-of-bounds access in SRH validation

This patch fixes an out-of-bounds access in seg6_validate_srh() when the
trailing data is less than sizeof(struct sr6_tlv).

Reported-by: Andrey Konovalov <>
Cc: Andrey Konovalov <>
Signed-off-by: David Lebrun <>
Signed-off-by: David S. Miller <>
5 years agoselftests/net: Fixes psock_fanout CBPF test case
Mike Maloney [Tue, 18 Apr 2017 15:14:16 +0000 (11:14 -0400)]
selftests/net: Fixes psock_fanout CBPF test case

'psock_fanout' has been failing since commit 4d7b9dc1f36a9 ("tools:
psock_lib: harden socket filter used by psock tests").  That commit
changed the CBPF filter to examine the full ethernet frame, and was
tested on 'psock_tpacket' which uses SOCK_RAW.  But 'psock_fanout' was
also using this same CBPF in two places, for filtering and fanout, on a
SOCK_DGRAM socket.

Change 'psock_fanout' to use SOCK_RAW so that the CBPF program used with
SO_ATTACH_FILTER can examine the entire frame.  Create a new CBPF
program for use with PACKET_FANOUT_DATA which ignores the header, as it
cannot see the ethernet header.

Tested: Ran tools/testing/selftests/net/psock_{fanout,tpacket} 10 times,
and they all passed.

Fixes: 4d7b9dc1f36a9 ("tools: psock_lib: harden socket filter used by psock tests")
Signed-off-by: 'Mike Maloney <>'
Signed-off-by: David S. Miller <>
5 years agomac80211: reject ToDS broadcast data frames
Johannes Berg [Thu, 20 Apr 2017 19:32:16 +0000 (21:32 +0200)]
mac80211: reject ToDS broadcast data frames

AP/AP_VLAN modes don't accept any real 802.11 multicast data
frames, but since they do need to accept broadcast management
frames the same is currently permitted for data frames. This
opens a security problem because such frames would be decrypted
with the GTK, and could even contain unicast L3 frames.

Since the spec says that ToDS frames must always have the BSSID
as the RA (addr1), reject any other data frames.

The problem was originally reported in "Predicting, Decrypting,
and Abusing WPA2/802.11 Group Keys" at usenix
and brought to my attention by Jouni.

Reported-by: Jouni Malinen <>
Signed-off-by: Johannes Berg <>
Dave, I didn't want to send you a new pull request for a single
commit yet again - can you apply this one patch as is?
Signed-off-by: David S. Miller <>
5 years agoMerge tag 'trace-v4.11-rc5-5' of git://
Linus Torvalds [Thu, 20 Apr 2017 19:30:10 +0000 (12:30 -0700)]
Merge tag 'trace-v4.11-rc5-5' of git://git./linux/kernel/git/rostedt/linux-trace

Pull two more ftrace fixes from Steven Rostedt:
 "While continuing my development, I uncovered two more small bugs.

  One is a race condition when enabling the snapshot function probe
  trigger. It enables the probe before allocating the snapshot, and if
  the probe triggers first, it stops tracing with a warning that the
  snapshot buffer was not allocated.

  The seconds is that the snapshot file should show how to use it when
  it is empty. But a bug fix from long ago broke the "is empty" test and
  the snapshot file no longer displays the help message"

* tag 'trace-v4.11-rc5-5' of git://
  ring-buffer: Have ring_buffer_iter_empty() return true when empty
  tracing: Allocate the snapshot buffer before enabling probe

5 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 20 Apr 2017 19:26:10 +0000 (12:26 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/hid

Pull HID fixes from Jiri Kosina:
 "Two last-minute regression fixes for Wacom driver from Jason Gerecke"

* 'for-linus' of git://
  HID: wacom: Override incorrect logical maximum contact identifier
  HID: wacom: Treat HID_DG_TOOLSERIALNUMBER as unsigned

5 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Thu, 20 Apr 2017 19:20:30 +0000 (12:20 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/s390/linux

Pull s390 fix from Martin Schwidefsky:
 "There is one more fix I would like to see in 4.11: The combination of
  KVM, CMMA and heavy paging can cause data corruption, the fix is to
  clear the _PAGE_UNUSED bit in set_pte_at()"

* 'for-linus' of git://
  s390/mm: fix CMMA vs KSM vs others

5 years agoMerge tag 'keys-fixes-20170419' of git://
Linus Torvalds [Thu, 20 Apr 2017 18:12:51 +0000 (11:12 -0700)]
Merge tag 'keys-fixes-20170419' of git://git./linux/kernel/git/dhowells/linux-fs

Pull keyrings fixes from David Howells:

 (1) Disallow keyrings whose name begins with a '.' to be joined

 (2) Change the name of the dead type to ".dead" to prevent user access

 (3) Fix keyctl_set_reqkey_keyring() to not leak thread keyrings

* tag 'keys-fixes-20170419' of git://
  KEYS: fix keyctl_set_reqkey_keyring() to not leak thread keyrings
  KEYS: Change the name of the dead type to ".dead" to prevent user access
  KEYS: Disallow keyrings beginning with '.' to be joined as session keyrings

5 years agoMerge tag 'mac80211-for-davem-2017-04-18' of git://
David S. Miller [Thu, 20 Apr 2017 17:36:42 +0000 (13:36 -0400)]
Merge tag 'mac80211-for-davem-2017-04-18' of git://git./linux/kernel/git/jberg/mac80211

Johannes Berg says:

A single fix, for the MU-MIMO monitor mode, that fixes
bad SKB accesses if the SKB was paged, which is the case
for the only driver supporting this - iwlwifi.

Signed-off-by: David S. Miller <>
5 years agogenirq/affinity: Fix calculating vectors to assign
Keith Busch [Wed, 19 Apr 2017 23:51:10 +0000 (19:51 -0400)]
genirq/affinity: Fix calculating vectors to assign

The vectors_per_node is calculated from the remaining available vectors.
The current vector starts after pre_vectors, so we need to subtract that
from the current to properly account for the number of remaining vectors
to assign.

Fixes: 3412386b531 ("irq/affinity: Fix extra vecs calculation")
Reported-by: Andrei Vagin <>
Signed-off-by: Keith Busch <>
Signed-off-by: Thomas Gleixner <>
5 years agommc: sdhci-esdhc-imx: increase the pad I/O drive strength for DDR50 card
Haibo Chen [Wed, 19 Apr 2017 02:53:51 +0000 (10:53 +0800)]
mmc: sdhci-esdhc-imx: increase the pad I/O drive strength for DDR50 card

Currently for DDR50 card, it need tuning in default. We meet tuning fail
issue for DDR50 card and some data CRC error when DDR50 sd card works.

This is because the default pad I/O drive strength can't make sure DDR50
card work stable. So increase the pad I/O drive strength for DDR50 card,
and use pins_100mhz.

This fixes DDR50 card support for IMX since DDR50 tuning was enabled from
commit 9faac7b95ea4 ("mmc: sdhci: enable tuning for DDR50")

Tested-and-reported-by: Tim Harvey <>
Signed-off-by: Haibo Chen <>
Cc: # v4.4+
Acked-by: Dong Aisheng <>
Acked-by: Adrian Hunter <>
Signed-off-by: Ulf Hansson <>
5 years agoHID: wacom: Override incorrect logical maximum contact identifier
Jason Gerecke [Wed, 19 Apr 2017 21:47:24 +0000 (14:47 -0700)]
HID: wacom: Override incorrect logical maximum contact identifier

It apears that devices designed around Wacom's G11 chipset (e.g. Lenovo
ThinkPad Yoga 260, Lenovo ThinkPad X1 Yoga, Dell XPS 12 9250, Dell Venue
8 Pro 5855, etc.) suffer from a common issue in their HID descriptors.
The logical maximum is not updated for the "Contact Identifier" usage,
leaving it as just "1" despite these devices being capable of tracking
far more touches.

Commit 60a221869803 began ignoring usages with out-of-range values,
causing problems for devices based on this chipset. Touches after
the first will have an out-of-range Contact Identifier, and ignoring
that usage will cause the kernel to incorrectly slot each finger's
events (along with all the knock-on userspace effects that entails).

This commit checks for these buggy descriptors and updates the maximum
where required. Prior chipsets have used "255" as the maximum (and the
G11, at least, doesn't seem to actually use IDs outside the range of
1..CONTACTMAX) so continue using this value.

Fixes: 60a221869803 ("HID: wacom: generic: add support for touchring")
Signed-off-by: Jason Gerecke <>
Signed-off-by: Jiri Kosina <>
5 years agoring-buffer: Have ring_buffer_iter_empty() return true when empty
Steven Rostedt (VMware) [Wed, 19 Apr 2017 18:29:46 +0000 (14:29 -0400)]
ring-buffer: Have ring_buffer_iter_empty() return true when empty

I noticed that reading the snapshot file when it is empty no longer gives a
status. It suppose to show the status of the snapshot buffer as well as how
to allocate and use it. For example:

 ># cat snapshot
 # tracer: nop
 # * Snapshot is allocated *
 # Snapshot commands:
 # echo 0 > snapshot : Clears and frees snapshot buffer
 # echo 1 > snapshot : Allocates snapshot buffer, if not already allocated.
 #                      Takes a snapshot of the main buffer.
 # echo 2 > snapshot : Clears snapshot buffer (but does not allocate or free)
 #                      (Doesn't have to be '2' works with any number that
 #                       is not a '0' or '1')

But instead it just showed an empty buffer:

 ># cat snapshot
 # tracer: nop
 # entries-in-buffer/entries-written: 0/0   #P:4
 #                              _-----=> irqs-off
 #                             / _----=> need-resched
 #                            | / _---=> hardirq/softirq
 #                            || / _--=> preempt-depth
 #                            ||| /     delay
 #           TASK-PID   CPU#  ||||    TIMESTAMP  FUNCTION
 #              | |       |   ||||       |         |

What happened was that it was using the ring_buffer_iter_empty() function to
see if it was empty, and if it was, it showed the status. But that function
was returning false when it was empty. The reason was that the iter header
page was on the reader page, and the reader page was empty, but so was the
buffer itself. The check only tested to see if the iter was on the commit
page, but the commit page was no longer pointing to the reader page, but as
all pages were empty, the buffer is also.

Fixes: 651e22f2701b ("ring-buffer: Always reset iterator to reader page")
Signed-off-by: Steven Rostedt (VMware) <>
5 years agoMerge tag 'clk-fixes-for-linus' of git://
Linus Torvalds [Thu, 20 Apr 2017 00:16:18 +0000 (17:16 -0700)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fixes from Stephen Boyd"

 - one stm32f4 fix for a change that introduced the PLL_I2S and PLL_SAI

 - two Allwinner clk driver build fixes

 - two Allwinner CPU clk driver fixes where we see random CPUFreq
   crashes because the CPU's PLL locks up sometimes when we change the

* tag 'clk-fixes-for-linus' of git://
  clk: sunxi-ng: a33: gate then ungate PLL CPU clk after rate change
  clk: sunxi-ng: Add clk notifier to gate then ungate PLL clocks
  clk: sunxi-ng: fix build failure in ccu-sun9i-a80 driver
  clk: sunxi-ng: fix build error without CONFIG_RESET_CONTROLLER
  clk: stm32f4: fix: exclude values 0 and 1 for PLLQ

5 years agoMerge branch 'for-next' of git://
Linus Torvalds [Thu, 20 Apr 2017 00:12:46 +0000 (17:12 -0700)]
Merge branch 'for-next' of git://

Pull CIFS fix from Steve French:
 "One more cifs fix for stable"

* 'for-next' of git://
  cifs: Do not send echoes before Negotiate is complete

5 years agonsfs: mark dentry with DCACHE_RCUACCESS
Cong Wang [Wed, 19 Apr 2017 22:11:00 +0000 (15:11 -0700)]
nsfs: mark dentry with DCACHE_RCUACCESS

Andrey reported a use-after-free in __ns_get_path():

  spin_lock include/linux/spinlock.h:299 [inline]
  lockref_get_not_dead+0x19/0x80 lib/lockref.c:179
  __ns_get_path+0x197/0x860 fs/nsfs.c:66
  open_related_ns+0xda/0x200 fs/nsfs.c:143
  sock_ioctl+0x39d/0x440 net/socket.c:1001
  vfs_ioctl fs/ioctl.c:45 [inline]
  do_vfs_ioctl+0x1bf/0x1780 fs/ioctl.c:685
  SYSC_ioctl fs/ioctl.c:700 [inline]
  SyS_ioctl+0x8f/0xc0 fs/ioctl.c:691

We are under rcu read lock protection at that point:

        d = atomic_long_read(&ns->stashed);
        if (!d)
                goto slow;
        dentry = (struct dentry *)d;
        if (!lockref_get_not_dead(&dentry->d_lockref))
                goto slow;

but don't use a proper RCU API on the free path, therefore a parallel
__d_free() could free it at the same time.  We need to mark the stashed
dentry with DCACHE_RCUACCESS so that __d_free() will be called after all
readers leave RCU.

Fixes: e149ed2b805f ("take the targets of /proc/*/ns/* symlinks to separate fs")
Cc: Alexander Viro <>
Cc: Andrew Morton <>
Reported-by: Andrey Konovalov <>
Signed-off-by: Cong Wang <>
Signed-off-by: Linus Torvalds <>
5 years agomm: make mm_percpu_wq non freezable
Michal Hocko [Wed, 19 Apr 2017 07:52:46 +0000 (09:52 +0200)]
mm: make mm_percpu_wq non freezable

Geert has reported a freeze during PM resume and some additional
debugging has shown that the device_resume worker cannot make a forward
progress because it waits for an event which is stuck waiting in

  INFO: task kworker/u4:0:5 blocked for more than 120 seconds.
        Not tainted 4.11.0-rc7-koelsch-00029-g005882e53d62f25d-dirty #3476
  "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
  kworker/u4:0    D    0     5      2 0x00000000
  Workqueue: events_unbound async_run_entry_fn
  bash            D    0  1703   1694 0x00000000
  Showing busy workqueues and worker pools:
  workqueue mm_percpu_wq: flags=0xc
    pwq 2: cpus=1 node=0 flags=0x0 nice=0 active=0/0
      delayed: drain_local_pages_wq, vmstat_update
    pwq 0: cpus=0 node=0 flags=0x0 nice=0 active=0/0
      delayed: drain_local_pages_wq BAR(1703), vmstat_update

Tetsuo has properly noted that mm_percpu_wq is created as WQ_FREEZABLE
so it is frozen this early during resume so we are effectively
deadlocked.  Fix this by dropping WQ_FREEZABLE when creating
mm_percpu_wq.  We really want to have it operational all the time.

Fixes: ce612879ddc7 ("mm: move pcp and lru-pcp draining into single wq")
Reported-and-tested-by: Geert Uytterhoeven <>
Debugged-by: Tetsuo Handa <>
Signed-off-by: Michal Hocko <>
Signed-off-by: Linus Torvalds <>
5 years agoMerge tag 'backlight-for-v4.11' of git://
Linus Torvalds [Wed, 19 Apr 2017 21:45:51 +0000 (14:45 -0700)]
Merge tag 'backlight-for-v4.11' of git://

Pull backlight fix from Daniel Thompson:
 "Normally pull requests for backlight come from Lee Jones (and will
  continue to do so) but the bug fixed here is annoying for few people
  so I'm providing a little holiday cover.

  Fix a single bug in the PWM backlight driver and make it play nice
  with a wider range of GPIO devices. This bug is a regression and was
  independently discovered by Geert Uytterhoevan and Paul Kocialkowski
  (and is tested by both)"

* tag 'backlight-for-v4.11' of git://
  backlight: pwm_bl: Fix GPIO out for unimplemented .get_direction()

5 years agoACPI / power: Avoid maybe-uninitialized warning
Arnd Bergmann [Wed, 19 Apr 2017 17:47:04 +0000 (19:47 +0200)]
ACPI / power: Avoid maybe-uninitialized warning

gcc -O2 cannot always prove that the loop in acpi_power_get_inferred_state()
is enterered at least once, so it assumes that cur_state might not get

drivers/acpi/power.c: In function 'acpi_power_get_inferred_state':
drivers/acpi/power.c:222:9: error: 'cur_state' may be used uninitialized in this function [-Werror=maybe-uninitialized]

This sets the variable to zero at the start of the loop, to ensure that
there is well-defined behavior even for an empty list. This gets rid of
the warning.

The warning first showed up when the -Os flag got removed in a bug fix
patch in linux-4.11-rc5.

I would suggest merging this addon patch on top of that bug fix to avoid
introducing a new warning in the stable kernels.

Fixes: 61b79e16c68d (ACPI: Fix incompatibility with mcount-based function graph tracing)
Cc: All applicable <>
Signed-off-by: Arnd Bergmann <>
Signed-off-by: Rafael J. Wysocki <>
5 years agomtip32xx: pass BLK_MQ_F_NO_SCHED
Ming Lei [Sat, 15 Apr 2017 12:38:23 +0000 (20:38 +0800)]
mtip32xx: pass BLK_MQ_F_NO_SCHED

The recent introduced MQ IO scheduler breaks mtip32xx in the
following way.

mtip32xx use the 'request_index' passed to .init_request() as
hardware tag index for initializing hardware queue, and it
actually require that rq->tag is always same with 'request_index'
passed to .init_request(). Current blk-mq IO scheduler can't
guarantee this point, so this patch passes BLK_MQ_F_NO_SCHED
and at least make mtip32xx working.

This patch fixes the following strange hardware failure. The
issue can be triggered easily when doing I/O with mq-deadline

[  186.972578] {1}[Hardware Error]: Hardware error from APEI Generic Hardware Error Source: 32993
[  186.972578] {1}[Hardware Error]: event severity: fatal
[  186.972579] {1}[Hardware Error]:  Error 0, type: fatal
[  186.972580] {1}[Hardware Error]:   section_type: PCIe error
[  186.972580] {1}[Hardware Error]:   port_type: 0, PCIe end point
[  186.972581] {1}[Hardware Error]:   version: 1.0
[  186.972581] {1}[Hardware Error]:   command: 0x0407, status: 0x0010
[  186.972582] {1}[Hardware Error]:   device_id: 0000:07:00.0
[  186.972582] {1}[Hardware Error]:   slot: 4
[  186.972583] {1}[Hardware Error]:   secondary_bus: 0x00
[  186.972583] {1}[Hardware Error]:   vendor_id: 0x1344, device_id: 0x5150
[  186.972584] {1}[Hardware Error]:   class_code: 008001
[  186.972585] Kernel panic - not syncing: Fatal hardware error!

Reported-by: Jozef Mikovic <>
Signed-off-by: Ming Lei <>
Signed-off-by: Jens Axboe <>
5 years agoblock: respect BLK_MQ_F_NO_SCHED
Ming Lei [Sat, 15 Apr 2017 12:38:22 +0000 (20:38 +0800)]
block: respect BLK_MQ_F_NO_SCHED

If one driver claims that it doesn't support io scheduler via
BLK_MQ_F_NO_SCHED, we should not allow to change and show the
availabe io schedulers.

This patch adds check to enhance this behaviour.

Signed-off-by: Ming Lei <>
Signed-off-by: Jens Axboe <>
5 years agobacklight: pwm_bl: Fix GPIO out for unimplemented .get_direction()
Geert Uytterhoeven [Tue, 4 Apr 2017 10:54:35 +0000 (12:54 +0200)]
backlight: pwm_bl: Fix GPIO out for unimplemented .get_direction()

Commit 7613c922315e308a ("backlight: pwm_bl: Move the checks for initial
power state to a separate function") not just moved some code, but made
slight changes in semantics.

If a gpiochip doesn't implement the optional .get_direction() callback,
gpiod_get_direction always returns -EINVAL, which is never equal to
GPIOF_DIR_IN, leading to the GPIO not being configured for output.

To avoid this, invert the test and check for not GPIOF_DIR_OUT instead,
like the original code did.

This restores the display on r8a7740/armadillo.

Fixes: 7613c922315e308a ("backlight: pwm_bl: Move the checks for initial power state to a separate function")
Signed-off-by: Geert Uytterhoeven <>
Reviewed-by: Peter Ujfalusi <>
Acked-by: Philipp Zabel <>
Acked-by: Daniel Thompson <>
Signed-off-by: Daniel Thompson <>
5 years agotracing: Allocate the snapshot buffer before enabling probe
Steven Rostedt (VMware) [Wed, 19 Apr 2017 16:07:08 +0000 (12:07 -0400)]
tracing: Allocate the snapshot buffer before enabling probe

Currently the snapshot trigger enables the probe and then allocates the
snapshot. If the probe triggers before the allocation, it could cause the
snapshot to fail and turn tracing off. It's best to allocate the snapshot
buffer first, and then enable the trigger. If something goes wrong in the
enabling of the trigger, the snapshot buffer is still allocated, but it can
also be freed by the user by writting zero into the snapshot buffer file.

Also add a check of the return status of alloc_snapshot().

Fixes: 77fd5c15e3 ("tracing: Add snapshot trigger to function probes")
Signed-off-by: Steven Rostedt (VMware) <>
5 years agoASoC: rsnd: enable clock-frequency for both 44.1kHz/48kHz
Kuninori Morimoto [Wed, 19 Apr 2017 00:45:52 +0000 (00:45 +0000)]
ASoC: rsnd: enable clock-frequency for both 44.1kHz/48kHz

Current clock-frequency allows only 1 clock, but ADG can
handle both 44.1kHz/48kHz base clocks. This patch enables these.

On Salvator-X board, AUDIO_CLKOUT which is generated by ADG
is connected to ak4613 MCKI, and it should be synchronized with
LRCK. Thus, we need both 44.1kHz/48kHz base clock-frequency.
Otherwise, either one sounds strange in high frequency sound.

Signed-off-by: Kuninori Morimoto <>
Tested-by: Hiroyuki Yokoyama <>
Signed-off-by: Mark Brown <>
5 years agoASoC: rsnd: rsnd_mod_make_sure() is not under DEBUG
Kuninori Morimoto [Wed, 19 Apr 2017 00:41:24 +0000 (00:41 +0000)]
ASoC: rsnd: rsnd_mod_make_sure() is not under DEBUG

rsnd_mod_make_sure() will be used any situation,
thus, under DEBUG is not realistic.
This patch move it to non DEBUG area

Signed-off-by: Kuninori Morimoto <>
Signed-off-by: Mark Brown <>