openwrt/staging/blogic.git
7 years agomembarrier: Provide register expedited private command
Mathieu Desnoyers [Thu, 19 Oct 2017 17:30:15 +0000 (13:30 -0400)]
membarrier: Provide register expedited private command

This introduces a "register private expedited" membarrier command which
allows eventual removal of important memory barrier constraints on the
scheduler fast-paths. It changes how the "private expedited" membarrier
command (new to 4.14) is used from user-space.

This new command allows processes to register their intent to use the
private expedited command.  This affects how the expedited private
command introduced in 4.14-rc is meant to be used, and should be merged
before 4.14 final.

Processes are now required to register before using
MEMBARRIER_CMD_PRIVATE_EXPEDITED, otherwise that command returns EPERM.

This fixes a problem that arose when designing requested extensions to
sys_membarrier() to allow JITs to efficiently flush old code from
instruction caches.  Several potential algorithms are much less painful
if the user register intent to use this functionality early on, for
example, before the process spawns the second thread.  Registering at
this time removes the need to interrupt each and every thread in that
process at the first expedited sys_membarrier() system call.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoMerge branch 'parisc-4.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller...
Linus Torvalds [Thu, 19 Oct 2017 20:18:58 +0000 (16:18 -0400)]
Merge branch 'parisc-4.14-3' of git://git./linux/kernel/git/deller/parisc-linux

Pull parisc fixes from Helge Deller:
 "Three small important fixes for the parisc architecture:

   - Export __cmpxchg_u64() symbol on 32bit kernel too. This unbreaks
     building the kernel with ixgbe kernel module. From Guenter Roeck.

   - Fix 64-bit atomic cmpxchg kernel helper function for 32-bit kernel
     in LWS code for userspace. This unbreaks e.g. the 64-bit variant of
     the glibc function __sync_fetch_and_add() with a 32-bit parisc
     kernel. From John David Anglin, tagged for backport to v3.13+.

   - Detect nonsynchronous CPU-internal cr16 cycle counters more
     reliable. This avoids stalled CPU warnings by the kernel soft
     lockup detector. From me, tagged for backport to v4.13+"

* 'parisc-4.14-3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/parisc-linux:
  parisc: Fix detection of nonsynchronous cr16 cycle counters
  parisc: Export __cmpxchg_u64 unconditionally
  parisc: Fix double-word compare and exchange in LWS code on 32-bit kernels

7 years agoMerge tag 'sound-4.14-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Thu, 19 Oct 2017 20:15:17 +0000 (16:15 -0400)]
Merge tag 'sound-4.14-rc6' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "We've got slightly more fixes than wished, but heading to a good
  shape. Most of changes are about HD-audio fixes, one for a buggy code
  that went into 4.13, and another for avoiding a crash due to buggy
  BIOS.

  Apart from HD-audio, a sequencer core change that is only for UP
  config (which must be pretty rare nowadays), and a USB-audio quirk as
  usual"

* tag 'sound-4.14-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda - Fix incorrect TLV callback check introduced during set_fs() removal
  ALSA: hda: Remove superfluous '-' added by printk conversion
  ALSA: hda: Abort capability probe at invalid register read
  ALSA: seq: Enable 'use' locking in all configurations
  ALSA: usb-audio: Add native DSD support for Pro-Ject Pre Box S2 Digital

7 years agoparisc: Fix detection of nonsynchronous cr16 cycle counters
Helge Deller [Wed, 18 Oct 2017 20:25:00 +0000 (22:25 +0200)]
parisc: Fix detection of nonsynchronous cr16 cycle counters

For CPUs which have an unknown or invalid CPU location (physical location)
assume that their cycle counters aren't syncronized across CPUs.

Signed-off-by: Helge Deller <deller@gmx.de>
Fixes: c8c3735997a3 ("parisc: Enhance detection of synchronous cr16 clocksources")
Cc: stable@vger.kernel.org # 4.13+
Signed-off-by: Helge Deller <deller@gmx.de>
7 years agoparisc: Export __cmpxchg_u64 unconditionally
Guenter Roeck [Sun, 24 Sep 2017 17:28:09 +0000 (10:28 -0700)]
parisc: Export __cmpxchg_u64 unconditionally

__cmpxchg_u64 is built and used outside CONFIG_64BIT and thus needs to
be exported. This fixes the following build error seen when building
parisc:allmodconfig.

ERROR: "__cmpxchg_u64" [drivers/net/ethernet/intel/i40e/i40e.ko] undefined!

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Helge Deller <deller@gmx.de>
7 years agoparisc: Fix double-word compare and exchange in LWS code on 32-bit kernels
John David Anglin [Sat, 30 Sep 2017 21:24:23 +0000 (17:24 -0400)]
parisc: Fix double-word compare and exchange in LWS code on 32-bit kernels

As discussed on the debian-hppa list, double-wordcompare and exchange
operations fail on 32-bit kernels.  Looking at the code, I realized that
the ",ma" completer does the wrong thing in the  "ldw,ma  4(%r26), %r29"
instruction.  This increments %r26 and causes the following store to
write to the wrong location.

Note by Helge Deller:
The patch applies cleanly to stable kernel series if this upstream
commit is merged in advance:
f4125cfdb300 ("parisc: Avoid trashing sr2 and sr3 in LWS code").

Signed-off-by: John David Anglin <dave.anglin@bell.net>
Tested-by: Christoph Biedl <debian.axhn@manchmal.in-ulm.de>
Fixes: 89206491201c ("parisc: Implement new LWS CAS supporting 64 bit operations.")
Cc: stable@vger.kernel.org # 3.13+
Signed-off-by: Helge Deller <deller@gmx.de>
7 years agoMerge tag 'xfs-4.14-fixes-6' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Wed, 18 Oct 2017 18:51:50 +0000 (14:51 -0400)]
Merge tag 'xfs-4.14-fixes-6' of git://git./fs/xfs/xfs-linux

Pull xfs fixes from Darrick Wong:

 - fix some more CONFIG_XFS_RT related build problems

 - fix data loss when writeback at eof races eofblocks gc and loses

 - invalidate page cache after fs finishes a dio write

 - remove dirty page state when invalidating pages so releasepage does
   the right thing when handed a dirty page

* tag 'xfs-4.14-fixes-6' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: move two more RT specific functions into CONFIG_XFS_RT
  xfs: trim writepage mapping to within eof
  fs: invalidate page cache after end_io() in dio completion
  xfs: cancel dirty pages on invalidation

7 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Wed, 18 Oct 2017 18:43:40 +0000 (14:43 -0400)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "Three small fixes:

   - A fix for skd, it was using kfree() to free a structure allocate
     with kmem_cache_alloc().

   - Stable fix for nbd, fixing a regression using the normal ioctl
     based tools.

   - Fix for a previous fix in this series, that fixed up
     inconsistencies between buffered and direct IO"

* 'for-linus' of git://git.kernel.dk/linux-block:
  fs: Avoid invalidation in interrupt context in dio_complete()
  nbd: don't set the device size until we're connected
  skd: Use kmem_cache_free

7 years agoMerge tag 'enforcement-4.14-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 18 Oct 2017 10:47:59 +0000 (06:47 -0400)]
Merge tag 'enforcement-4.14-rc6' of git://git./linux/kernel/git/gregkh/driver-core

Pull enforcement policy update from Greg KH:
 "Documentation: Add a file explaining the requested Linux kernel
  license enforcement policy

  Here's a new file to the kernel's Documentation directory. It adds a
  short document describing the views of how the Linux kernel community
  feels about enforcing the license of the kernel.

  The patch has been reviewed by a large number of kernel developers
  already, as seen by their acks on the patch, and their agreement of
  the statement with their names on it. The location of the file was
  also agreed upon by the Documentation maintainer, so all should be
  good there.

  For some background information about this statement, see this article
  written by some of the kernel developers involved in drafting it:

http://kroah.com/log/blog/2017/10/16/linux-kernel-community-enforcement-statement/

  and this article that answers a number of questions that came up in
  the discussion of this statement with the kernel developer community:

http://kroah.com/log/blog/2017/10/16/linux-kernel-community-enforcement-statement-faq/

  If anyone has any further questions about it, please let me, and the
  TAB members, know and we will be glad to help answer them"

* tag 'enforcement-4.14-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  Documentation: Add a file explaining the Linux kernel license enforcement policy

7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Wed, 18 Oct 2017 10:45:52 +0000 (06:45 -0400)]
Merge branch 'for-linus' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Martin Schwidefsky:
 "Two bug fixes:

   - A fix for cputime accounting vs CPU hotplug

   - Add two options to zfcpdump_defconfig to make SCSI dump work again"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390: fix zfcpdump-config
  s390/cputime: fix guest/irq/softirq times after CPU hotplug

7 years agoMerge tag 'trace-v4.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Wed, 18 Oct 2017 10:43:30 +0000 (06:43 -0400)]
Merge tag 'trace-v4.14-rc3' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fix from Steven Rostedt:
 "Testing a new trace event format, I triggered a bug by doing:

    # modprobe trace-events-sample
    # echo 1 > /sys/kernel/debug/tracing/events/sample-trace/enable
    # rmmod trace-events-sample

  This would cause an oops. The issue is that I added another trace
  event sample that reused a reg function of another trace event to
  create a thread to call the tracepoints. The problem was that the reg
  function couldn't handle nested calls (reg; reg; unreg; unreg;) and
  created two threads (instead of one) and only removed one on exit.

  This isn't a critical bug as the bug is only in sample code. But
  sample code should be free of known bugs to prevent others from
  copying it. This is why this is also marked for stable"

* tag 'trace-v4.14-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing/samples: Fix creation and deletion of simple_thread_fn creation

7 years agoALSA: hda - Fix incorrect TLV callback check introduced during set_fs() removal
Takashi Iwai [Mon, 16 Oct 2017 09:39:28 +0000 (11:39 +0200)]
ALSA: hda - Fix incorrect TLV callback check introduced during set_fs() removal

The commit 99b5c5bb9a54 ("ALSA: hda - Remove the use of set_fs()")
converted the get_kctl_0dB_offset() call for killing set_fs() usage in
HD-audio codec code.  The conversion assumed that the TLV callback
used in HD-audio code is only snd_hda_mixer_amp() and applies the TLV
calculation locally.

Although this assumption is correct, and all slave kctls are actually
with that callback, the current code is still utterly buggy; it
doesn't hit this condition and falls back to the next check.  It's
because the function gets called after adding slave kctls to vmaster.
By assigning a slave kctl, the slave kctl object is faked inside
vmaster code, and the whole kctl ops are overridden.  Thus the
callback op points to a different value from what we've assumed.

More badly, as reported by the KERNEXEC and UDEREF features of PaX,
the code flow turns into the unexpected pitfall.  The next fallback
check is SNDRV_CTL_ELEM_ACCESS_TLV_READ access bit, and this always
hits for each kctl with TLV.  Then it evaluates the callback function
pointer wrongly as if it were a TLV array.  Although currently its
side-effect is fairly limited, this incorrect reference may lead to an
unpleasant result.

For addressing the regression, this patch introduces a new helper to
vmaster code, snd_ctl_apply_vmaster_slaves().  This works similarly
like the existing map_slaves() in hda_codec.c: it loops over the slave
list of the given master, and applies the given function to each
slave.  Then the initializer function receives the right kctl object
and we can compare the correct pointer instead of the faked one.

Also, for catching the similar breakage in future, give an error
message when the unexpected TLV callback is found and bail out
immediately.

Fixes: 99b5c5bb9a54 ("ALSA: hda - Remove the use of set_fs()")
Reported-by: PaX Team <pageexec@freemail.hu>
Cc: <stable@vger.kernel.org> # v4.13
Signed-off-by: Takashi Iwai <tiwai@suse.de>
7 years agoALSA: hda: Remove superfluous '-' added by printk conversion
Takashi Iwai [Tue, 17 Oct 2017 09:58:17 +0000 (11:58 +0200)]
ALSA: hda: Remove superfluous '-' added by printk conversion

While converting the error messages to the standard macros in the
commit 4e76a8833fac ("ALSA: hda - Replace with standard printk"), a
superfluous '-' slipped in the code mistakenly.  Its influence is
almost negligible, merely shows a dB value as negative integer instead
of positive integer (or vice versa) in the rare error message.
So let's kill this embarrassing byte to show more correct value.

Fixes: 4e76a8833fac ("ALSA: hda - Replace with standard printk")
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
7 years agoALSA: hda: Abort capability probe at invalid register read
Takashi Iwai [Tue, 17 Oct 2017 14:38:55 +0000 (16:38 +0200)]
ALSA: hda: Abort capability probe at invalid register read

The loop in snd_hdac_bus_parse_capabilities() may go to nirvana when
it hits an invalid register value read:

 BUG: unable to handle kernel paging request at ffffad5dc41f3fff
 IP: pci_azx_readl+0x5/0x10 [snd_hda_intel]
 Call Trace:
  snd_hdac_bus_parse_capabilities+0x3c/0x1f0 [snd_hda_core]
  azx_probe_continue+0x7d5/0x940 [snd_hda_intel]
  .....

This happened on a new Intel machine, and we need to check the value
and abort the loop accordingly.

[Note: the fixes tag below indicates only the commit where this patch
 can be applied; the original problem was introduced even before that
 commit]

Fixes: 6720b38420a0 ("ALSA: hda - move bus_parse_capabilities to core")
Cc: <stable@vger.kernel.org>
Acked-by: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
7 years agoALSA: seq: Enable 'use' locking in all configurations
Ben Hutchings [Tue, 17 Oct 2017 23:45:49 +0000 (00:45 +0100)]
ALSA: seq: Enable 'use' locking in all configurations

The 'use' locking macros are no-ops if neither SMP or SND_DEBUG is
enabled.  This might once have been OK in non-preemptible
configurations, but even in that case snd_seq_read() may sleep while
relying on a 'use' lock.  So always use the proper implementations.

Cc: stable@vger.kernel.org
Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
7 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Tue, 17 Oct 2017 19:54:41 +0000 (15:54 -0400)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Four mostly error leg fixes and one more important regression in a
  prior commit (the qla2xxx one)"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: fc: check for rport presence in fc_block_scsi_eh
  scsi: qla2xxx: Fix uninitialized work element
  scsi: libiscsi: fix shifting of DID_REQUEUE host byte
  scsi: libfc: fix a deadlock in fc_rport_work
  scsi: fixup kernel warning during rmmod()

7 years agotracing/samples: Fix creation and deletion of simple_thread_fn creation
Steven Rostedt (VMware) [Tue, 17 Oct 2017 18:55:24 +0000 (14:55 -0400)]
tracing/samples: Fix creation and deletion of simple_thread_fn creation

Commit 7496946a8 ("tracing: Add samples of DECLARE_EVENT_CLASS() and
DEFINE_EVENT()") added template examples for all the events. It created a
DEFINE_EVENT_FN() example which reused the foo_bar_reg and foo_bar_unreg
functions.

Enabling both the TRACE_EVENT_FN() and DEFINE_EVENT_FN() example trace
events caused the foo_bar_reg to be called twice, creating the test thread
twice. The foo_bar_unreg would remove it only once, even if it was called
multiple times, leaving a thread existing when the module is unloaded,
causing an oops.

Add a ref count and allow foo_bar_reg() and foo_bar_unreg() be called by
multiple trace events.

Cc: stable@vger.kernel.org
Fixes: 7496946a8 ("tracing: Add samples of DECLARE_EVENT_CLASS() and DEFINE_EVENT()")
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
7 years agofs: Avoid invalidation in interrupt context in dio_complete()
Lukas Czerner [Tue, 17 Oct 2017 14:43:09 +0000 (08:43 -0600)]
fs: Avoid invalidation in interrupt context in dio_complete()

Currently we try to defer completion of async DIO to the process context
in case there are any mapped pages associated with the inode so that we
can invalidate the pages when the IO completes. However the check is racy
and the pages can be mapped afterwards. If this happens we might end up
calling invalidate_inode_pages2_range() in dio_complete() in interrupt
context which could sleep. This can be reproduced by generic/451.

Fix this by passing the information whether we can or can't invalidate
to the dio_complete(). Thanks Eryu Guan for reporting this and Jan Kara
for suggesting a fix.

Fixes: 332391a9935d ("fs: Fix page cache inconsistency when mixing buffered and AIO DIO")
Reported-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Tested-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
7 years agoMerge tag 'media/v4.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Tue, 17 Oct 2017 10:23:09 +0000 (06:23 -0400)]
Merge tag 'media/v4.14-2' of git://git./linux/kernel/git/mchehab/linux-media

Pull media fixes from Mauro Carvalho Chehab:
 "Core fixes:
   - cec: Respond to unregistered initiators, when applicable
   - dvb_frontend: only use kref after initialized

  Driver-specific fixes:
   - qcom, camss: Make function vfe_set_selection static
   - qcom: VIDEO_QCOM_CAMSS should depend on HAS_DMA
   - s5p-cec: add NACK detection support
   - media: staging/imx: Fix uninitialized variable warning
   - dib3000mc: i2c transfers over usb cannot be done from stack
   - venus: init registered list on streamoff"

* tag 'media/v4.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  media: dvb_frontend: only use kref after initialized
  media: platform: VIDEO_QCOM_CAMSS should depend on HAS_DMA
  media: cec: Respond to unregistered initiators, when applicable
  media: s5p-cec: add NACK detection support
  media: staging/imx: Fix uninitialized variable warning
  media: qcom: camss: Make function vfe_set_selection static
  media: venus: init registered list on streamoff
  media: dvb: i2c transfers over usb cannot be done from stack

7 years agoxfs: move two more RT specific functions into CONFIG_XFS_RT
Arnd Bergmann [Fri, 13 Oct 2017 16:47:47 +0000 (09:47 -0700)]
xfs: move two more RT specific functions into CONFIG_XFS_RT

The last cleanup introduced two harmless warnings:

fs/xfs/xfs_fsmap.c:480:1: warning: '__xfs_getfsmap_rtdev' defined but not used
fs/xfs/xfs_fsmap.c:372:1: warning: 'xfs_getfsmap_rtdev_rtbitmap_helper' defined but not used

This moves those two functions as well.

Fixes: bb9c2e543325 ("xfs: move more RT specific code under CONFIG_XFS_RT")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
7 years agoxfs: trim writepage mapping to within eof
Brian Foster [Fri, 13 Oct 2017 16:47:46 +0000 (09:47 -0700)]
xfs: trim writepage mapping to within eof

The writeback rework in commit fbcc02561359 ("xfs: Introduce
writeback context for writepages") introduced a subtle change in
behavior with regard to the block mapping used across the
->writepages() sequence. The previous xfs_cluster_write() code would
only flush pages up to EOF at the time of the writepage, thus
ensuring that any pages due to file-extending writes would be
handled on a separate cycle and with a new, updated block mapping.

The updated code establishes a block mapping in xfs_writepage_map()
that could extend beyond EOF if the file has post-eof preallocation.
Because we now use the generic writeback infrastructure and pass the
cached mapping to each writepage call, there is no implicit EOF
limit in place. If eofblocks trimming occurs during ->writepages(),
any post-eof portion of the cached mapping becomes invalid. The
eofblocks code has no means to serialize against writeback because
there are no pages associated with post-eof blocks. Therefore if an
eofblocks trim occurs and is followed by a file-extending buffered
write, not only has the mapping become invalid, but we could end up
writing a page to disk based on the invalid mapping.

Consider the following sequence of events:

- A buffered write creates a delalloc extent and post-eof
  speculative preallocation.
- Writeback starts and on the first writepage cycle, the delalloc
  extent is converted to real blocks (including the post-eof blocks)
  and the mapping is cached.
- The file is closed and xfs_release() trims post-eof blocks. The
  cached writeback mapping is now invalid.
- Another buffered write appends the file with a delalloc extent.
- The concurrent writeback cycle picks up the just written page
  because the writeback range end is LLONG_MAX. xfs_writepage_map()
  attributes it to the (now invalid) cached mapping and writes the
  data to an incorrect location on disk (and where the file offset is
  still backed by a delalloc extent).

This problem is reproduced by xfstests test generic/464, which
triggers racing writes, appends, open/closes and writeback requests.

To address this problem, trim the mapping used during writeback to
within EOF when the mapping is validated. This ensures the mapping
is revalidated for any pages encountered beyond EOF as of the time
the current mapping was cached or last validated.

Reported-by: Eryu Guan <eguan@redhat.com>
Diagnosed-by: Eryu Guan <eguan@redhat.com>
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
7 years agofs: invalidate page cache after end_io() in dio completion
Eryu Guan [Fri, 13 Oct 2017 16:47:46 +0000 (09:47 -0700)]
fs: invalidate page cache after end_io() in dio completion

Commit 332391a9935d ("fs: Fix page cache inconsistency when mixing
buffered and AIO DIO") moved page cache invalidation from
iomap_dio_rw() to iomap_dio_complete() for iomap based direct write
path, but before the dio->end_io() call, and it re-introdued the bug
fixed by commit c771c14baa33 ("iomap: invalidate page caches should
be after iomap_dio_complete() in direct write").

I found this because fstests generic/418 started failing on XFS with
v4.14-rc3 kernel, which is the regression test for this specific
bug.

So similarly, fix it by moving dio->end_io() (which does the
unwritten extent conversion) before page cache invalidation, to make
sure next buffer read reads the final real allocations not unwritten
extents. I also add some comments about why should end_io() go first
in case we get it wrong again in the future.

Note that, there's no such problem in the non-iomap based direct
write path, because we didn't remove the page cache invalidation
after the ->direct_IO() in generic_file_direct_write() call, but I
decided to fix dio_complete() too so we don't leave a landmine
there, also be consistent with iomap_dio_complete().

Fixes: 332391a9935d ("fs: Fix page cache inconsistency when mixing buffered and AIO DIO")
Signed-off-by: Eryu Guan <eguan@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Lukas Czerner <lczerner@redhat.com>
7 years agoxfs: cancel dirty pages on invalidation
Dave Chinner [Fri, 13 Oct 2017 16:47:45 +0000 (09:47 -0700)]
xfs: cancel dirty pages on invalidation

Recently we've had warnings arise from the vm handing us pages
without bufferheads attached to them. This should not ever occur
in XFS, but we don't defend against it properly if it does. The only
place where we remove bufferheads from a page is in
xfs_vm_releasepage(), but we can't tell the difference here between
"page is dirty so don't release" and "page is dirty but is being
invalidated so release it".

In some places that are invalidating pages ask for pages to be
released and follow up afterward calling ->releasepage by checking
whether the page was dirty and then aborting the invalidation. This
is a possible vector for releasing buffers from a page but then
leaving it in the mapping, so we really do need to avoid dirty pages
in xfs_vm_releasepage().

To differentiate between invalidated pages and normal pages, we need
to clear the page dirty flag when invalidating the pages. This can
be done through xfs_vm_invalidatepage(), and will result
xfs_vm_releasepage() seeing the page as clean which matches the
bufferhead state on the page after calling block_invalidatepage().

Hence we can re-add the page dirty check in xfs_vm_releasepage to
catch the case where we might be releasing a page that is actually
dirty and so should not have the bufferheads on it removed. This
will remove one possible vector of "dirty page with no bufferheads"
and so help narrow down the search for the root cause of that
problem.

Signed-Off-By: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
7 years agoALSA: usb-audio: Add native DSD support for Pro-Ject Pre Box S2 Digital
Jussi Laako [Sun, 15 Oct 2017 09:41:32 +0000 (12:41 +0300)]
ALSA: usb-audio: Add native DSD support for Pro-Ject Pre Box S2 Digital

Add native DSD support quirk for Pro-Ject Pre Box S2 Digital USB id
2772:0230.

Signed-off-by: Jussi Laako <jussi@sonarnerd.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
7 years agoDocumentation: Add a file explaining the Linux kernel license enforcement policy
Greg Kroah-Hartman [Wed, 4 Oct 2017 14:17:55 +0000 (16:17 +0200)]
Documentation: Add a file explaining the Linux kernel license enforcement policy

This adds a short document describing the views of how the Linux kernel
community feels about enforcing the license of the kernel.

Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Alex Elder (Linaro) <elder@linaro.org>
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Andy Gross <andy.gross@linaro.org>
Acked-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Anna Schumaker <schumaker.anna@gmail.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Acked-by: Bart Van Assche <bart.vanassche@wdc.com>
Acked-by: Bhumika Goyal <bhumirks@gmail.com>
Acked-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Acked-by: Borislav Petkov <bp@suse.de>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Acked-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Colin Ian King <colin.king@canonical.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Darrick J. Wong (Oracle) <darrick.wong@oracle.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Acked-by: David Kershner <david.kershner@unisys.com>
Acked-by: David S. Miller <davem@davemloft.net>
Acked-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Acked-by: Doug Ledford <dledford@redhat.com>
Acked-by: Fabio Estevam <festevam@gmail.com>
Acked-by: Felipe Balbi <balbi@kernel.org>
Acked-by: Florian Westphal <fw@strlen.de>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Hannes Reinecke <hare@suse.com>
Acked-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Acked-by: Heiko Stuebner <heiko@sntech.de>
Acked-by: Heiner Kallweit <hkallweit1@gmail.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Ivan Safonov <insafonov@gmail.com>
Acked-by: Jaegeuk Kim <jaegeuk@kernel.org>
Acked-by: Jan Kara (SUSE) <jack@suse.cz>
Acked-by: Javier Martinez Canillas <javier@dowhile0.org>
Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Acked-by: Jens Axboe <axboe@kernel.dk>
Acked-by: Jes Sorensen <Jes.Sorensen@gmail.com>
Acked-by: Jiri Kosina <jkosina@suse.cz>
Acked-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Joe Perches <joe@perches.com>
Acked-by: Joerg Roedel (SUSE) <jroedel@suse.de>
Acked-by: Johan Hovold <johan@kernel.org>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Juergen Gross <jgross@suse.com>
Acked-by: Julia Lawall <Julia.Lawall@lip6.fr>
Acked-by: K. Y. Srinivasan <kys@microsoft.com>
Acked-by: Khalid Aziz <khalid@gonehiking.org>
Acked-by: Krzysztof Kozlowski <krzk@kernel.org>
Acked-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Acked-by: Larry Finger <Larry.Finger@lwfinger.net>
Acked-by: Laura Abbott <laura@labbott.name>
Acked-by: Lee Jones <lee.jones@linaro.org>
Acked-by: Leon Romanovsky <leon@kernel.org>
Acked-by: Linus Walleij (Linaro) <linus.walleij@linaro.org>
Acked-by: Lv Zheng <zetalog@gmail.com>
Acked-by: Martin K. Petersen (Oracle) <martin.petersen@oracle.com>
Acked-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Masami Hiramatsu <mhiramat@kernel.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mike Marshall <hubcap@omnibond.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Neil Armstrong <narmstrong@baylibre.com>
Acked-by: Olof Johansson <olof@lixom.net>
Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Paul Burton <paul.burton@mips.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Rafael J. Wysocki <rafael@kernel.org>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Richard Weinberger <richard@nod.at>
Acked-by: Rik van Riel <riel@surriel.com>
Acked-by: Rob Clark <robdclark@gmail.com>
Acked-by: Rob Herring <robh@kernel.org>
Acked-by: Sebastian Reichel (Collabora) <sre@kernel.org>
Acked-by: Shawn Guo <shawnguo@kernel.org>
Acked-by: Shuah Khan <shuahkh@osg.samsung.com>
Acked-by: Simon Horman <horms@verge.net.au>
Acked-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Acked-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Acked-by: Sven Eckelmann <sven@narfation.org>
Acked-by: Takashi Iwai (SUSE) <tiwai@suse.de>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Thierry Reding <thierry.reding@gmail.com>
Acked-by: Tony Luck <tony.luck@gmail.com>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Vinod Koul <vkoul@kernel.org>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Acked-by: Wei Yongjun <weiyongjun1@huawei.com>
Acked-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
7 years agos390: fix zfcpdump-config
Dimitri John Ledkov [Thu, 12 Oct 2017 10:15:09 +0000 (11:15 +0100)]
s390: fix zfcpdump-config

zipl from s390-tools generates root=/dev/ram0 kernel cmdline for
zfcpdump, thus BLK_DEV_RAM is required.

zfcpdump initrd mounts DEBUG_FS, thus is also required.

Bug-Ubuntu: https://launchpad.net/bugs/1722735
Bug-Ubuntu: https://launchpad.net/bugs/1719290

Signed-off-by: Dimitri John Ledkov <xnox@ubuntu.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
7 years agos390/cputime: fix guest/irq/softirq times after CPU hotplug
Christian Borntraeger [Wed, 4 Oct 2017 12:46:17 +0000 (14:46 +0200)]
s390/cputime: fix guest/irq/softirq times after CPU hotplug

On CPU hotplug some cpu stats contain bogus values:

$ cat /proc/stat
cpu 0 0 49 1280 0 0 0 3 0 0
cpu0 0 0 49 618 0 0 0 3 0 0
cpu1 0 0 0 662 0 0 0 0 0 0
[...]
$ echo 0 > /sys/devices/system/cpu/cpu1/online
$ echo 1 > /sys/devices/system/cpu/cpu1/online
$ cat /proc/stat
cpu 0 0 49 3200 0 450359962737 450359962737 3 0 0
cpu0 0 0 49 1956 0 0 0 3 0 0
cpu1 0 0 0 1244 0 450359962737 450359962737 0 0 0
[...]

pcpu_attach_task() needs the same assignments as vtime_task_switch.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Fixes: b7394a5f4ce9 ("sched/cputime, s390: Implement delayed accounting of system time")
Cc: stable@vger.kernel.org # 4.11+
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
7 years agoLinux 4.14-rc5
Linus Torvalds [Mon, 16 Oct 2017 01:01:12 +0000 (21:01 -0400)]
Linux 4.14-rc5

7 years agoMerge tag 'char-misc-4.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Sun, 15 Oct 2017 11:50:38 +0000 (07:50 -0400)]
Merge tag 'char-misc-4.14-rc5' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are 4 patches to resolve some char/misc driver issues found these
  past weeks.

  One of them is a mei bugfix and another is a new mei device id. There
  is also a hyper-v fix for a reported issue, and a binder issue fix for
  a problem reported by a few people.

  All of these have been in my tree for a while, I don't know if
  linux-next is really testing much this month. But 0-day is happy with
  them :)"

* tag 'char-misc-4.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  binder: fix use-after-free in binder_transaction()
  Drivers: hv: vmbus: Fix bugs in rescind handling
  mei: me: add gemini lake devices id
  mei: always use domain runtime pm callbacks.

7 years agoMerge tag 'usb-4.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 15 Oct 2017 11:49:16 +0000 (07:49 -0400)]
Merge tag 'usb-4.14-rc5' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are a handful of USB driver fixes for 4.14-rc5.

  There is the "usual" usb-serial fixes and device ids, USB gadget
  fixes, and some more fixes found by the fuzz testing that is happening
  on the USB layer right now.

  All of these have been in my tree this week with no reported issues"

* tag 'usb-4.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: usbtest: fix NULL pointer dereference
  usb: gadget: configfs: Fix memory leak of interface directory data
  usb: gadget: composite: Fix use-after-free in usb_composite_overwrite_options
  usb: misc: usbtest: Fix overflow in usbtest_do_ioctl()
  usb: renesas_usbhs: Fix DMAC sequence for receiving zero-length packet
  USB: dummy-hcd: Fix deadlock caused by disconnect detection
  usb: phy: tegra: Fix phy suspend for UDC
  USB: serial: console: fix use-after-free after failed setup
  USB: serial: console: fix use-after-free on disconnect
  USB: serial: qcserial: add Dell DW5818, DW5819
  USB: serial: cp210x: add support for ELV TFD500
  USB: serial: cp210x: fix partnum regression
  USB: serial: option: add support for TP-Link LTE module
  USB: serial: ftdi_sio: add id for Cypress WICED dev board

7 years agoMerge tag 'dmaengine-fix-4.14-rc5' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Sun, 15 Oct 2017 11:47:07 +0000 (07:47 -0400)]
Merge tag 'dmaengine-fix-4.14-rc5' of git://git.infradead.org/users/vkoul/slave-dma

Pull dmaengine fixes from Vinod Koul:
 "Here are fixes for this round

   - fix spinlock usage amd fifo response for altera driver

   - fix ti crossbar race condition

   - fix edma memcpy align"

* tag 'dmaengine-fix-4.14-rc5' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: altera: fix spinlock usage
  dmaengine: altera: fix response FIFO emptying
  dmaengine: ti-dma-crossbar: Fix possible race condition with dma_inuse
  dmaengine: edma: Align the memcpy acnt array size with the transfer

7 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 14 Oct 2017 19:26:38 +0000 (15:26 -0400)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "A landry list of fixes:

   - fix reboot breakage on some PCID-enabled system

   - fix crashes/hangs on some PCID-enabled systems

   - fix microcode loading on certain older CPUs

   - various unwinder fixes

   - extend an APIC quirk to more hardware systems and disable APIC
     related warning on virtualized systems

   - various Hyper-V fixes

   - a macro definition robustness fix

   - remove jprobes IRQ disabling

   - various mem-encryption fixes"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/microcode: Do the family check first
  x86/mm: Flush more aggressively in lazy TLB mode
  x86/apic: Update TSC_DEADLINE quirk with additional SKX stepping
  x86/apic: Silence "FW_BUG TSC_DEADLINE disabled due to Errata" on hypervisors
  x86/mm: Disable various instrumentations of mm/mem_encrypt.c and mm/tlb.c
  x86/hyperv: Fix hypercalls with extended CPU ranges for TLB flushing
  x86/hyperv: Don't use percpu areas for pcpu_flush/pcpu_flush_ex structures
  x86/hyperv: Clear vCPU banks between calls to avoid flushing unneeded vCPUs
  x86/unwind: Disable unwinder warnings on 32-bit
  x86/unwind: Align stack pointer in unwinder dump
  x86/unwind: Use MSB for frame pointer encoding on 32-bit
  x86/unwind: Fix dereference of untrusted pointer
  x86/alternatives: Fix alt_max_short macro to really be a max()
  x86/mm/64: Fix reboot interaction with CR4.PCIDE
  kprobes/x86: Remove IRQ disabling from jprobe handlers
  kprobes/x86: Set up frame pointer in kprobe trampoline

7 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 14 Oct 2017 19:20:38 +0000 (15:20 -0400)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "Three fixes that address an SMP balancing performance regression"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/core: Ensure load_balance() respects the active_mask
  sched/core: Address more wake_affine() regressions
  sched/core: Fix wake_affine() performance regression

7 years agoMerge branch 'ras-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 14 Oct 2017 19:19:11 +0000 (15:19 -0400)]
Merge branch 'ras-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull RAS fixes from Ingo Molnar:
 "A boot parameter fix, plus a header export fix"

* 'ras-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mce: Hide mca_cfg
  RAS/CEC: Use the right length for "cec_disable"

7 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 14 Oct 2017 19:16:49 +0000 (15:16 -0400)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "Some tooling fixes plus three kernel fixes: a memory leak fix, a
  statistics fix and a crash fix"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/intel/uncore: Fix memory leaks on allocation failures
  perf/core: Fix cgroup time when scheduling descendants
  perf/core: Avoid freeing static PMU contexts when PMU is unregistered
  tools include uapi bpf.h: Sync kernel ABI header with tooling header
  perf pmu: Unbreak perf record for arm/arm64 with events with explicit PMU
  perf script: Add missing separator for "-F ip,brstack" (and brstackoff)
  perf callchain: Compare dsos (as well) for CCKEY_FUNCTION

7 years agoMerge branch 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 14 Oct 2017 19:14:20 +0000 (15:14 -0400)]
Merge branch 'locking-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull locking fixes from Ingo Molnar:
 "Two lockdep fixes for bugs introduced by the cross-release dependency
  tracking feature - plus a commit that disables it because performance
  regressed in an absymal fashion on some systems"

* 'locking-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  locking/lockdep: Disable cross-release features for now
  locking/selftest: Avoid false BUG report
  locking/lockdep: Fix stacktrace mess

7 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 14 Oct 2017 19:11:21 +0000 (15:11 -0400)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fixes from Ingo Molnar:
 "A CPU hotplug related fix, plus two related sanity checks"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq/cpuhotplug: Enforce affinity setting on startup of managed irqs
  genirq/cpuhotplug: Add sanity check for effective affinity mask
  genirq: Warn when effective affinity is not updated

7 years agoMerge branch 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 14 Oct 2017 19:09:08 +0000 (15:09 -0400)]
Merge branch 'core-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull objtool fix from Ingo Molnar:
 "A single objtool fix: avoid silently broken ORC debuginfo builds and
  error out instead"

* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  objtool: Upgrade libelf-devel warning to error for CONFIG_ORC_UNWINDER

7 years agox86/microcode: Do the family check first
Borislav Petkov [Thu, 12 Oct 2017 11:23:16 +0000 (13:23 +0200)]
x86/microcode: Do the family check first

On CPUs like AMD's Geode, for example, we shouldn't even try to load
microcode because they do not support the modern microcode loading
interface.

However, we do the family check *after* the other checks whether the
loader has been disabled on the command line or whether we're running in
a guest.

So move the family checks first in order to exit early if we're being
loaded on an unsupported family.

Reported-and-tested-by: Sven Glodowski <glodi1@arcor.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: <stable@vger.kernel.org> # 4.11..
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://bugzilla.suse.com/show_bug.cgi?id=1061396
Link: http://lkml.kernel.org/r/20171012112316.977-1-bp@alien8.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agolocking/lockdep: Disable cross-release features for now
Ingo Molnar [Sat, 14 Oct 2017 07:26:59 +0000 (09:26 +0200)]
locking/lockdep: Disable cross-release features for now

Johan Hovold reported a big lockdep slowdown on his system, caused by lockdep:

> I had noticed that the BeagleBone Black boot time appeared to have
> increased significantly with 4.14 and yesterday I finally had time to
> investigate it.
>
> Boot time (from "Linux version" to login prompt) had in fact doubled
> since 4.13 where it took 17 seconds (with my current config) compared to
> the 35 seconds I now see with 4.14-rc4.
>
> I quick bisect pointed to lockdep and specifically the following commit:
>
> 28a903f63ec0 ("locking/lockdep: Handle non(or multi)-acquisition of a crosslock")

Because the final v4.14 release is close, disable the cross-release lockdep
features for now.

Bisected-by: Johan Hovold <johan@kernel.org>
Debugged-by: Johan Hovold <johan@kernel.org>
Reported-by: Johan Hovold <johan@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Byungchul Park <byungchul.park@lge.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tony Lindgren <tony@atomide.com>
Cc: kernel-team@lge.com
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-mm@kvack.org
Cc: linux-omap@vger.kernel.org
Link: http://lkml.kernel.org/r/20171014072659.f2yr6mhm5ha3eou7@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agoMerge branch '4.14-fixes' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus
Linus Torvalds [Sat, 14 Oct 2017 10:49:15 +0000 (06:49 -0400)]
Merge branch '4.14-fixes' of git://git.linux-mips.org/ralf/upstream-linus

Pull MIPS fixes from Ralf Baechle:
 "More MIPS fixes for 4.14:

   - Loongson 1: Set the default number of RX and TX queues to
     accomodate for recent changes of stmmac driver.

   - BPF: Fix uninitialised target compiler error.

   - Fix cmpxchg on 32 bit signed ints for 64 bit kernels with
     !kernel_uses_llsc

   - Fix generic-board-config.sh for builds using O=

   - Remove pr_err() calls from fpu_emu() for a case which is not a
     kernel error"

* '4.14-fixes' of git://git.linux-mips.org/pub/scm/ralf/upstream-linus:
  MIPS: math-emu: Remove pr_err() calls from fpu_emu()
  MIPS: Fix generic-board-config.sh for builds using O=
  MIPS: Fix cmpxchg on 32b signed ints for 64b kernel with !kernel_uses_llsc
  MIPS: loongson1: set default number of rx and tx queues for stmmac
  MIPS: bpf: Fix uninitialised target compiler error

7 years agox86/mm: Flush more aggressively in lazy TLB mode
Andy Lutomirski [Mon, 9 Oct 2017 16:50:49 +0000 (09:50 -0700)]
x86/mm: Flush more aggressively in lazy TLB mode

Since commit:

  94b1b03b519b ("x86/mm: Rework lazy TLB mode and TLB freshness tracking")

x86's lazy TLB mode has been all the way lazy: when running a kernel thread
(including the idle thread), the kernel keeps using the last user mm's
page tables without attempting to maintain user TLB coherence at all.

From a pure semantic perspective, this is fine -- kernel threads won't
attempt to access user pages, so having stale TLB entries doesn't matter.

Unfortunately, I forgot about a subtlety.  By skipping TLB flushes,
we also allow any paging-structure caches that may exist on the CPU
to become incoherent.  This means that we can have a
paging-structure cache entry that references a freed page table, and
the CPU is within its rights to do a speculative page walk starting
at the freed page table.

I can imagine this causing two different problems:

 - A speculative page walk starting from a bogus page table could read
   IO addresses.  I haven't seen any reports of this causing problems.

 - A speculative page walk that involves a bogus page table can install
   garbage in the TLB.  Such garbage would always be at a user VA, but
   some AMD CPUs have logic that triggers a machine check when it notices
   these bogus entries.  I've seen a couple reports of this.

Boris further explains the failure mode:

> It is actually more of an optimization which assumes that paging-structure
> entries are in WB DRAM:
>
> "TlbCacheDis: cacheable memory disable. Read-write. 0=Enables
> performance optimization that assumes PML4, PDP, PDE, and PTE entries
> are in cacheable WB-DRAM; memory type checks may be bypassed, and
> addresses outside of WB-DRAM may result in undefined behavior or NB
> protocol errors. 1=Disables performance optimization and allows PML4,
> PDP, PDE and PTE entries to be in any memory type. Operating systems
> that maintain page tables in memory types other than WB- DRAM must set
> TlbCacheDis to insure proper operation."
>
> The MCE generated is an NB protocol error to signal that
>
> "Link: A specific coherent-only packet from a CPU was issued to an
> IO link. This may be caused by software which addresses page table
> structures in a memory type other than cacheable WB-DRAM without
> properly configuring MSRC001_0015[TlbCacheDis]. This may occur, for
> example, when page table structure addresses are above top of memory. In
> such cases, the NB will generate an MCE if it sees a mismatch between
> the memory operation generated by the core and the link type."
>
> I'm assuming coherent-only packets don't go out on IO links, thus the
> error.

To fix this, reinstate TLB coherence in lazy mode.  With this patch
applied, we do it in one of two ways:

 - If we have PCID, we simply switch back to init_mm's page tables
   when we enter a kernel thread -- this seems to be quite cheap
   except for the cost of serializing the CPU.

 - If we don't have PCID, then we set a flag and switch to init_mm
   the first time we would otherwise need to flush the TLB.

The /sys/kernel/debug/x86/tlb_use_lazy_mode debug switch can be changed
to override the default mode for benchmarking.

In theory, we could optimize this better by only flushing the TLB in
lazy CPUs when a page table is freed.  Doing that would require
auditing the mm code to make sure that all page table freeing goes
through tlb_remove_page() as well as reworking some data structures
to implement the improved flush logic.

Reported-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Reported-by: Adam Borowski <kilobyte@angband.pl>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Johannes Hirte <johannes.hirte@datenkhaos.de>
Cc: Kees Cook <keescook@chromium.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Roman Kagan <rkagan@virtuozzo.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 94b1b03b519b ("x86/mm: Rework lazy TLB mode and TLB freshness tracking")
Link: http://lkml.kernel.org/r/20171009170231.fkpraqokz6e4zeco@pd.tnic
Signed-off-by: Ingo Molnar <mingo@kernel.org>
7 years agoMerge tag 'drm-fixes-for-v4.14-rc5' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sat, 14 Oct 2017 01:10:35 +0000 (18:10 -0700)]
Merge tag 'drm-fixes-for-v4.14-rc5' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Couple of the arm people seem to wake up so this has imx and msm
  fixes, along with a bunch of i915 stable bounds fixes and an amdgpu
  regression fix.

  All seems pretty okay for now"

* tag 'drm-fixes-for-v4.14-rc5' of git://people.freedesktop.org/~airlied/linux:
  drm/msm: fix _NO_IMPLICIT fencing case
  drm/msm: fix error path cleanup
  drm/msm/mdp5: Remove extra pm_runtime_put call in mdp5_crtc_cursor_set()
  drm/msm/dsi: Use correct pm_runtime_put variant during host_init
  drm/msm: fix return value check in _msm_gem_kernel_new()
  drm/msm: use proper memory barriers for updating tail/head
  drm/msm/mdp5: add missing max size for 8x74 v1
  drm/amdgpu: fix placement flags in amdgpu_ttm_bind
  drm/i915/bios: parse DDI ports also for CHV for HDMI DDC pin and DP AUX channel
  gpu: ipu-v3: pre: implement workaround for ERR009624
  gpu: ipu-v3: prg: wait for double buffers to be filled on channel startup
  gpu: ipu-v3: Allow channel burst locking on i.MX6 only
  drm/i915: Read timings from the correct transcoder in intel_crtc_mode_get()
  drm/i915: Order two completing nop_submit_request
  drm/i915: Silence compiler warning for hsw_power_well_enable()
  drm/i915: Use crtc_state_is_legacy_gamma in intel_color_check
  drm/i915/edp: Increase the T12 delay quirk to 1300ms
  drm/i915/edp: Get the Panel Power Off timestamp after panel is off
  sync_file: Return consistent status in SYNC_IOC_FILE_INFO
  drm/atomic: Unref duplicated drm_atomic_state in drm_atomic_helper_resume()

7 years agoMerge tag 'drm-intel-fixes-2017-10-11' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Fri, 13 Oct 2017 23:59:20 +0000 (09:59 +1000)]
Merge tag 'drm-intel-fixes-2017-10-11' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

drm/i915 fixes for 4.14-rc5:

Three fixes for stable:

- Use crtc_state_is_legacy_gamma in intel_color_check (Maarten)
- Read timings from the correct transcoder (Ville).
- Fix HDMI on BSW (Jani).

Other fixes:

- eDP fixes (Manasi)
- Silence compiler warnings (Chris)
- Order two completing nop_submit_request (Chris)

* tag 'drm-intel-fixes-2017-10-11' of git://anongit.freedesktop.org/drm/drm-intel:
  drm/i915/bios: parse DDI ports also for CHV for HDMI DDC pin and DP AUX channel
  drm/i915: Read timings from the correct transcoder in intel_crtc_mode_get()
  drm/i915: Order two completing nop_submit_request
  drm/i915: Silence compiler warning for hsw_power_well_enable()
  drm/i915: Use crtc_state_is_legacy_gamma in intel_color_check
  drm/i915/edp: Increase the T12 delay quirk to 1300ms
  drm/i915/edp: Get the Panel Power Off timestamp after panel is off

7 years agoMerge branch 'msm-fixes-4.14-rc4' of git://people.freedesktop.org/~robclark/linux...
Dave Airlie [Fri, 13 Oct 2017 23:38:49 +0000 (09:38 +1000)]
Merge branch 'msm-fixes-4.14-rc4' of git://people.freedesktop.org/~robclark/linux into drm-fixes

bunch of msm fixes

* 'msm-fixes-4.14-rc4' of git://people.freedesktop.org/~robclark/linux:
  drm/msm: fix _NO_IMPLICIT fencing case
  drm/msm: fix error path cleanup
  drm/msm/mdp5: Remove extra pm_runtime_put call in mdp5_crtc_cursor_set()
  drm/msm/dsi: Use correct pm_runtime_put variant during host_init
  drm/msm: fix return value check in _msm_gem_kernel_new()
  drm/msm: use proper memory barriers for updating tail/head
  drm/msm/mdp5: add missing max size for 8x74 v1

7 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 13 Oct 2017 23:29:08 +0000 (16:29 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "18 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  mm, swap: use page-cluster as max window of VMA based swap readahead
  mm: page_vma_mapped: ensure pmd is loaded with READ_ONCE outside of lock
  kmemleak: clear stale pointers from task stacks
  fs/binfmt_misc.c: node could be NULL when evicting inode
  fs/mpage.c: fix mpage_writepage() for pages with buffers
  linux/kernel.h: add/correct kernel-doc notation
  tty: fall back to N_NULL if switching to N_TTY fails during hangup
  Revert "vmalloc: back off when the current task is killed"
  mm/cma.c: take __GFP_NOWARN into account in cma_alloc()
  scripts/kallsyms.c: ignore symbol type 'n'
  userfaultfd: selftest: exercise -EEXIST only in background transfer
  mm: only display online cpus of the numa node
  mm: remove unnecessary WARN_ONCE in page_vma_mapped_walk().
  mm/mempolicy: fix NUMA_INTERLEAVE_HIT counter
  include/linux/of.h: provide of_n_{addr,size}_cells wrappers for !CONFIG_OF
  mm/madvise.c: add description for MADV_WIPEONFORK and MADV_KEEPONFORK
  lib/Kconfig.debug: kernel hacking menu: runtime testing: keep tests together
  mm/migrate: fix indexing bug (off by one) and avoid out of bound access

7 years agomm, swap: use page-cluster as max window of VMA based swap readahead
Huang Ying [Fri, 13 Oct 2017 22:58:29 +0000 (15:58 -0700)]
mm, swap: use page-cluster as max window of VMA based swap readahead

When the VMA based swap readahead was introduced, a new knob

  /sys/kernel/mm/swap/vma_ra_max_order

was added as the max window of VMA swap readahead.  This is to make it
possible to use different max window for VMA based readahead and
original physical readahead.  But Minchan Kim pointed out that this will
cause a regression because setting page-cluster sysctl to zero cannot
disable swap readahead with the change.

To fix the regression, the page-cluster sysctl is used as the max window
of both the VMA based swap readahead and original physical swap
readahead.  If more fine grained control is needed in the future, more
knobs can be added as the subordinate knobs of the page-cluster sysctl.

The vma_ra_max_order knob is deleted.  Because the knob was introduced
in v4.14-rc1, and this patch is targeting being merged before v4.14
releasing, there should be no existing users of this newly added ABI.

Link: http://lkml.kernel.org/r/20171011070847.16003-1-ying.huang@intel.com
Fixes: ec560175c0b6fce ("mm, swap: VMA based swap readahead")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reported-by: Minchan Kim <minchan@kernel.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Shaohua Li <shli@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Tim Chen <tim.c.chen@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: page_vma_mapped: ensure pmd is loaded with READ_ONCE outside of lock
Will Deacon [Fri, 13 Oct 2017 22:58:25 +0000 (15:58 -0700)]
mm: page_vma_mapped: ensure pmd is loaded with READ_ONCE outside of lock

Loading the pmd without holding the pmd_lock exposes us to races with
concurrent updaters of the page tables but, worse still, it also allows
the compiler to cache the pmd value in a register and reuse it later on,
even if we've performed a READ_ONCE in between and seen a more recent
value.

In the case of page_vma_mapped_walk, this leads to the following crash
when the pmd loaded for the initial pmd_trans_huge check is all zeroes
and a subsequent valid table entry is loaded by check_pmd.  We then
proceed into map_pte, but the compiler re-uses the zero entry inside
pte_offset_map, resulting in a junk pointer being installed in
pvmw->pte:

  PC is at check_pte+0x20/0x170
  LR is at page_vma_mapped_walk+0x2e0/0x540
  [...]
  Process doio (pid: 2463, stack limit = 0xffff00000f2e8000)
  Call trace:
    check_pte+0x20/0x170
    page_vma_mapped_walk+0x2e0/0x540
    page_mkclean_one+0xac/0x278
    rmap_walk_file+0xf0/0x238
    rmap_walk+0x64/0xa0
    page_mkclean+0x90/0xa8
    clear_page_dirty_for_io+0x84/0x2a8
    mpage_submit_page+0x34/0x98
    mpage_process_page_bufs+0x164/0x170
    mpage_prepare_extent_to_map+0x134/0x2b8
    ext4_writepages+0x484/0xe30
    do_writepages+0x44/0xe8
    __filemap_fdatawrite_range+0xbc/0x110
    file_write_and_wait_range+0x48/0xd8
    ext4_sync_file+0x80/0x4b8
    vfs_fsync_range+0x64/0xc0
    SyS_msync+0x194/0x1e8

This patch fixes the problem by ensuring that READ_ONCE is used before
the initial checks on the pmd, and this value is subsequently used when
checking whether or not the pmd is present.  pmd_check is removed and
the pmd_present check is inlined directly.

Link: http://lkml.kernel.org/r/1507222630-5839-1-git-send-email-will.deacon@arm.com
Fixes: f27176cfc363 ("mm: convert page_mkclean_one() to use page_vma_mapped_walk()")
Signed-off-by: Will Deacon <will.deacon@arm.com>
Tested-by: Yury Norov <ynorov@caviumnetworks.com>
Tested-by: Richard Ruigrok <rruigrok@codeaurora.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agokmemleak: clear stale pointers from task stacks
Konstantin Khlebnikov [Fri, 13 Oct 2017 22:58:22 +0000 (15:58 -0700)]
kmemleak: clear stale pointers from task stacks

Kmemleak considers any pointers on task stacks as references.  This
patch clears newly allocated and reused vmap stacks.

Link: http://lkml.kernel.org/r/150728990124.744199.8403409836394318684.stgit@buzz
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agofs/binfmt_misc.c: node could be NULL when evicting inode
Eryu Guan [Fri, 13 Oct 2017 22:58:18 +0000 (15:58 -0700)]
fs/binfmt_misc.c: node could be NULL when evicting inode

inode->i_private is assigned by a Node pointer only after registering a
new binary format, so it could be NULL if inode was created by
bm_fill_super() (or iput() was called by the error path in
bm_register_write()), and this could result in NULL pointer dereference
when evicting such an inode.  e.g.  mount binfmt_misc filesystem then
umount it immediately:

  mount -t binfmt_misc binfmt_misc /proc/sys/fs/binfmt_misc
  umount /proc/sys/fs/binfmt_misc

will result in

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000013
  IP: bm_evict_inode+0x16/0x40 [binfmt_misc]
  ...
  Call Trace:
   evict+0xd3/0x1a0
   iput+0x17d/0x1d0
   dentry_unlink_inode+0xb9/0xf0
   __dentry_kill+0xc7/0x170
   shrink_dentry_list+0x122/0x280
   shrink_dcache_parent+0x39/0x90
   do_one_tree+0x12/0x40
   shrink_dcache_for_umount+0x2d/0x90
   generic_shutdown_super+0x1f/0x120
   kill_litter_super+0x29/0x40
   deactivate_locked_super+0x43/0x70
   deactivate_super+0x45/0x60
   cleanup_mnt+0x3f/0x70
   __cleanup_mnt+0x12/0x20
   task_work_run+0x86/0xa0
   exit_to_usermode_loop+0x6d/0x99
   syscall_return_slowpath+0xba/0xf0
   entry_SYSCALL_64_fastpath+0xa3/0xa

Fix it by making sure Node (e) is not NULL.

Link: http://lkml.kernel.org/r/20171010100642.31786-1-eguan@redhat.com
Fixes: 83f918274e4b ("exec: binfmt_misc: shift filp_close(interp_file) from kill_node() to bm_evict_inode()")
Signed-off-by: Eryu Guan <eguan@redhat.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agofs/mpage.c: fix mpage_writepage() for pages with buffers
Matthew Wilcox [Fri, 13 Oct 2017 22:58:15 +0000 (15:58 -0700)]
fs/mpage.c: fix mpage_writepage() for pages with buffers

When using FAT on a block device which supports rw_page, we can hit
BUG_ON(!PageLocked(page)) in try_to_free_buffers().  This is because we
call clean_buffers() after unlocking the page we've written.  Introduce
a new clean_page_buffers() which cleans all buffers associated with a
page and call it from within bdev_write_page().

[akpm@linux-foundation.org: s/PAGE_SIZE/~0U/ per Linus and Matthew]
Link: http://lkml.kernel.org/r/20171006211541.GA7409@bombadil.infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Reported-by: Toshi Kani <toshi.kani@hpe.com>
Reported-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Tested-by: Toshi Kani <toshi.kani@hpe.com>
Acked-by: Johannes Thumshirn <jthumshirn@suse.de>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agolinux/kernel.h: add/correct kernel-doc notation
Randy Dunlap [Fri, 13 Oct 2017 22:58:11 +0000 (15:58 -0700)]
linux/kernel.h: add/correct kernel-doc notation

Add kernel-doc notation for some macros.  Correct kernel-doc comments &
typos for a few macros.

Link: http://lkml.kernel.org/r/76fa1403-1511-be4c-e9c4-456b43edfad3@infradead.org
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agotty: fall back to N_NULL if switching to N_TTY fails during hangup
Johannes Weiner [Fri, 13 Oct 2017 22:58:08 +0000 (15:58 -0700)]
tty: fall back to N_NULL if switching to N_TTY fails during hangup

We have seen NULL-pointer dereference crashes in tty->disc_data when the
N_TTY fallback driver failed to open during hangup.  The immediate cause
of this open to fail has been addressed in the preceding patch to
vmalloc(), but this code could be more robust.

As Alan pointed out in commit 8a8dabf2dd68 ("tty: handle the case where
we cannot restore a line discipline"), the N_TTY driver, historically
the safe fallback that could never fail, can indeed fail, but the
surrounding code is not prepared to handle this.  To avoid crashes he
added a new N_NULL driver to take N_TTY's place as the last resort.

Hook that fallback up to the hangup path.  Update tty_ldisc_reinit() to
reflect the reality that n_tty_open can indeed fail.

Link: http://lkml.kernel.org/r/20171004185959.GC2136@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Alan Cox <alan@llwyncelyn.cymru>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoRevert "vmalloc: back off when the current task is killed"
Johannes Weiner [Fri, 13 Oct 2017 22:58:05 +0000 (15:58 -0700)]
Revert "vmalloc: back off when the current task is killed"

This reverts commits 5d17a73a2ebe ("vmalloc: back off when the current
task is killed") and 171012f56127 ("mm: don't warn when vmalloc() fails
due to a fatal signal").

Commit 5d17a73a2ebe ("vmalloc: back off when the current task is
killed") made all vmalloc allocations from a signal-killed task fail.
We have seen crashes in the tty driver from this, where a killed task
exiting tries to switch back to N_TTY, fails n_tty_open because of the
vmalloc failing, and later crashes when dereferencing tty->disc_data.

Arguably, relying on a vmalloc() call to succeed in order to properly
exit a task is not the most robust way of doing things.  There will be a
follow-up patch to the tty code to fall back to the N_NULL ldisc.

But the justification to make that vmalloc() call fail like this isn't
convincing, either.  The patch mentions an OOM victim exhausting the
memory reserves and thus deadlocking the machine.  But the OOM killer is
only one, improbable source of fatal signals.  It doesn't make sense to
fail allocations preemptively with plenty of memory in most cases.

The patch doesn't mention real-life instances where vmalloc sites would
exhaust memory, which makes it sound more like a theoretical issue to
begin with.  But just in case, the OOM access to memory reserves has
been restricted on the allocator side in cd04ae1e2dc8 ("mm, oom: do not
rely on TIF_MEMDIE for memory reserves access"), which should take care
of any theoretical concerns on that front.

Revert this patch, and the follow-up that suppresses the allocation
warnings when we fail the allocations due to a signal.

Link: http://lkml.kernel.org/r/20171004185906.GB2136@cmpxchg.org
Fixes: 171012f56127 ("mm: don't warn when vmalloc() fails due to a fatal signal")
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Alan Cox <alan@llwyncelyn.cymru>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/cma.c: take __GFP_NOWARN into account in cma_alloc()
Boris Brezillon [Fri, 13 Oct 2017 22:58:01 +0000 (15:58 -0700)]
mm/cma.c: take __GFP_NOWARN into account in cma_alloc()

cma_alloc() unconditionally prints an INFO message when the CMA
allocation fails.  Make this message conditional on the non-presence of
__GFP_NOWARN in gfp_mask.

This patch aims at removing INFO messages that are displayed when the
VC4 driver tries to allocate buffer objects.  From the driver
perspective an allocation failure is acceptable, and the driver can
possibly do something to make following allocation succeed (like
flushing the VC4 internal cache).

Link: http://lkml.kernel.org/r/20171004125447.15195-1-boris.brezillon@free-electrons.com
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Acked-by: Laura Abbott <labbott@redhat.com>
Cc: Jaewon Kim <jaewon31.kim@samsung.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Eric Anholt <eric@anholt.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoscripts/kallsyms.c: ignore symbol type 'n'
Guenter Roeck [Fri, 13 Oct 2017 22:57:58 +0000 (15:57 -0700)]
scripts/kallsyms.c: ignore symbol type 'n'

gcc on aarch64 may emit synbols of type 'n' if the kernel is built with
'-frecord-gcc-switches'.  In most cases, those symbols are reported with
nm as

000000000000000e n $d

and with objdump as

0000000000000000 l    d  .GCC.command.line 0000000000000000 .GCC.command.line
000000000000000e l       .GCC.command.line 0000000000000000 $d

Those symbols are detected in is_arm_mapping_symbol() and ignored.
However, if "--prefix-symbols=<prefix>" is configured as well, the
situation is different.  For example, in efi/libstub, arm64 images are
built with

'--prefix-alloc-sections=.init --prefix-symbols=__efistub_'.

In combination with '-frecord-gcc-switches', the symbols are now reported
by nm as:

000000000000000e n __efistub_$d
and by objdump as:
0000000000000000 l    d  .GCC.command.line 0000000000000000 .GCC.command.line
000000000000000e l       .GCC.command.line 0000000000000000 __efistub_$d

Those symbols are no longer ignored and included in the base address
calculation.  This results in a base address of 000000000000000e, which
in turn causes kallsyms to abort with

    kallsyms failure:
relative symbol value 0xffffff900800a000 out of range in relative mode

The problem is seen in little endian arm64 builds with CONFIG_EFI
enabled and with '-frecord-gcc-switches' set in KCFLAGS.

Explicitly ignore symbols of type 'n' since those are clearly debug
symbols.

Link: http://lkml.kernel.org/r/1507136063-3139-1-git-send-email-linux@roeck-us.net
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agouserfaultfd: selftest: exercise -EEXIST only in background transfer
Andrea Arcangeli [Fri, 13 Oct 2017 22:57:54 +0000 (15:57 -0700)]
userfaultfd: selftest: exercise -EEXIST only in background transfer

I was stress testing some backports and with high load, after some time,
the latest version of the selftest showed some false positive in
connection with the uffdio_copy_retry.  This seems to fix it while still
exercising -EEXIST in the background transfer once in a while.

The fork child will quit after the last UFFDIO_COPY is run, so a
repeated UFFDIO_COPY may not return -EEXIST.  This change restricts the
-EEXIST stress to the background transfer where the memory can't go away
from under it.

Also updated uffdio_zeropage, so the interface is consistent.

Link: http://lkml.kernel.org/r/20171004171541.1495-2-aarcange@redhat.com
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Pavel Emelyanov <xemul@virtuozzo.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: only display online cpus of the numa node
Zhen Lei [Fri, 13 Oct 2017 22:57:50 +0000 (15:57 -0700)]
mm: only display online cpus of the numa node

When I execute numactl -H (which reads /sys/devices/system/node/nodeX/cpumap
and displays cpumask_of_node for each node), I get different result
on X86 and arm64.  For each numa node, the former only displayed online
CPUs, and the latter displayed all possible CPUs.  Unfortunately, both
Linux documentation and numactl manual have not described it clear.

I sent a mail to ask for help, and Michal Hocko replied that he
preferred to print online cpus because it doesn't really make much sense
to bind anything on offline nodes.

Will said:
 "I suspect the vast majority (if not all) code that reads this file was
  developed for x86, so having the same behaviour for arm64 sounds like
  something we should do ASAP before people try to special case with
  things like #ifdef __aarch64__. I'd rather have this in 4.14 if
  possible."

Link: http://lkml.kernel.org/r/1506678805-15392-2-git-send-email-thunder.leizhen@huawei.com
Signed-off-by: Zhen Lei <thunder.leizhen@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Tianhong Ding <dingtianhong@huawei.com>
Cc: Hanjun Guo <guohanjun@huawei.com>
Cc: Libin <huawei.libin@huawei.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm: remove unnecessary WARN_ONCE in page_vma_mapped_walk().
Zi Yan [Fri, 13 Oct 2017 22:57:47 +0000 (15:57 -0700)]
mm: remove unnecessary WARN_ONCE in page_vma_mapped_walk().

A non present pmd entry can appear after pmd_lock is taken in
page_vma_mapped_walk(), even if THP migration is not enabled.  The
WARN_ONCE is unnecessary.

Link: http://lkml.kernel.org/r/20171003142606.12324-1-zi.yan@sent.com
Fixes: 616b8371539a ("mm: thp: enable thp migration in generic path")
Signed-off-by: Zi Yan <zi.yan@cs.rutgers.edu>
Reported-by: Abdul Haleem <abdhalee@linux.vnet.ibm.com>
Tested-by: Abdul Haleem <abdhalee@linux.vnet.ibm.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/mempolicy: fix NUMA_INTERLEAVE_HIT counter
Andrey Ryabinin [Fri, 13 Oct 2017 22:57:43 +0000 (15:57 -0700)]
mm/mempolicy: fix NUMA_INTERLEAVE_HIT counter

Commit 3a321d2a3dde ("mm: change the call sites of numa statistics
items") separated NUMA counters from zone counters, but the
NUMA_INTERLEAVE_HIT call site wasn't updated to use the new interface.
So alloc_page_interleave() actually increments NR_ZONE_INACTIVE_FILE
instead of NUMA_INTERLEAVE_HIT.

Fix this by using __inc_numa_state() interface to increment
NUMA_INTERLEAVE_HIT.

Link: http://lkml.kernel.org/r/20171003191003.8573-1-aryabinin@virtuozzo.com
Fixes: 3a321d2a3dde ("mm: change the call sites of numa statistics items")
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Kemi Wang <kemi.wang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoinclude/linux/of.h: provide of_n_{addr,size}_cells wrappers for !CONFIG_OF
Arnd Bergmann [Fri, 13 Oct 2017 22:57:40 +0000 (15:57 -0700)]
include/linux/of.h: provide of_n_{addr,size}_cells wrappers for !CONFIG_OF

The pci-rcar driver is enabled for compile tests, and this has shown that
the driver cannot build without CONFIG_OF, following the inclusion of
commit f8f2fe7355fb ("PCI: rcar: Use new OF interrupt mapping when possible"):

  drivers/pci/host/pcie-rcar.c: In function 'pci_dma_range_parser_init':
  drivers/pci/host/pcie-rcar.c:1039:2: error: implicit declaration of function 'of_n_addr_cells' [-Werror=implicit-function-declaration]
    parser->pna = of_n_addr_cells(node);
    ^

As pointed out by Ben Dooks and Geert Uytterhoeven, this is actually
supposed to build fine, which we can achieve if we make the declaration
of of_irq_parse_and_map_pci conditional on CONFIG_OF and provide an
empty inline function otherwise, as we do for a lot of other of
interfaces.

This lets us build the rcar_pci driver again without CONFIG_OF for build
testing.  All platforms using this driver select OF, so this doesn't
change anything for the users.

[akpm@linux-foundation.org: be consistent with surrounding code]
Link: http://lkml.kernel.org/r/20170911200805.3363318-1-arnd@arndb.de
Fixes: c25da4778803 ("PCI: rcar: Add Renesas R-Car PCIe driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Frank Rowand <frank.rowand@sony.com>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Magnus Damm <damm@opensource.se>
Cc: Ben Dooks <ben.dooks@codethink.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/madvise.c: add description for MADV_WIPEONFORK and MADV_KEEPONFORK
Yang Shi [Fri, 13 Oct 2017 22:57:37 +0000 (15:57 -0700)]
mm/madvise.c: add description for MADV_WIPEONFORK and MADV_KEEPONFORK

mm/madvise.c has a brief description about all MADV_ flags.  Add a
description for the newly added MADV_WIPEONFORK and MADV_KEEPONFORK.

Although man page has the similar information, but it'd better to keep
the consistent with other flags.

Link: http://lkml.kernel.org/r/1506117328-88228-1-git-send-email-yang.s@alibaba-inc.com
Signed-off-by: Yang Shi <yang.s@alibaba-inc.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agolib/Kconfig.debug: kernel hacking menu: runtime testing: keep tests together
Randy Dunlap [Fri, 13 Oct 2017 22:57:33 +0000 (15:57 -0700)]
lib/Kconfig.debug: kernel hacking menu: runtime testing: keep tests together

Expand the "Runtime testing" menu by including more entries inside it
instead of after it.  This is just Kconfig symbol movement.

This causes the (arch-independent) Runtime tests to be presented
(listed) all in one place instead of in multiple places.

Link: http://lkml.kernel.org/r/c194e5c4-2042-bf94-a2d8-7aa13756e257@infradead.org
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Kees Cook <keescook@chromium.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: "Luis R. Rodriguez" <mcgrof@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agomm/migrate: fix indexing bug (off by one) and avoid out of bound access
Mark Hairgrove [Fri, 13 Oct 2017 22:57:30 +0000 (15:57 -0700)]
mm/migrate: fix indexing bug (off by one) and avoid out of bound access

Index was incremented before last use and thus the second array could
dereference to an invalid address (not mentioning the fact that it did
not properly clear the entry we intended to clear).

Link: http://lkml.kernel.org/r/1506973525-16491-1-git-send-email-jglisse@redhat.com
Fixes: 8315ada7f095bf ("mm/migrate: allow migrate_vma() to alloc new page on empty entry")
Signed-off-by: Mark Hairgrove <mhairgrove@nvidia.com>
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
Cc: Reza Arbab <arbab@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agoMerge tag 'pci-v4.14-fixes-5' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Fri, 13 Oct 2017 20:28:19 +0000 (13:28 -0700)]
Merge tag 'pci-v4.14-fixes-5' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fixes from Bjorn Helgaas:

 - revert Tegra MSI change that caused a regression on TrimSlice
   (Thierry Reding)

 - finish Aardvark conversion to host bridge IRQ mapping. We missed
   Aardvark the first time around so it regressed (Thomas Petazzoni)

* tag 'pci-v4.14-fixes-5' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI: aardvark: Move to struct pci_host_bridge IRQ mapping functions
  Revert "PCI: tegra: Do not allocate MSI target memory"

7 years agoMerge tag 'iommu-fixes-v4.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 13 Oct 2017 18:49:38 +0000 (11:49 -0700)]
Merge tag 'iommu-fixes-v4.14-rc4' of git://git./linux/kernel/git/joro/iommu

Pull IOMMU fixes from Joerg Roedel:

 - keep an important data structure in the Exynos driver around after
   kernel-init to fix a kernel-oops

 - keep SWIOTLB enabled when SME is active in the AMD IOMMU driver

 - add a missing IOTLB sync to the AMD IOMMU driver

* tag 'iommu-fixes-v4.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/amd: Finish TLB flush in amd_iommu_unmap()
  iommu/exynos: Remove initconst attribute to avoid potential kernel oops
  iommu/amd: Do not disable SWIOTLB if SME is active

7 years agoMerge tag 'devicetree-fixes-for-4.14-2' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 13 Oct 2017 18:44:09 +0000 (11:44 -0700)]
Merge tag 'devicetree-fixes-for-4.14-2' of git://git./linux/kernel/git/robh/linux

Pull DeviceTree fixes from Rob Herring:

 - Fix memory leak in error case of of_console_check

 - Increase number of reserved memory regions to 32. 16 was not enough
   on some Power systems.

 - Fix OF node refcounting for of_fwnode_graph_get_port_parent

* tag 'devicetree-fixes-for-4.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  device property: preserve usecount for node passed to of_fwnode_graph_get_port_parent()
  drivers: of: increase MAX_RESERVED_REGIONS to 32
  of: do not leak console options

7 years agoMerge tag 'powerpc-4.14-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Fri, 13 Oct 2017 18:39:28 +0000 (11:39 -0700)]
Merge tag 'powerpc-4.14-5' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "A fix for a bad bug (written by me) in our livepatch handler. Removal
  of an over-zealous lockdep_assert_cpus_held() in our topology code. A
  fix to the recently added emulation of cntlz[wd]. And three small
  fixes to the recently added IMC PMU driver.

  Thanks to: Anju T Sudhakar, Balbir Singh, Kamalesh Babulal, Naveen N.
  Rao, Sandipan Das, Santosh Sivaraj, Thiago Jung Bauermann"

* tag 'powerpc-4.14-5' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/perf: Fix IMC initialization crash
  powerpc/perf: Add ___GFP_NOWARN flag to alloc_pages_node()
  powerpc/perf: Fix for core/nest imc call trace on cpuhotplug
  powerpc: Don't call lockdep_assert_cpus_held() from arch_update_cpu_topology()
  powerpc/lib/sstep: Fix count leading zeros instructions
  powerpc/livepatch: Fix livepatch stack access

7 years agoMerge tag 'for-linus-4.14c-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 13 Oct 2017 18:35:03 +0000 (11:35 -0700)]
Merge tag 'for-linus-4.14c-rc5-tag' of git://git./linux/kernel/git/xen/tip

Pull xen fixlet from Juergen Gross:
 "A minor fix correcting the cpu hotplug name for Xen guests"

* tag 'for-linus-4.14c-rc5-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
  xen/vcpu: Use a unified name about cpu hotplug state for pv and pvhvm

7 years agoiommu/amd: Finish TLB flush in amd_iommu_unmap()
Joerg Roedel [Fri, 13 Oct 2017 12:32:37 +0000 (14:32 +0200)]
iommu/amd: Finish TLB flush in amd_iommu_unmap()

The function only sends the flush command to the IOMMU(s),
but does not wait for its completion when it returns. Fix
that.

Fixes: 601367d76bd1 ('x86/amd-iommu: Remove iommu_flush_domain function')
Cc: stable@vger.kernel.org # >= 2.6.33
Signed-off-by: Joerg Roedel <jroedel@suse.de>
7 years agopowerpc/perf: Fix IMC initialization crash
Anju T Sudhakar [Fri, 13 Oct 2017 05:59:41 +0000 (11:29 +0530)]
powerpc/perf: Fix IMC initialization crash

Panic observed with latest firmware, and upstream kernel:

 NIP init_imc_pmu+0x8c/0xcf0
 LR  init_imc_pmu+0x2f8/0xcf0
 Call Trace:
   init_imc_pmu+0x2c8/0xcf0 (unreliable)
   opal_imc_counters_probe+0x300/0x400
   platform_drv_probe+0x64/0x110
   driver_probe_device+0x3d8/0x580
   __driver_attach+0x14c/0x1a0
   bus_for_each_dev+0x8c/0xf0
   driver_attach+0x34/0x50
   bus_add_driver+0x298/0x350
   driver_register+0x9c/0x180
   __platform_driver_register+0x5c/0x70
   opal_imc_driver_init+0x2c/0x40
   do_one_initcall+0x64/0x1d0
   kernel_init_freeable+0x280/0x374
   kernel_init+0x24/0x160
   ret_from_kernel_thread+0x5c/0x74

While registering nest imc at init, cpu-hotplug callback
nest_pmu_cpumask_init() makes an OPAL call to stop the engine. And if
the OPAL call fails, imc_common_cpuhp_mem_free() is invoked to cleanup
memory and cpuhotplug setup.

But when cleaning up the attribute group, we are dereferencing the
attribute element array without checking whether the backing element
is not NULL. This causes the kernel panic.

Add a check for the backing element prior to dereferencing the
attribute element, to handle the failing case gracefully.

Signed-off-by: Anju T Sudhakar <anju@linux.vnet.ibm.com>
Reported-by: Pridhiviraj Paidipeddi <ppaidipe@linux.vnet.ibm.com>
[mpe: Trim change log]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agoMerge tag 'imx-drm-fixes-2017-10-12' of git://git.pengutronix.de/git/pza/linux into...
Dave Airlie [Fri, 13 Oct 2017 07:32:30 +0000 (17:32 +1000)]
Merge tag 'imx-drm-fixes-2017-10-12' of git://git.pengutronix.de/git/pza/linux into drm-fixes

drm/imx: i.MX5 regression fix and i.MX6QP PRE/PRG stability fixes

- Disable channel burst locking on IPUv3EX (i.MX51) and IPUv3M (i.MX53).
  This fixes a regression introduced by commit 790cb4c7c954 ("drm/imx: lock
  scanout transfers for consecutive bursts").
- Give PRG a head start. Waiting for both double buffers to fill up before
  enabling the IPU improves startup reliability.
- Avoid PRE control register updates during unsafe window, workaround for
  ERR009624.

* tag 'imx-drm-fixes-2017-10-12' of git://git.pengutronix.de/git/pza/linux:
  gpu: ipu-v3: pre: implement workaround for ERR009624
  gpu: ipu-v3: prg: wait for double buffers to be filled on channel startup
  gpu: ipu-v3: Allow channel burst locking on i.MX6 only

7 years agoMerge branch 'drm-fixes-4.14' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Fri, 13 Oct 2017 06:23:34 +0000 (16:23 +1000)]
Merge branch 'drm-fixes-4.14' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

single amdgpu regression fix.

* 'drm-fixes-4.14' of git://people.freedesktop.org/~agd5f/linux:
  drm/amdgpu: fix placement flags in amdgpu_ttm_bind

7 years agoMerge tag 'devprop-4.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 13 Oct 2017 00:37:23 +0000 (17:37 -0700)]
Merge tag 'devprop-4.14-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull device properties framework fixes from Rafael Wysocki:
 "These fix an issue related to device removal introduced during the 4.9
  cycle and fix up new functionality added recently.

  Specifics:

   - Fix a device properties management issue, introduced during the 4.9
     cycle, that causes device properties associated with a parent
     device to go away on a removal of its child in some cases (Jarkko
     Nikula).

   - Fix inconsistencies in error codes returned by a new function
     helper in the device properties framework depending on the
     underlying low-level firmware interface, DT or ACPI, by making the
     meaning of error codes returned in the ACPI case agree with the
     meaning of DT error codes in analogous situations (Sakari Ailus)"

* tag 'devprop-4.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI: properties: Fix __acpi_node_get_property_reference() return codes
  ACPI: properties: Align return codes of __acpi_node_get_property_reference()
  device property: Track owner device of device property

7 years agoMerge tag 'xfs-4.14-fixes-5' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux
Linus Torvalds [Thu, 12 Oct 2017 21:51:13 +0000 (14:51 -0700)]
Merge tag 'xfs-4.14-fixes-5' of git://git./fs/xfs/xfs-linux

Pull xfs fixes from Darrick Wong:

 - Fix a stale kernel memory exposure when logging inodes.

 - Fix some build problems with CONFIG_XFS_RT=n

 - Don't change inode mode if the acl write fails, leaving the file
   totally inaccessible.

 - Fix a dangling pointer problem when removing an attr fork under
   memory pressure.

 - Don't crash while trying to invalidate a null buffer associated with
   a corrupt metadata pointer.

* tag 'xfs-4.14-fixes-5' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux:
  xfs: handle error if xfs_btree_get_bufs fails
  xfs: reinit btree pointer on attr tree inactivation walk
  xfs: Fix bool initialization/comparison
  xfs: don't change inode mode if ACL update fails
  xfs: move more RT specific code under CONFIG_XFS_RT
  xfs: Don't log uninitialised fields in inode structures

7 years agoscripts: fix faddr2line to work on last symbol
NeilBrown [Thu, 12 Oct 2017 03:22:04 +0000 (14:22 +1100)]
scripts: fix faddr2line to work on last symbol

If faddr2line is given a function name which is the last one listed by
"nm -n", it will fail because it never finds the next symbol.

So teach the awk script to catch that possibility, and use 'size' to
provide the end point of the last function.

Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
7 years agodrm/msm: fix _NO_IMPLICIT fencing case
Rob Clark [Tue, 12 Sep 2017 18:23:05 +0000 (14:23 -0400)]
drm/msm: fix _NO_IMPLICIT fencing case

We need to call reservation_object_reserve_shared() in both cases, but
this wasn't happening in the _NO_IMPLICIT submit case.

Fixes: f0a42bb ("drm/msm: submit support for in-fences")
Reported-by: Jordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
7 years agodrm/msm: fix error path cleanup
Rob Clark [Thu, 24 Aug 2017 18:24:29 +0000 (14:24 -0400)]
drm/msm: fix error path cleanup

If we fail to attach iommu, gpu->aspace could be IS_ERR()..

Signed-off-by: Rob Clark <robdclark@gmail.com>
7 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs
Linus Torvalds [Thu, 12 Oct 2017 17:56:06 +0000 (10:56 -0700)]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs

Pull quota fix from Jan Kara:
 "A fix for a regression in handling of quota grace times and warnings"

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  quota: Generate warnings for DQUOT_SPACE_NOFAIL allocations

7 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Thu, 12 Oct 2017 17:42:03 +0000 (10:42 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "Another latent bug related to PCID, an out-of-bounds access, and a
  submaintainer change being finally made official"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  MAINTAINERS: Add Paul Mackerras as maintainer for KVM/powerpc
  KVM: nVMX: fix guest CR4 loading when emulating L2 to L1 exit
  KVM: MMU: always terminate page walks at level 1
  KVM: nVMX: update last_nonleaf_level when initializing nested EPT

7 years agodevice property: preserve usecount for node passed to of_fwnode_graph_get_port_parent()
Niklas Söderlund [Tue, 22 Aug 2017 00:19:12 +0000 (02:19 +0200)]
device property: preserve usecount for node passed to of_fwnode_graph_get_port_parent()

Using CONFIG_OF_DYNAMIC=y uncovered an imbalance in the usecount of the
node being passed to of_fwnode_graph_get_port_parent(). Preserve the
usecount by using of_get_parent() instead of of_get_next_parent() which
don't decrement the usecount of the node passed to it.

Fixes: 3b27d00e7b6d7c88 ("device property: Move fwnode graph ops to firmware specific locations")
Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Rob Herring <robh@kernel.org>
7 years agodrivers: of: increase MAX_RESERVED_REGIONS to 32
Stewart Smith [Tue, 26 Sep 2017 08:40:00 +0000 (18:40 +1000)]
drivers: of: increase MAX_RESERVED_REGIONS to 32

There are two types of memory reservations firmware can ask the kernel
to make in the device tree: static and dynamic.
See Documentation/devicetree/bindings/reserved-memory/reserved-memory.txt

If you have greater than 16 entries in /reserved-memory (as we do on
POWER9 systems) you would get this scary looking error message:
 [    0.000000] OF: reserved mem: not enough space all defined regions.

This is harmless if all your reservations are static (which with OPAL on
POWER9, they are).

It is not harmless if you have any dynamic reservations after the 16th.

In the first pass over the fdt to find reservations, the child nodes of
/reserved-memory are added to a static array in of_reserved_mem.c so that
memory can be reserved in a 2nd pass. The array has 16 entries. This is why,
on my dual socket POWER9 system, I get that error 4 times with 20 static
reservations.

We don't have a problem on ppc though, as in arch/powerpc/kernel/prom.c
we look at the new style /reserved-ranges property to do reservations,
and this logic was introduced in 0962e8004e974 (well before any powernv
system shipped).

A Google search shows up no occurances of that exact error message, so we're
probably safe in that no machine that people use has memory not being reserved
when it should be.

The simple fix is to bump the length of the array to 32 which "should be
enough for everyone(TM)". The simple fix of not recording static allocations
in the array would cause problems for devices with "memory-region" properties.
A more future-proof fix is likely possible, although more invasive and this
simple fix is perfectly suitable in the meantime while a more future-proof
fix is developed.

Signed-off-by: Stewart Smith <stewart@linux.vnet.ibm.com>
Tested-by: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com>
Signed-off-by: Rob Herring <robh@kernel.org>
7 years agoof: do not leak console options
Sergey Senozhatsky [Tue, 26 Sep 2017 06:25:10 +0000 (15:25 +0900)]
of: do not leak console options

Do not strdup() console options. It seems that the only reason for
it to be strdup()-ed was a compilation warning: printk, UART and
console drivers, for some reason, expect char pointer instead of
const char pointer. So we can just pass `of_stdout_options', but
need to cast it to char pointer. A better fix would be to change
printk, console drivers and UART to accept const char `options';
but that will take time - there are lots of drivers to update.

The patch also fixes a possible memory leak: add_preferred_console()
can fail, but we don't kfree() options.

Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Rob Herring <robh@kernel.org>
7 years agodrm/msm/mdp5: Remove extra pm_runtime_put call in mdp5_crtc_cursor_set()
Archit Taneja [Fri, 6 Oct 2017 10:57:07 +0000 (16:27 +0530)]
drm/msm/mdp5: Remove extra pm_runtime_put call in mdp5_crtc_cursor_set()

While converting mdp5_enable/disable() calls to pm_runtime_get/put() API,
an extra call to pm_runtime_put_autosuspend() crept in
mdp5_crtc_cursor_set(). This results in calling the suspend handler
twice, and therefore clk_disables twice, which isn't a nice thing to do.

Fixes: d68fe15b1878 (drm/msm/mdp5: Use runtime PM get/put API instead ...)
Reported-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
7 years agodrm/msm/dsi: Use correct pm_runtime_put variant during host_init
Archit Taneja [Fri, 6 Oct 2017 10:57:06 +0000 (16:27 +0530)]
drm/msm/dsi: Use correct pm_runtime_put variant during host_init

The DSI runtime PM suspend/resume callbacks check whether
msm_host->cfg_hnd is non-NULL before trying to enable the bus clocks.
This is done to accommodate early calls to these functions that may
happen before the bus clocks are even initialized.

Calling pm_runtime_put_autosuspend() in dsi_host_init() can result in
racy behaviour since msm_host->cfg_hnd is set very soon after. If the
suspend callback happens too late, we end up trying to disable clocks
that were never enabled, resulting in a bunch of WARN_ON splats.

Use pm_runtime_put_sync() so that the suspend callback is called
immediately.

Reported-by: Nicolas Dechesne <nicolas.dechesne@linaro.org>
Signed-off-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
7 years agoMerge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Thu, 12 Oct 2017 16:32:53 +0000 (09:32 -0700)]
Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fixes from Herbert Xu:

 - fix crashes in skcipher/shash from zero-length input.

 - fix softirq GFP_KERNEL allocation in shash_setkey_unaligned.

 - error path bug fix in xts create function.

 - fix compiler warning regressions in axis and stm32

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: shash - Fix zero-length shash ahash digest crash
  crypto: skcipher - Fix crash on zero-length input
  crypto: shash - Fix a sleep-in-atomic bug in shash_setkey_unaligned
  crypto: xts - Fix an error handling path in 'create()'
  crypto: stm32 - Try to fix hash padding
  crypto: axis - hide an unused variable

7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/livep...
Linus Torvalds [Thu, 12 Oct 2017 16:21:56 +0000 (09:21 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/livepatching

Pull livepatching fix from Jiri Kosina:

 - bugfix for handling of coming modules (incorrect handling of failure)
   from Joe Lawrence

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/livepatching:
  livepatch: unpatch all klp_objects if klp_module_coming fails

7 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Thu, 12 Oct 2017 16:08:56 +0000 (09:08 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/hid

Pull HID fixes from Jiri Kosina:

 - fix for potential out-of-bounds memory access (found by fuzzing,
   likely requires specially crafted device to trigger) by Jaejoong Kim

 - two new device IDs for elecom driver from Alex Manoussakis

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
  HID: hid-elecom: extend to fix descriptor for HUGE trackball
  HID: usbhid: fix out-of-bounds bug

7 years agoMerge tag 'sound-4.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Thu, 12 Oct 2017 16:01:18 +0000 (09:01 -0700)]
Merge tag 'sound-4.14-rc5' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "It's been a busy week for defending the attacks from fuzzer people.

  This contains various USB-audio driver fixes and sequencer core fixes
  spotted by syzkaller and other fuzzer, as well as one quirk for a
  Plantronics USB audio device"

* tag 'sound-4.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: caiaq: Fix stray URB at probe error path
  ALSA: seq: Fix use-after-free at creating a port
  ALSA: usb-audio: Kill stray URB at exiting
  ALSA: line6: Fix leftover URB at error-path during probe
  ALSA: line6: Fix NULL dereference at podhd_disconnect()
  ALSA: line6: Fix missing initialization before error path
  ALSA: seq: Fix copy_from_user() call inside lock
  ALSA: usb-audio: Add sample rate quirk for Plantronics P610

7 years agoMerge branch 'waitid-fix'
Linus Torvalds [Thu, 12 Oct 2017 15:36:47 +0000 (08:36 -0700)]
Merge branch 'waitid-fix'

Merge waitid() fix from Kees Cook.

I'd have hoped that the unsafe_{get|put}_user() naming would have
avoided these kinds of stupid bugs, but no such luck.

* waitid-fix:
  waitid(): Add missing access_ok() checks

7 years agox86/apic: Update TSC_DEADLINE quirk with additional SKX stepping
Len Brown [Wed, 11 Oct 2017 21:16:04 +0000 (17:16 -0400)]
x86/apic: Update TSC_DEADLINE quirk with additional SKX stepping

SKX stepping-3 fixed the TSC_DEADLINE issue in a different ucode
version number than stepping-4.  Linux needs to know this stepping-3
specific version number to also enable the TSC_DEADLINE on stepping-3.

The steppings and ucode versions are documented in the SKX BIOS update:
https://downloadmirror.intel.com/26978/eng/ReleaseNotes_R00.01.0004.txt

Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: peterz@infradead.org
Link: https://lkml.kernel.org/r/60f2bbf7cf617e212b522e663f84225bfebc50e5.1507756305.git.len.brown@intel.com
7 years agox86/apic: Silence "FW_BUG TSC_DEADLINE disabled due to Errata" on hypervisors
Paolo Bonzini [Tue, 10 Oct 2017 10:12:57 +0000 (12:12 +0200)]
x86/apic: Silence "FW_BUG TSC_DEADLINE disabled due to Errata" on hypervisors

Commit 594a30fb1242 ("x86/apic: Silence "FW_BUG TSC_DEADLINE disabled
due to Errata" on CPUs without the feature", 2017-08-30) was also about
silencing the warning on VirtualBox; however, KVM does expose the TSC
deadline timer, and it's virtualized so that it is immune from CPU errata.

Therefore, booting 4.13 with "-cpu Haswell" shows this in the logs:

     [    0.000000] [Firmware Bug]: TSC_DEADLINE disabled due to Errata;
                    please update microcode to version: 0xb2 (or later)

Even if you had a hypervisor that does _not_ virtualize the TSC deadline
and rather exposes the hardware one, it should be the hypervisors task
to update microcode and possibly hide the flag from CPUID.  So just
hide the message when running on _any_ hypervisor, not just those that
do not support the TSC deadline timer.

The older check still makes sense, so keep it.

Fixes: bd9240a18e ("x86/apic: Add TSC_DEADLINE quirk due to errata")
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Hans de Goede <hdegoede@redhat.com>
Cc: kvm@vger.kernel.org
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/1507630377-54471-1-git-send-email-pbonzini@redhat.com
7 years agodrm/msm: fix return value check in _msm_gem_kernel_new()
Wei Yongjun [Wed, 11 Oct 2017 11:36:56 +0000 (11:36 +0000)]
drm/msm: fix return value check in _msm_gem_kernel_new()

In case of error, the function msm_gem_get_vaddr() returns ERR_PTR()
and never returns NULL. The NULL test in the return value check should
be replaced with IS_ERR().

Fixes: 8223286d62e2 ("drm/msm: Add a helper function for in-kernel
buffer allocations")
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Signed-off-by: Rob Clark <robdclark@gmail.com>
7 years agodrm/msm: use proper memory barriers for updating tail/head
Rob Clark [Mon, 2 Oct 2017 14:28:37 +0000 (10:28 -0400)]
drm/msm: use proper memory barriers for updating tail/head

Fixes intermittent corruption of cmdstream dump.

Signed-off-by: Rob Clark <robdclark@gmail.com>
7 years agodrm/msm/mdp5: add missing max size for 8x74 v1
Rob Clark [Thu, 12 Oct 2017 12:37:45 +0000 (08:37 -0400)]
drm/msm/mdp5: add missing max size for 8x74 v1

This should have same max width as v2.

Signed-off-by: Rob Clark <robdclark@gmail.com>
7 years agodrm/amdgpu: fix placement flags in amdgpu_ttm_bind
Christian König [Fri, 1 Sep 2017 07:22:56 +0000 (09:22 +0200)]
drm/amdgpu: fix placement flags in amdgpu_ttm_bind

Otherwise we lose the NO_EVICT flag and can try to evict pinned BOs.

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
7 years agopowerpc/perf: Add ___GFP_NOWARN flag to alloc_pages_node()
Anju T Sudhakar [Wed, 11 Oct 2017 12:57:39 +0000 (18:27 +0530)]
powerpc/perf: Add ___GFP_NOWARN flag to alloc_pages_node()

Stack trace output during a stress test:
 [    4.310049] Freeing initrd memory: 22592K
[    4.310646] rtas_flash: no firmware flash support
[    4.313341] cpuhp/64: page allocation failure: order:0, mode:0x14480c0(GFP_KERNEL|__GFP_ZERO|__GFP_THISNODE), nodemask=(null)
[    4.313465] cpuhp/64 cpuset=/ mems_allowed=0
[    4.313521] CPU: 64 PID: 392 Comm: cpuhp/64 Not tainted 4.11.0-39.el7a.ppc64le #1
[    4.313588] Call Trace:
[    4.313622] [c000000f1fb1b8e0] [c000000000c09388] dump_stack+0xb0/0xf0 (unreliable)
[    4.313694] [c000000f1fb1b920] [c00000000030ef6c] warn_alloc+0x12c/0x1c0
[    4.313753] [c000000f1fb1b9c0] [c00000000030ff68] __alloc_pages_nodemask+0xea8/0x1000
[    4.313823] [c000000f1fb1bbb0] [c000000000113a8c] core_imc_mem_init+0xbc/0x1c0
[    4.313892] [c000000f1fb1bc00] [c000000000113cdc] ppc_core_imc_cpu_online+0x14c/0x170
[    4.313962] [c000000f1fb1bc90] [c000000000125758] cpuhp_invoke_callback+0x198/0x5d0
[    4.314031] [c000000f1fb1bd00] [c00000000012782c] cpuhp_thread_fun+0x8c/0x3d0
[    4.314101] [c000000f1fb1bd60] [c0000000001678d0] smpboot_thread_fn+0x290/0x2a0
[    4.314169] [c000000f1fb1bdc0] [c00000000015ee78] kthread+0x168/0x1b0
[    4.314229] [c000000f1fb1be30] [c00000000000b368] ret_from_kernel_thread+0x5c/0x74
[    4.314313] Mem-Info:
[    4.314356] active_anon:0 inactive_anon:0 isolated_anon:0

core_imc_mem_init() at system boot use alloc_pages_node() to get memory
and alloc_pages_node() throws this stack dump when tried to allocate
memory from a node which has no memory behind it. Add a ___GFP_NOWARN
flag in allocation request as a fix.

Signed-off-by: Anju T Sudhakar <anju@linux.vnet.ibm.com>
Reported-by: Michael Ellerman <mpe@ellerman.id.au>
Reported-by: Venkat R.B <venkatb3@in.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agopowerpc/perf: Fix for core/nest imc call trace on cpuhotplug
Anju T Sudhakar [Wed, 4 Oct 2017 06:50:52 +0000 (12:20 +0530)]
powerpc/perf: Fix for core/nest imc call trace on cpuhotplug

Nest/core pmu units are enabled only when it is used. A reference count is
maintained for the events which uses the nest/core pmu units. Currently in
*_imc_counters_release function a WARN() is used for notification of any
underflow of ref count.

The case where event ref count hit a negative value is, when perf session is
started, followed by offlining of all cpus in a given core.
i.e. in cpuhotplug offline path ppc_core_imc_cpu_offline() function set the
ref->count to zero, if the current cpu which is about to offline is the last
cpu in a given core and make an OPAL call to disable the engine in that core.
And on perf session termination, perf->destroy (core_imc_counters_release) will
first decrement the ref->count for this core and based on the ref->count value
an opal call is made to disable the core-imc engine.
Now, since cpuhotplug path already clears the ref->count for core and disabled
the engine, perf->destroy() decrementing again at event termination make it
negative which in turn fires the WARN_ON. The same happens for nest units.

Add a check to see if the reference count is alreday zero, before decrementing
the count, so that the ref count will not hit a negative value.

Signed-off-by: Anju T Sudhakar <anju@linux.vnet.ibm.com>
Reviewed-by: Santosh Sivaraj <santosh@fossix.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
7 years agoMAINTAINERS: Add Paul Mackerras as maintainer for KVM/powerpc
Thomas Huth [Mon, 9 Oct 2017 14:34:01 +0000 (16:34 +0200)]
MAINTAINERS: Add Paul Mackerras as maintainer for KVM/powerpc

Paul is handling almost all of the powerpc related KVM patches nowadays,
so he should be mentioned in the MAINTAINERS file accordingly.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
7 years agoKVM: nVMX: fix guest CR4 loading when emulating L2 to L1 exit
Haozhong Zhang [Tue, 10 Oct 2017 07:01:22 +0000 (15:01 +0800)]
KVM: nVMX: fix guest CR4 loading when emulating L2 to L1 exit

When KVM emulates an exit from L2 to L1, it loads L1 CR4 into the
guest CR4. Before this CR4 loading, the guest CR4 refers to L2
CR4. Because these two CR4's are in different levels of guest, we
should vmx_set_cr4() rather than kvm_set_cr4() here. The latter, which
is used to handle guest writes to its CR4, checks the guest change to
CR4 and may fail if the change is invalid.

The failure may cause trouble. Consider we start
  a L1 guest with non-zero L1 PCID in use,
     (i.e. L1 CR4.PCIDE == 1 && L1 CR3.PCID != 0)
and
  a L2 guest with L2 PCID disabled,
     (i.e. L2 CR4.PCIDE == 0)
and following events may happen:

1. If kvm_set_cr4() is used in load_vmcs12_host_state() to load L1 CR4
   into guest CR4 (in VMCS01) for L2 to L1 exit, it will fail because
   of PCID check. As a result, the guest CR4 recorded in L0 KVM (i.e.
   vcpu->arch.cr4) is left to the value of L2 CR4.

2. Later, if L1 attempts to change its CR4, e.g., clearing VMXE bit,
   kvm_set_cr4() in L0 KVM will think L1 also wants to enable PCID,
   because the wrong L2 CR4 is used by L0 KVM as L1 CR4. As L1
   CR3.PCID != 0, L0 KVM will inject GP to L1 guest.

Fixes: 4704d0befb072 ("KVM: nVMX: Exiting from L2 to L1")
Cc: qemu-stable@nongnu.org
Signed-off-by: Haozhong Zhang <haozhong.zhang@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>