sfrench/cifs-2.6.git
5 years agoMerge tag 'trace-v4.20-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Fri, 30 Nov 2018 17:32:34 +0000 (09:32 -0800)]
Merge tag 'trace-v4.20-rc3' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "While rewriting the function graph tracer, I discovered a design flaw
  that was introduced by a patch that tried to fix one bug, but by doing
  so created another bug.

  As both bugs corrupt the output (but they do not crash the kernel), I
  decided to fix the design such that it could have both bugs fixed. The
  original fix, fixed time reporting of the function graph tracer when
  doing a max_depth of one. This was code that can test how much the
  kernel interferes with userspace. But in doing so, it could corrupt
  the time keeping of the function profiler.

  The issue is that the curr_ret_stack variable was being used for two
  different meanings. One was to keep track of the stack pointer on the
  ret_stack (shadow stack used by the function graph tracer), and the
  other use case was the graph call depth. Although, the two may be
  closely related, where they got updated was the issue that lead to the
  two different bugs that required the two use cases to be updated
  differently.

  The big issue with this fix is that it requires changing each
  architecture. The good news is, I was able to remove a lot of code
  that was duplicated within the architectures and place it into a
  single location. Then I could make the fix in one place.

  I pushed this code into linux-next to let it settle over a week, and
  before doing so, I cross compiled all the affected architectures to
  make sure that they built fine.

  In the mean time, I also pulled in a patch that fixes the sched_switch
  previous tasks state output, that was not actually correct"

* tag 'trace-v4.20-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  sched, trace: Fix prev_state output in sched_switch tracepoint
  function_graph: Have profiler use curr_ret_stack and not depth
  function_graph: Reverse the order of pushing the ret_stack and the callback
  function_graph: Move return callback before update of curr_ret_stack
  function_graph: Use new curr_ret_depth to manage depth instead of curr_ret_stack
  function_graph: Make ftrace_push_return_trace() static
  sparc/function_graph: Simplify with function_graph_enter()
  sh/function_graph: Simplify with function_graph_enter()
  s390/function_graph: Simplify with function_graph_enter()
  riscv/function_graph: Simplify with function_graph_enter()
  powerpc/function_graph: Simplify with function_graph_enter()
  parisc: function_graph: Simplify with function_graph_enter()
  nds32: function_graph: Simplify with function_graph_enter()
  MIPS: function_graph: Simplify with function_graph_enter()
  microblaze: function_graph: Simplify with function_graph_enter()
  arm64: function_graph: Simplify with function_graph_enter()
  ARM: function_graph: Simplify with function_graph_enter()
  x86/function_graph: Simplify with function_graph_enter()
  function_graph: Create function_graph_enter() to consolidate architecture code

5 years agoACPI/IORT: Fix iort_get_platform_device_domain() uninitialized pointer value
Lorenzo Pieralisi [Thu, 29 Nov 2018 09:55:59 +0000 (09:55 +0000)]
ACPI/IORT: Fix iort_get_platform_device_domain() uninitialized pointer value

Running the Clang static analyzer on IORT code detected the following
error:

Logic error: Branch condition evaluates to a garbage value

in

iort_get_platform_device_domain()

If the named component associated with a given device has no IORT
mappings, iort_get_platform_device_domain() exits its MSI mapping loop
with msi_parent pointer containing garbage, which can lead to erroneous
code path execution.

Initialize the msi_parent pointer, fixing the bug.

Fixes: d4f54a186667 ("ACPI: platform: setup MSI domain for ACPI based
platform device")
Reported-by: Patrick Bellasi <patrick.bellasi@arm.com>
Reviewed-by: Hanjun Guo <hanjun.guo@linaro.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Cc: Sudeep Holla <sudeep.holla@arm.com>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
5 years agoMerge tag 'drm-fixes-2018-11-30' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 30 Nov 2018 17:10:29 +0000 (09:10 -0800)]
Merge tag 'drm-fixes-2018-11-30' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "This weeks instalment of fixes. Looks fairly like business as usual
  and everything seems to rolling along. There was one MST fix applied
  and reverted in the misc tree, but otherwise nothing too strange in
  here.

  core:
   - incorrect master setting on error fix

  i915:
   - only GVT fixes this week:
      * one MOCS register load
      * rpm lock fix
      * use after free

  rcar-du:
   - regression fix for group start

  amdgpu:
   - DP MST fix
   - GPUVM fix for huge pages
   - RLC fix for vega20

  ast:
   - fix EDID reading stability
   - ioreg free fix

  meson:
   - sleep in irq fix
   - vblank fixes
   - array boundary fix"

* tag 'drm-fixes-2018-11-30' of git://anongit.freedesktop.org/drm/drm:
  drm/ast: fixed reading monitor EDID not stable issue
  drm/ast: Fix incorrect free on ioregs
  Revert "drm/dp_mst: Skip validating ports during destruction, just ref"
  drm/amdgpu: Add delay after enable RLC ucode
  drm/amdgpu: Avoid endless loop in GPUVM fragment processing
  drm/amdgpu: Cast to uint64_t before left shift
  drm/meson: add support for 1080p25 mode
  drm/meson: Fix OOB memory accesses in meson_viu_set_osd_lut()
  drm/meson: Enable fast_io in meson_dw_hdmi_regmap_config
  drm/meson: Fixes for drm_crtc_vblank_on/off support
  drm: set is_master to 0 upon drm_new_set_master() failure
  drm/dp_mst: Skip validating ports during destruction, just ref
  drm: rcar-du: Fix DU3 start/stop on M3-N
  drm/amd/dm: Understand why attaching path/tile properties are needed
  drm/amd/dm: Don't forget to attach MST encoders
  drm/i915/gvt: Avoid use-after-free iterating the gtt list
  drm/i915/gvt: ensure gpu is powered before do i915_gem_gtt_insert
  drm/i915/gvt: not to touch undefined MOCS registers

5 years agoMerge branch 'nvme-4.20' of git://git.infradead.org/nvme into for-linus
Jens Axboe [Fri, 30 Nov 2018 17:08:39 +0000 (10:08 -0700)]
Merge branch 'nvme-4.20' of git://git.infradead.org/nvme into for-linus

Pull NVMe fixes from Christoph:

"Various fixlets all over."

* 'nvme-4.20' of git://git.infradead.org/nvme:
  nvme-rdma: fix double freeing of async event data
  nvme: flush namespace scanning work just before removing namespaces
  nvme: warn when finding multi-port subsystems without multipathing enabled
  nvme-pci: fix surprise removal
  nvme-fc: initialize nvme_req(rq)->ctrl after calling __nvme_fc_init_request()
  nvme: Free ctrl device name on init failure

5 years agoblock: fix single range discard merge
Ming Lei [Fri, 30 Nov 2018 16:38:18 +0000 (00:38 +0800)]
block: fix single range discard merge

There are actually two kinds of discard merge:

- one is the normal discard merge, just like normal read/write request,
and call it single-range discard

- another is the multi-range discard, queue_max_discard_segments(rq->q) > 1

For the former case, queue_max_discard_segments(rq->q) is 1, and we
should handle this kind of discard merge like the normal read/write
request.

This patch fixes the following kernel panic issue[1], which is caused by
not removing the single-range discard request from elevator queue.

Guangwu has one raid discard test case, in which this issue is a bit
easier to trigger, and I verified that this patch can fix the kernel
panic issue in Guangwu's test case.

[1] kernel panic log from Jens's report

 BUG: unable to handle kernel NULL pointer dereference at 0000000000000148
 PGD 0 P4D 0.
 Oops: 0000 [#1] SMP PTI
 CPU: 37 PID: 763 Comm: kworker/37:1H Not tainted \
4.20.0-rc3-00649-ge64d9a554a91-dirty #14  Hardware name: Wiwynn \
Leopard-Orv2/Leopard-DDR BW, BIOS LBM08   03/03/2017       Workqueue: kblockd \
blk_mq_run_work_fn                                            RIP: \
0010:blk_mq_get_driver_tag+0x81/0x120                                       Code: 24 \
10 48 89 7c 24 20 74 21 83 fa ff 0f 95 c0 48 8b 4c 24 28 65 48 33 0c 25 28 00 00 00 \
0f 85 96 00 00 00 48 83 c4 30 5b 5d c3 <48> 8b 87 48 01 00 00 8b 40 04 39 43 20 72 37 \
f6 87 b0 00 00 00 02  RSP: 0018:ffffc90004aabd30 EFLAGS: 00010246                     \
  RAX: 0000000000000003 RBX: ffff888465ea1300 RCX: ffffc90004aabde8
 RDX: 00000000ffffffff RSI: ffffc90004aabde8 RDI: 0000000000000000
 RBP: 0000000000000000 R08: ffff888465ea1348 R09: 0000000000000000
 R10: 0000000000001000 R11: 00000000ffffffff R12: ffff888465ea1300
 R13: 0000000000000000 R14: ffff888465ea1348 R15: ffff888465d10000
 FS:  0000000000000000(0000) GS:ffff88846f9c0000(0000) knlGS:0000000000000000
 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
 CR2: 0000000000000148 CR3: 000000000220a003 CR4: 00000000003606e0
 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
 Call Trace:
  blk_mq_dispatch_rq_list+0xec/0x480
  ? elv_rb_del+0x11/0x30
  blk_mq_do_dispatch_sched+0x6e/0xf0
  blk_mq_sched_dispatch_requests+0xfa/0x170
  __blk_mq_run_hw_queue+0x5f/0xe0
  process_one_work+0x154/0x350
  worker_thread+0x46/0x3c0
  kthread+0xf5/0x130
  ? process_one_work+0x350/0x350
  ? kthread_destroy_worker+0x50/0x50
  ret_from_fork+0x1f/0x30
 Modules linked in: sb_edac x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel \
kvm switchtec irqbypass iTCO_wdt iTCO_vendor_support efivars cdc_ether usbnet mii \
cdc_acm i2c_i801 lpc_ich mfd_core ipmi_si ipmi_devintf ipmi_msghandler acpi_cpufreq \
button sch_fq_codel nfsd nfs_acl lockd grace auth_rpcgss oid_registry sunrpc nvme \
nvme_core fuse sg loop efivarfs autofs4  CR2: 0000000000000148                        \

 ---[ end trace 340a1fb996df1b9b ]---
 RIP: 0010:blk_mq_get_driver_tag+0x81/0x120
 Code: 24 10 48 89 7c 24 20 74 21 83 fa ff 0f 95 c0 48 8b 4c 24 28 65 48 33 0c 25 28 \
00 00 00 0f 85 96 00 00 00 48 83 c4 30 5b 5d c3 <48> 8b 87 48 01 00 00 8b 40 04 39 43 \
20 72 37 f6 87 b0 00 00 00 02

Fixes: 445251d0f4d329a ("blk-mq: fix discard merge with scheduler attached")
Reported-by: Jens Axboe <axboe@kernel.dk>
Cc: Guangwu Zhang <guazhang@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Jianchao Wang <jianchao.w.wang@oracle.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agostackleak: Disable function tracing and kprobes for stackleak_erase()
Alexander Popov [Mon, 12 Nov 2018 21:08:48 +0000 (00:08 +0300)]
stackleak: Disable function tracing and kprobes for stackleak_erase()

The stackleak_erase() function is called on the trampoline stack at the
end of syscall. This stack is not big enough for ftrace and kprobes
operations, e.g. it can be exhausted if we use kprobe_events for
stackleak_erase().

So let's disable function tracing and kprobes of stackleak_erase().

Reported-by: kernel test robot <lkp@intel.com>
Fixes: 10e9ae9fabaf ("gcc-plugins: Add STACKLEAK plugin for tracking the kernel stack")
Signed-off-by: Alexander Popov <alex.popov@linux.com>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Reviewed-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
5 years agoMerge tag 'pstore-v4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/kees...
Linus Torvalds [Fri, 30 Nov 2018 17:03:15 +0000 (09:03 -0800)]
Merge tag 'pstore-v4.20-rc5' of git://git./linux/kernel/git/kees/linux

Pull pstore fix from Kees Cook:
 "Fix corrupted compression due to unlucky size choice with ECC"

* tag 'pstore-v4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
  pstore/ram: Correctly calculate usable PRZ bytes

5 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Fri, 30 Nov 2018 16:57:31 +0000 (08:57 -0800)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "This is a bit later than usual for our first -rc but I'm not seeing
  anything worry-some in the RDMA tree right now. Quiet so far this -rc
  cycle, only a few internal driver related bugs and a small series
  fixing ODP bugs found by more advanced testing.

  A set of small driver and core code fixes:

   - Small series fixing longtime user triggerable bugs in the ODP
     processing inside mlx5 and core code

   - Various small driver malfunctions and crashes (use after, free,
     error unwind, implementation bugs)

   - A misfunction of the RDMA GID cache that can be triggered by the
     administrator"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/mlx5: Initialize return variable in case pagefault was skipped
  IB/mlx5: Fix page fault handling for MW
  IB/umem: Set correct address to the invalidation function
  IB/mlx5: Skip non-ODP MR when handling a page fault
  RDMA/hns: Bugfix pbl configuration for rereg mr
  iser: set sector for ambiguous mr status errors
  RDMA/rdmavt: Fix rvt_create_ah function signature
  IB/mlx5: Avoid load failure due to unknown link width
  IB/mlx5: Fix XRC QP support after introducing extended atomic
  RDMA/bnxt_re: Avoid accessing the device structure after it is freed
  RDMA/bnxt_re: Fix system hang when registration with L2 driver fails
  RDMA/core: Add GIDs while changing MAC addr only for registered ndev
  RDMA/mlx5: Fix fence type for IB_WR_LOCAL_INV WR
  net/mlx5: Fix XRC SRQ umem valid bits

5 years agonvme-rdma: fix double freeing of async event data
Prabhath Sajeepa [Wed, 28 Nov 2018 18:11:29 +0000 (11:11 -0700)]
nvme-rdma: fix double freeing of async event data

Some error paths in configuration of admin queue free data buffer
associated with async request SQE without resetting the data buffer
pointer to NULL, This buffer is also freed up again if the controller
is shutdown or reset.

Signed-off-by: Prabhath Sajeepa <psajeepa@purestorage.com>
Reviewed-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme: flush namespace scanning work just before removing namespaces
Sagi Grimberg [Wed, 21 Nov 2018 23:17:37 +0000 (15:17 -0800)]
nvme: flush namespace scanning work just before removing namespaces

nvme_stop_ctrl can be called also for reset flow and there is no need to
flush the scan_work as namespaces are not being removed. This can cause
deadlock in rdma, fc and loop drivers since nvme_stop_ctrl barriers
before controller teardown (and specifically I/O cancellation of the
scan_work itself) takes place, but the scan_work will be blocked anyways
so there is no need to flush it.

Instead, move scan_work flush to nvme_remove_namespaces() where it really
needs to flush.

Reported-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Keith Busch <keith.busch@intel.com>
Reviewed by: James Smart <jsmart2021@gmail.com>
Tested-by: Ewan D. Milne <emilne@redhat.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
5 years agonvme: warn when finding multi-port subsystems without multipathing enabled
Christoph Hellwig [Tue, 20 Nov 2018 15:57:54 +0000 (16:57 +0100)]
nvme: warn when finding multi-port subsystems without multipathing enabled

Without CONFIG_NVME_MULTIPATH enabled a multi-port subsystem might
show up as invididual devices and cause problems, warn about it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
5 years agofscache, cachefiles: remove redundant variable 'cache'
Colin Ian King [Tue, 17 Jul 2018 08:53:42 +0000 (09:53 +0100)]
fscache, cachefiles: remove redundant variable 'cache'

Variable 'cache' is being assigned but is never used hence it is
redundant and can be removed.

Cleans up clang warning:
warning: variable 'cache' set but not used [-Wunused-but-set-variable]

Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agocachefiles: avoid deprecated get_seconds()
Arnd Bergmann [Fri, 13 Jul 2018 14:27:44 +0000 (16:27 +0200)]
cachefiles: avoid deprecated get_seconds()

get_seconds() returns an unsigned long can overflow on some architectures
and is deprecated because of that. In cachefs, we cast that number to
a a 32-bit integer, which will overflow in year 2106 on all architectures.

As confirmed by David Howells, the overflow probably isn't harmful
in the end, since the timestamps are only used to make the file names
unique, but they don't strictly have to be in monotonically increasing
order since the files only exist in order to be deleted as quickly
as possible.

Moving to ktime_get_real_seconds() avoids the deprecated interface.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agocachefiles: Explicitly cast enumerated type in put_object
Nathan Chancellor [Mon, 24 Sep 2018 17:33:44 +0000 (10:33 -0700)]
cachefiles: Explicitly cast enumerated type in put_object

Clang warns when one enumerated type is implicitly converted to another.

fs/cachefiles/namei.c:247:50: warning: implicit conversion from
enumeration type 'enum cachefiles_obj_ref_trace' to different
enumeration type 'enum fscache_obj_ref_trace' [-Wenum-conversion]
        cache->cache.ops->put_object(&xobject->fscache,
cachefiles_obj_put_wait_retry);

Silence this warning by explicitly casting to fscache_obj_ref_trace,
which is also done in put_object.

Reported-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agofscache: fix race between enablement and dropping of object
NeilBrown [Fri, 26 Oct 2018 06:16:29 +0000 (17:16 +1100)]
fscache: fix race between enablement and dropping of object

It was observed that a process blocked indefintely in
__fscache_read_or_alloc_page(), waiting for FSCACHE_COOKIE_LOOKING_UP
to be cleared via fscache_wait_for_deferred_lookup().

At this time, ->backing_objects was empty, which would normaly prevent
__fscache_read_or_alloc_page() from getting to the point of waiting.
This implies that ->backing_objects was cleared *after*
__fscache_read_or_alloc_page was was entered.

When an object is "killed" and then "dropped",
FSCACHE_COOKIE_LOOKING_UP is cleared in fscache_lookup_failure(), then
KILL_OBJECT and DROP_OBJECT are "called" and only in DROP_OBJECT is
->backing_objects cleared.  This leaves a window where
something else can set FSCACHE_COOKIE_LOOKING_UP and
__fscache_read_or_alloc_page() can start waiting, before
->backing_objects is cleared

There is some uncertainty in this analysis, but it seems to be fit the
observations.  Adding the wake in this patch will be handled correctly
by __fscache_read_or_alloc_page(), as it checks if ->backing_objects
is empty again, after waiting.

Customer which reported the hang, also report that the hang cannot be
reproduced with this fix.

The backtrace for the blocked process looked like:

PID: 29360  TASK: ffff881ff2ac0f80  CPU: 3   COMMAND: "zsh"
 #0 [ffff881ff43efbf8] schedule at ffffffff815e56f1
 #1 [ffff881ff43efc58] bit_wait at ffffffff815e64ed
 #2 [ffff881ff43efc68] __wait_on_bit at ffffffff815e61b8
 #3 [ffff881ff43efca0] out_of_line_wait_on_bit at ffffffff815e625e
 #4 [ffff881ff43efd08] fscache_wait_for_deferred_lookup at ffffffffa04f2e8f [fscache]
 #5 [ffff881ff43efd18] __fscache_read_or_alloc_page at ffffffffa04f2ffe [fscache]
 #6 [ffff881ff43efd58] __nfs_readpage_from_fscache at ffffffffa0679668 [nfs]
 #7 [ffff881ff43efd78] nfs_readpage at ffffffffa067092b [nfs]
 #8 [ffff881ff43efda0] generic_file_read_iter at ffffffff81187a73
 #9 [ffff881ff43efe50] nfs_file_read at ffffffffa066544b [nfs]
#10 [ffff881ff43efe70] __vfs_read at ffffffff811fc756
#11 [ffff881ff43efee8] vfs_read at ffffffff811fccfa
#12 [ffff881ff43eff18] sys_read at ffffffff811fda62
#13 [ffff881ff43eff50] entry_SYSCALL_64_fastpath at ffffffff815e986e

Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agofs: fix lost error code in dio_complete
Maximilian Heyne [Fri, 30 Nov 2018 15:35:14 +0000 (08:35 -0700)]
fs: fix lost error code in dio_complete

commit e259221763a40403d5bb232209998e8c45804ab8 ("fs: simplify the
generic_write_sync prototype") reworked callers of generic_write_sync(),
and ended up dropping the error return for the directio path. Prior to
that commit, in dio_complete(), an error would be bubbled up the stack,
but after that commit, errors passed on to dio_complete were eaten up.

This was reported on the list earlier, and a fix was proposed in
https://lore.kernel.org/lkml/20160921141539.GA17898@infradead.org/, but
never followed up with.  We recently hit this bug in our testing where
fencing io errors, which were previously erroring out with EIO, were
being returned as success operations after this commit.

The fix proposed on the list earlier was a little short -- it would have
still called generic_write_sync() in case `ret` already contained an
error. This fix ensures generic_write_sync() is only called when there's
no pending error in the write. Additionally, transferred is replaced
with ret to bring this code in line with other callers.

Fixes: e259221763a4 ("fs: simplify the generic_write_sync prototype")
Reported-by: Ravi Nankani <rnankani@amazon.com>
Signed-off-by: Maximilian Heyne <mheyne@amazon.de>
Reviewed-by: Christoph Hellwig <hch@lst.de>
CC: Torsten Mehlan <tomeh@amazon.de>
CC: Uwe Dannowski <uwed@amazon.de>
CC: Amit Shah <aams@amazon.de>
CC: David Woodhouse <dwmw@amazon.co.uk>
CC: stable@vger.kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: avoid extra bio reference for async O_DIRECT
Christoph Hellwig [Fri, 30 Nov 2018 08:23:48 +0000 (09:23 +0100)]
block: avoid extra bio reference for async O_DIRECT

The bio referencing has a trick that doesn't do any actual atomic
inc/dec on the reference count until we have to elevator to > 1. For the
async IO O_DIRECT case, we can't use the simple DIO variants, so we use
__blkdev_direct_IO(). It always grabs an extra reference to the bio
after allocation, which means we then enter the slower path of actually
having to do atomic_inc/dec on the count.

We don't need to do that for the async case, unless we end up going
multi-bio, in which case we're already doing huge amounts of IO. For the
smaller IO case (< BIO_MAX_PAGES), we can do without the extra ref.

Based on an earlier patch (and commit log) from Jens Axboe.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agotracing/fgraph: Fix set_graph_function from showing interrupts
Steven Rostedt (VMware) [Thu, 29 Nov 2018 13:50:27 +0000 (08:50 -0500)]
tracing/fgraph: Fix set_graph_function from showing interrupts

The tracefs file set_graph_function is used to only function graph functions
that are listed in that file (or all functions if the file is empty). The
way this is implemented is that the function graph tracer looks at every
function, and if the current depth is zero and the function matches
something in the file then it will trace that function. When other functions
are called, the depth will be greater than zero (because the original
function will be at depth zero), and all functions will be traced where the
depth is greater than zero.

The issue is that when a function is first entered, and the handler that
checks this logic is called, the depth is set to zero. If an interrupt comes
in and a function in the interrupt handler is traced, its depth will be
greater than zero and it will automatically be traced, even if the original
function was not. But because the logic only looks at depth it may trace
interrupts when it should not be.

The recent design change of the function graph tracer to fix other bugs
caused the depth to be zero while the function graph callback handler is
being called for a longer time, widening the race of this happening. This
bug was actually there for a longer time, but because the race window was so
small it seldom happened. The Fixes tag below is for the commit that widen
the race window, because that commit belongs to a series that will also help
fix the original bug.

Cc: stable@kernel.org
Fixes: 39eb456dacb5 ("function_graph: Use new curr_ret_depth to manage depth instead of curr_ret_stack")
Reported-by: Joe Lawrence <joe.lawrence@redhat.com>
Tested-by: Joe Lawrence <joe.lawrence@redhat.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 years agotracepoint: Use __idx instead of idx in DO_TRACE macro to make it unique
Zenghui Yu [Wed, 28 Nov 2018 03:35:23 +0000 (03:35 +0000)]
tracepoint: Use __idx instead of idx in DO_TRACE macro to make it unique

After enabling KVM event tracing, almost all of trace_kvm_exit()'s
printk shows

"kvm_exit: IRQ: ..."

even if the actual exception_type is NOT IRQ.  More specifically,
trace_kvm_exit() is defined in virt/kvm/arm/trace.h by TRACE_EVENT.

This slight problem may have existed after commit e6753f23d961
("tracepoint: Make rcuidle tracepoint callers use SRCU"). There are
two variables in trace_kvm_exit() and __DO_TRACE() which have the
same name, *idx*. Thus the actual value of *idx* will be overwritten
when tracing. Fix it by adding a simple prefix.

Cc: Joel Fernandes <joel@joelfernandes.org>
Cc: Wang Haibin <wanghaibin.wang@huawei.com>
Cc: linux-trace-devel@vger.kernel.org
Cc: stable@vger.kernel.org
Fixes: e6753f23d961 ("tracepoint: Make rcuidle tracepoint callers use SRCU")
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 years agoafs: Use d_instantiate() rather than d_add() and don't d_drop()
David Howells [Tue, 13 Nov 2018 23:20:35 +0000 (23:20 +0000)]
afs: Use d_instantiate() rather than d_add() and don't d_drop()

Use d_instantiate() rather than d_add() and don't d_drop() in
afs_vnode_new_inode().  The dentry shouldn't be removed as it's not
changing its name.

Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
5 years agoafs: Fix missing net error handling
David Howells [Tue, 13 Nov 2018 23:20:28 +0000 (23:20 +0000)]
afs: Fix missing net error handling

kAFS can be given certain network errors (EADDRNOTAVAIL, EHOSTDOWN and
ERFKILL) that it doesn't handle in its server/address rotation algorithms.
They cause the probing and rotation to abort immediately rather than
rotating.

Fix this by:

 (1) Abstracting out the error prioritisation from the VL and FS rotation
     algorithms into a common function and expand usage into the server
     probing code.

     When multiple errors are available, this code selects the one we'd
     prefer to return.

 (2) Add handling for EADDRNOTAVAIL, EHOSTDOWN and ERFKILL.

Fixes: 0fafdc9f888b ("afs: Fix file locking")
Fixes: 0338747d8454 ("afs: Probe multiple fileservers simultaneously")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
5 years agoafs: Fix validation/callback interaction
David Howells [Tue, 13 Nov 2018 23:20:21 +0000 (23:20 +0000)]
afs: Fix validation/callback interaction

When afs_validate() is called to validate a vnode (inode), there are two
unhandled cases in the fastpath at the top of the function:

 (1) If the vnode is promised (AFS_VNODE_CB_PROMISED is set), the break
     counters match and the data has expired, then there's an implicit case
     in which the vnode needs revalidating.

     This has no consequences since the default "valid = false" set at the
     top of the function happens to do the right thing.

 (2) If the vnode is not promised and it hasn't been deleted
     (AFS_VNODE_DELETED is not set) then there's a default case we're not
     handling in which the vnode is invalid.  If the vnode is invalid, we
     need to bring cb_s_break and cb_v_break up to date before we refetch
     the status.

     As a consequence, once the server loses track of the client
     (ie. sufficient time has passed since we last sent it an operation),
     it will send us a CB.InitCallBackState* operation when we next try to
     talk to it.  This calls afs_init_callback_state() which increments
     afs_server::cb_s_break, but this then doesn't propagate to the
     afs_vnode record.

     The result being that every afs_validate() call thereafter sends a
     status fetch operation to the server.

Clarify and fix this by:

 (A) Setting valid in all the branches rather than initialising it at the
     top so that the compiler catches where we've missed.

 (B) Restructuring the logic in the 'promised' branch so that we set valid
     to false if the callback is due to expire (or has expired) and so that
     the final case is that the vnode is still valid.

 (C) Adding an else-statement that ups cb_s_break and cb_v_break if the
     promised and deleted cases don't match.

Fixes: c435ee34551e ("afs: Overhaul the callback handling")
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
5 years agoMerge tag 'acpi-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Thu, 29 Nov 2018 23:54:12 +0000 (15:54 -0800)]
Merge tag 'acpi-4.20-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI fix from Rafael Wysocki:
 "Fix a recent regression in ACPICA releted to the Generic Serial Bus
  protocol handling and causing it to read or write too little or too
  much data in some cases, so incorrect data may be written to hardware
  as a result (Hans de Goede)"

* tag 'acpi-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPICA: Fix handling of buffer-size in acpi_ex_write_data_to_field()

5 years agoMerge tag 'pm-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Thu, 29 Nov 2018 23:07:30 +0000 (15:07 -0800)]
Merge tag 'pm-4.20-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael Wysocki:
 "These fix two issues in the operating performance points (OPP)
  framework.

  Specifics:

   - Fix the handling of the "operating-points-v2" property to avoid
     failures if multiple phandles are present in it which is legitimate
     (Viresh Kumar).

   - Drop the unnecessary static initialization of the .owner field in
     the ti_opp_supply_driver structure (YueHaibing)"

* tag 'pm-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  OPP: Fix parsing of multiple phandles in "operating-points-v2" property
  opp: ti-opp-supply: Fix platform_no_drv_owner.cocci warnings

5 years agoRDMA/mlx5: Initialize return variable in case pagefault was skipped
Leon Romanovsky [Thu, 29 Nov 2018 10:25:29 +0000 (12:25 +0200)]
RDMA/mlx5: Initialize return variable in case pagefault was skipped

Pagefaults occurred in non-ODP MR are completely valid events, so
initialize return variable to 0.

Fixes: 4d5422a309de ("IB/mlx5: Skip non-ODP MR when handling a page fault")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
5 years agopstore/ram: Correctly calculate usable PRZ bytes
Kees Cook [Thu, 1 Nov 2018 23:17:22 +0000 (16:17 -0700)]
pstore/ram: Correctly calculate usable PRZ bytes

The actual number of bytes stored in a PRZ is smaller than the
bytes requested by platform data, since there is a header on each
PRZ. Additionally, if ECC is enabled, there are trailing bytes used
as well. Normally this mismatch doesn't matter since PRZs are circular
buffers and the leading "overflow" bytes are just thrown away. However, in
the case of a compressed record, this rather badly corrupts the results.

This corruption was visible with "ramoops.mem_size=204800 ramoops.ecc=1".
Any stored crashes would not be uncompressable (producing a pstorefs
"dmesg-*.enc.z" file), and triggering errors at boot:

  [    2.790759] pstore: crypto_comp_decompress failed, ret = -22!

Backporting this depends on commit 70ad35db3321 ("pstore: Convert console
write to use ->write_buf")

Reported-by: Joel Fernandes <joel@joelfernandes.org>
Fixes: b0aad7a99c1d ("pstore: Add compression support to pstore")
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
5 years agosbitmap: don't loop for find_next_zero_bit() for !round_robin
Jens Axboe [Thu, 29 Nov 2018 19:35:16 +0000 (12:35 -0700)]
sbitmap: don't loop for find_next_zero_bit() for !round_robin

If we aren't forced to do round robin tag allocation, just use the
allocation hint to find the index for the tag word, don't use it for the
offset inside the word. This avoids a potential extra round trip in the
bit looping, and since we're fetching this cacheline, we may as well
check the whole word from the start.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoMerge branch 'acpica-fixes'
Rafael J. Wysocki [Thu, 29 Nov 2018 20:21:39 +0000 (21:21 +0100)]
Merge branch 'acpica-fixes'

* acpica-fixes:
  ACPICA: Fix handling of buffer-size in acpi_ex_write_data_to_field()

5 years agoMerge tag 'selinux-pr-20181129' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 29 Nov 2018 18:15:06 +0000 (10:15 -0800)]
Merge tag 'selinux-pr-20181129' of git://git./linux/kernel/git/pcmoore/selinux

Pull SELinux fix from Paul Moore:
 "One more SELinux fix for v4.20: add some missing netlink message to
  SELinux permission mappings. The netlink messages were added in v4.19,
  but unfortunately we didn't catch it then because the mechanism to
  catch these things was bypassed.

  In addition to adding the mappings, we're adding some comments to the
  code to hopefully prevent bypasses in the future"

* tag 'selinux-pr-20181129' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/selinux:
  selinux: add support for RTM_NEWCHAIN, RTM_DELCHAIN, and RTM_GETCHAIN

5 years agoMerge tag 's390-4.20-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Thu, 29 Nov 2018 18:10:42 +0000 (10:10 -0800)]
Merge tag 's390-4.20-3' of git://git./linux/kernel/git/s390/linux

Pull s390 updates from Martin Schwidefsky:

 - Add two missing kfree calls on error paths in the vfio-ccw code

 - Make sure that all data structures of a mediated vfio-ccw device are
   initialized before registering it

 - Fix a sparse warning in vfio-ccw

 - A followup patch for the pgtable_bytes accounting, the page table
   downgrade for compat processes missed a mm_dec_nr_pmds()

 - Reject sampling requests in the PMU init function of the CPU
   measurement counter facility

 - With the vfio AP driver an AP queue needs to be reset on every device
   probe as the alternative driver could have modified the device state

* tag 's390-4.20-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/mm: correct pgtable_bytes on page table downgrade
  s390/zcrypt: reinit ap queue state machine during device probe
  s390/cpum_cf: Reject request for sampling in event initialization
  s390/cio: Fix cleanup when unsupported IDA format is used
  s390/cio: Fix cleanup of pfn_array alloc failure
  vfio: ccw: Register mediated device once all structures are initialized
  s390/cio: make vfio_ccw_io_region static

5 years agoMerge tag 'sound-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Thu, 29 Nov 2018 18:03:42 +0000 (10:03 -0800)]
Merge tag 'sound-4.20-rc5' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "As a usual pattern, we've got relatively large updates at rc5:

   - A fix for races in ALSA control user elements

   - ASoC DAPM regression due to component refactoring

   - A fix in error handling of ASoC iteration macro

   - ASoC Intel SST Skylake kconfig fix; a new Kconfig will appear as a
     consequence, but in the end it's a good cleanup

   - HD-audio and USB-audio quirks as always

   - Assort of ASoC driver fixes (pcm186x, Intel cht, rockchip, pcm3060,
     rsnd, omap, wm_adsp, qcom, sunxi, stm32)"

* tag 'sound-4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound: (34 commits)
  ALSA: usb-audio: Add vendor and product name for Dell WD19 Dock
  ALSA: hda/realtek - Support ALC300
  ALSA: hda/realtek - Add auto-mute quirk for HP Spectre x360 laptop
  ALSA: hda/realtek - fix the pop noise on headphone for lenovo laptops
  ALSA: control: Fix race between adding and removing a user element
  ALSA: sparc: Fix invalid snd_free_pages() at error path
  ALSA: wss: Fix invalid snd_free_pages() at error path
  ALSA: hda/realtek - fix headset mic detection for MSI MS-B171
  ALSA: hda: Add ASRock N68C-S UCC the power_save blacklist
  ALSA: ac97: Fix incorrect bit shift at AC97-SPSA control write
  ASoC: omap-dmic: Add pm_qos handling to avoid overruns with CPU_IDLE
  ASoC: omap-mcpdm: Add pm_qos handling to avoid under/overruns with CPU_IDLE
  ASoC: omap-mcbsp: Fix latency value calculation for pm_qos
  ASoC: acpi: fix: continue searching when machine is ignored
  ASoC: Intel: Skylake: fix Kconfigs, make HDaudio codec optional
  MAINTAINERS: add ASoC maintainers for sound dt-bindings
  ASoC: pcm186x: Fix device reset-registers trigger value
  ASoC: dapm: Recalculate audio map forcely when card instantiated
  ASoC: omap-abe-twl6040: Fix missing audio card caused by deferred probing
  ASoC: pcm3060: Rename output widgets
  ...

5 years agoMerge tag 'fixes_for_v4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 29 Nov 2018 17:56:00 +0000 (09:56 -0800)]
Merge tag 'fixes_for_v4.20-rc5' of git://git./linux/kernel/git/jack/linux-fs

Pull ext2 and udf fixes from Jan Kara:
 "Three small ext2 and udf fixes"

* tag 'fixes_for_v4.20-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  ext2: fix potential use after free
  ext2: initialize opts.s_mount_opt as zero before using it
  udf: Allow mounting volumes with incorrect identification strings

5 years agoblk-mq: use plug for devices that implement ->commits_rqs()
Jens Axboe [Thu, 29 Nov 2018 17:03:42 +0000 (10:03 -0700)]
blk-mq: use plug for devices that implement ->commits_rqs()

If we have that hook, we know the driver handles bd->last == true in
a smart fashion. If it does, even for multiple hardware queues, it's
a good idea to flush batches of requests to the device, if we have
batches of requests from the submitter.

Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: use bd->last == true for list inserts
Jens Axboe [Sat, 24 Nov 2018 17:15:46 +0000 (10:15 -0700)]
blk-mq: use bd->last == true for list inserts

If we are issuing a list of requests, we know if we're at the last one.
If we fail issuing, ensure that we call ->commits_rqs() to flush any
potential previous requests.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoataflop: implement mq_ops->commit_rqs() hook
Jens Axboe [Wed, 28 Nov 2018 13:09:07 +0000 (06:09 -0700)]
ataflop: implement mq_ops->commit_rqs() hook

We need this for blk-mq to kick things into gear, if we told it that
we had more IO coming, but then failed to deliver on that promise.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agovirtio_blk: implement mq_ops->commit_rqs() hook
Jens Axboe [Mon, 26 Nov 2018 18:00:12 +0000 (11:00 -0700)]
virtio_blk: implement mq_ops->commit_rqs() hook

We need this for blk-mq to kick things into gear, if we told it that
we had more IO coming, but then failed to deliver on that promise.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonvme: implement mq_ops->commit_rqs() hook
Jens Axboe [Thu, 29 Nov 2018 17:02:29 +0000 (10:02 -0700)]
nvme: implement mq_ops->commit_rqs() hook

Split the command submission and the SQ doorbell ring, and add the
doorbell ring as our ->commit_rqs() hook. This allows a list of
requests to be issued, with nvme only writing the SQ update when
it's necessary. This is more efficient if we have lists of requests
to issue, particularly on virtualized hardware, where writing the
SQ doorbell is more expensive than on real hardware. For those cases,
performance increases of 2-3x have been observed.

The use case for this is plugged IO, where blk-mq flushes a batch of
requests at the time.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblk-mq: add mq_ops->commit_rqs()
Jens Axboe [Wed, 28 Nov 2018 00:02:25 +0000 (17:02 -0700)]
blk-mq: add mq_ops->commit_rqs()

blk-mq passes information to the hardware about any given request being
the last that we will issue in this sequence. The point is that hardware
can defer costly doorbell type writes to the last request. But if we run
into errors issuing a sequence of requests, we may never send the request
with bd->last == true set. For that case, we need a hook that tells the
hardware that nothing else is coming right now.

For failures returned by the drivers ->queue_rq() hook, the driver is
responsible for flushing pending requests, if it uses bd->last to
optimize that part. This works like before, no changes there.

Reviewed-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: improve logic around when to sort a plug list
Jens Axboe [Wed, 28 Nov 2018 00:13:56 +0000 (17:13 -0700)]
block: improve logic around when to sort a plug list

Only do it if we have requests for multiple queues in the same
plug.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agopvcalls-front: fixes incorrect error handling
Pan Bian [Thu, 22 Nov 2018 02:07:12 +0000 (10:07 +0800)]
pvcalls-front: fixes incorrect error handling

kfree() is incorrectly used to release the pages allocated by
__get_free_page() and __get_free_pages(). Use the matching deallocators
i.e., free_page() and free_pages(), respectively.

Signed-off-by: Pan Bian <bianpan2016@163.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Juergen Gross <jgross@suse.com>
5 years agoRevert "xen/balloon: Mark unallocated host memory as UNUSABLE"
Igor Druzhinin [Tue, 27 Nov 2018 20:58:21 +0000 (20:58 +0000)]
Revert "xen/balloon: Mark unallocated host memory as UNUSABLE"

This reverts commit b3cf8528bb21febb650a7ecbf080d0647be40b9f.

That commit unintentionally broke Xen balloon memory hotplug with
"hotplug_unpopulated" set to 1. As long as "System RAM" resource
got assigned under a new "Unusable memory" resource in IO/Mem tree
any attempt to online this memory would fail due to general kernel
restrictions on having "System RAM" resources as 1st level only.

The original issue that commit has tried to workaround fa564ad96366
("x86/PCI: Enable a 64bit BAR on AMD Family 15h (Models 00-1f, 30-3f,
60-7f)") also got amended by the following 03a551734 ("x86/PCI: Move
and shrink AMD 64-bit window to avoid conflict") which made the
original fix to Xen ballooning unnecessary.

Signed-off-by: Igor Druzhinin <igor.druzhinin@citrix.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
5 years agoxen: xlate_mmu: add missing header to fix 'W=1' warning
Srikanth Boddepalli [Tue, 27 Nov 2018 14:23:27 +0000 (19:53 +0530)]
xen: xlate_mmu: add missing header to fix 'W=1' warning

Add a missing header otherwise compiler warns about missed prototype:

drivers/xen/xlate_mmu.c:183:5: warning: no previous prototype for 'xen_xlate_unmap_gfn_range?' [-Wmissing-prototypes]
  int xen_xlate_unmap_gfn_range(struct vm_area_struct *vma,
      ^~~~~~~~~~~~~~~~~~~~~~~~~

Signed-off-by: Srikanth Boddepalli <boddepalli.srikanth@gmail.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Joey Pabalinas <joeypabalinas@gmail.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
5 years agoxen/x86: add diagnostic printout to xen_mc_flush() in case of error
Juergen Gross [Fri, 23 Nov 2018 16:24:51 +0000 (17:24 +0100)]
xen/x86: add diagnostic printout to xen_mc_flush() in case of error

Failure of an element of a Xen multicall is signalled via a WARN()
only if the kernel is compiled with MC_DEBUG. It is impossible to
know which element failed and why it did so.

Change that by printing the related information even without MC_DEBUG,
even if maybe in some limited form (e.g. without information which
caller produced the failing element).

Move the printing out of the switch statement in order to have the
same information for a single call.

Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
5 years agoarm64: ftrace: Fix to enable syscall events on arm64
Masami Hiramatsu [Thu, 29 Nov 2018 05:39:33 +0000 (14:39 +0900)]
arm64: ftrace: Fix to enable syscall events on arm64

Since commit 4378a7d4be30 ("arm64: implement syscall wrappers")
introduced "__arm64_" prefix to all syscall wrapper symbols in
sys_call_table, syscall tracer can not find corresponding
metadata from syscall name. In the result, we have no syscall
ftrace events on arm64 kernel, and some bpf testcases are failed
on arm64.

To fix this issue, this introduces custom
arch_syscall_match_sym_name() which skips first 8 bytes when
comparing the syscall and symbol names.

Fixes: 4378a7d4be30 ("arm64: implement syscall wrappers")
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Tested-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Cc: stable@vger.kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
5 years agoarm64: Add workaround for Cortex-A76 erratum 1286807
Catalin Marinas [Mon, 19 Nov 2018 11:27:28 +0000 (11:27 +0000)]
arm64: Add workaround for Cortex-A76 erratum 1286807

On the affected Cortex-A76 cores (r0p0 to r3p0), if a virtual address
for a cacheable mapping of a location is being accessed by a core while
another core is remapping the virtual address to a new physical page
using the recommended break-before-make sequence, then under very rare
circumstances TLBI+DSB completes before a read using the translation
being invalidated has been observed by other observers. The workaround
repeats the TLBI+DSB operation and is shared with the Qualcomm Falkor
erratum 1009

Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
5 years agoselinux: add support for RTM_NEWCHAIN, RTM_DELCHAIN, and RTM_GETCHAIN
Paul Moore [Wed, 28 Nov 2018 17:57:33 +0000 (12:57 -0500)]
selinux: add support for RTM_NEWCHAIN, RTM_DELCHAIN, and RTM_GETCHAIN

Commit 32a4f5ecd738 ("net: sched: introduce chain object to uapi")
added new RTM_* definitions without properly updating SELinux, this
patch adds the necessary SELinux support.

While there was a BUILD_BUG_ON() in the SELinux code to protect from
exactly this case, it was bypassed in the broken commit.  In order to
hopefully prevent this from happening in the future, add additional
comments which provide some instructions on how to resolve the
BUILD_BUG_ON() failures.

Fixes: 32a4f5ecd738 ("net: sched: introduce chain object to uapi")
Cc: <stable@vger.kernel.org> # 4.19
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Paul Moore <paul@paul-moore.com>
5 years agoblk-mq: Add a NULL check in blk_mq_free_map_and_requests()
Dan Carpenter [Thu, 29 Nov 2018 10:56:54 +0000 (13:56 +0300)]
blk-mq: Add a NULL check in blk_mq_free_map_and_requests()

I recently found some code which called blk_mq_free_map_and_requests()
with a NULL set->tags pointer.  I fixed the caller, but it seems like a
good idea to add a NULL check here as well.  Now we can call:

blk_mq_free_tag_set(set);
blk_mq_free_tag_set(set);

twice in a row and it's harmless.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoataflop: fix error handling in atari_floppy_init()
Dan Carpenter [Thu, 29 Nov 2018 10:55:19 +0000 (13:55 +0300)]
ataflop: fix error handling in atari_floppy_init()

Smatch complains that there is an off by one if the allocation fails in:

DMABuffer = atari_stram_alloc(BUFFER_SIZE+512, "ataflop");

In that situation, "i" would be point to one element beyond the end of
the unit[] array.

There is a second bug because the error handling calls
blk_mq_free_tag_set(&unit[i].tag_set); regardless of whether
"disk->queue" is NULL or non-NULL.  So if blk_mq_init_sq_queue() fails,
then that means unit[i].tag_set->tags is NULL and it leads to an Oops.

It's easiest to call put_disk() before the goto to clean up the partial
iteration.  Then the earlier unit[] elements are fully allocated so we
can remove the checks whether "disk->queue" is NULL and the code is
simpler.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agodmaengine: at_hdmac: fix module unloading
Richard Genoud [Tue, 27 Nov 2018 16:06:35 +0000 (17:06 +0100)]
dmaengine: at_hdmac: fix module unloading

of_dma_controller_free() was not called on module onloading.
This lead to a soft lockup:
watchdog: BUG: soft lockup - CPU#0 stuck for 23s!
Modules linked in: at_hdmac [last unloaded: at_hdmac]
when of_dma_request_slave_channel() tried to call ofdma->of_dma_xlate().

Cc: stable@vger.kernel.org
Fixes: bbe89c8e3d59 ("at_hdmac: move to generic DMA binding")
Acked-by: Ludovic Desroches <ludovic.desroches@microchip.com>
Signed-off-by: Richard Genoud <richard.genoud@gmail.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
5 years agodmaengine: at_hdmac: fix memory leak in at_dma_xlate()
Richard Genoud [Tue, 27 Nov 2018 16:06:34 +0000 (17:06 +0100)]
dmaengine: at_hdmac: fix memory leak in at_dma_xlate()

The leak was found when opening/closing a serial port a great number of
time, increasing kmalloc-32 in slabinfo.

Each time the port was opened, dma_request_slave_channel() was called.
Then, in at_dma_xlate(), atslave was allocated with devm_kzalloc() and
never freed. (Well, it was free at module unload, but that's not what we
want).
So, here, kzalloc is more suited for the job since it has to be freed in
atc_free_chan_resources().

Cc: stable@vger.kernel.org
Fixes: bbe89c8e3d59 ("at_hdmac: move to generic DMA binding")
Reported-by: Mario Forner <m.forner@be4energy.com>
Suggested-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Acked-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Acked-by: Ludovic Desroches <ludovic.desroches@microchip.com>
Signed-off-by: Richard Genoud <richard.genoud@gmail.com>
Signed-off-by: Vinod Koul <vkoul@kernel.org>
5 years agoMerge tag 'fixes-for-v4.20-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git...
Greg Kroah-Hartman [Thu, 29 Nov 2018 10:39:35 +0000 (11:39 +0100)]
Merge tag 'fixes-for-v4.20-rc4' of git://git./linux/kernel/git/balbi/usb into usb-linus

Felipe writes:

USB: fixes for v4.20-rc4

In this second set of fixes for the current -rc cycle, we have some
regressions fixes for the old omap_udc driver done by Aaro Koskinen.

We're also reverting an old patch on dwc3 which is, now, known to
break USB certification in some cases.

We have a fix on u_ether for an unsafe list iteration.

* tag 'fixes-for-v4.20-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/balbi/usb:
  usb: gadget: u_ether: fix unsafe list iteration
  USB: omap_udc: fix rejection of out transfers when DMA is used
  USB: omap_udc: fix USB gadget functionality on Palm Tungsten E
  USB: omap_udc: fix omap_udc_start() on 15xx machines
  USB: omap_udc: fix crashes on probe error and module removal
  USB: omap_udc: use devm_request_irq()
  Revert "usb: dwc3: gadget: skip Set/Clear Halt when invalid"

5 years agoMerge tag 'drm-misc-fixes-2018-11-28-1' of git://anongit.freedesktop.org/drm/drm...
Dave Airlie [Thu, 29 Nov 2018 00:11:02 +0000 (10:11 +1000)]
Merge tag 'drm-misc-fixes-2018-11-28-1' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

- mst: Don't try to validate ports while destroying them (Lyude)
- Revert: Don't try to validate ports while destroying them (Lyude)
- core: Don't set device to master unless set_master succeeds (Sergio)
- meson: Do vblank_on/off on enable/disable (Neil)
- meson: Use fast_io regmap option to avoid sleeping in irq ctx (Lyude)
- meson: Don't walk off the end of the OSD EOTF LUTs (Lyude)

Cc: Lyude Paul <lyude@redhat.com>
Cc: Sergio Correia <sergio@correia.cc>
Cc: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Sean Paul <sean@poorly.run>
Link: https://patchwork.freedesktop.org/patch/msgid/20181128212936.GA21379@art_vandelay
5 years agoMerge branch 'drm-fixes-4.20' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Thu, 29 Nov 2018 00:07:33 +0000 (10:07 +1000)]
Merge branch 'drm-fixes-4.20' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

Fixes for 4.20.  Nothing major.
- DC DP MST fix
- GPUVM fix for huge page mapping
- RLC fix for vega20

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181128195905.2897-1-alexander.deucher@amd.com
5 years agoMerge tag 'drm-intel-fixes-2018-11-28' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Wed, 28 Nov 2018 23:59:12 +0000 (09:59 +1000)]
Merge tag 'drm-intel-fixes-2018-11-28' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

Just gvt-fixes-2018-11-26

""One to correct MOCS registers load on engine list, one for rpm lock
warning fix, and another for use-after-free fix for partial ggtt
list destroy. "

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20181128180648.GA17585@jlahtine-desk.ger.corp.intel.com
5 years agoMerge tag 'du-fixes-20181126' of git://linuxtv.org/pinchartl/media into drm-fixes
Dave Airlie [Wed, 28 Nov 2018 23:56:29 +0000 (09:56 +1000)]
Merge tag 'du-fixes-20181126' of git://linuxtv.org/pinchartl/media into drm-fixes

R-Car DU v4.20 regression fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/8134504.ZSXK7gKU4H@avalon
5 years agodrm/ast: fixed reading monitor EDID not stable issue
Y.C. Chen [Thu, 22 Nov 2018 03:56:28 +0000 (11:56 +0800)]
drm/ast: fixed reading monitor EDID not stable issue

v1: over-sample data to increase the stability with some specific monitors
v2: refine to avoid infinite loop
v3: remove un-necessary "volatile" declaration

[airlied: fix two checkpatch warnings]

Signed-off-by: Y.C. Chen <yc_chen@aspeedtech.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1542858988-1127-1-git-send-email-yc_chen@aspeedtech.com
5 years agodrm/ast: Fix incorrect free on ioregs
Sam Bobroff [Mon, 5 Nov 2018 05:57:47 +0000 (16:57 +1100)]
drm/ast: Fix incorrect free on ioregs

If the platform has no IO space, ioregs is placed next to the already
allocated regs. In this case, it should not be separately freed.

This prevents a kernel warning from __vunmap "Trying to vfree()
nonexistent vm area" when unloading the driver.

Fixes: 0dd68309b9c5 ("drm/ast: Try to use MMIO registers when PIO isn't supported")
Signed-off-by: Sam Bobroff <sbobroff@linux.ibm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
5 years agoRevert "drm/dp_mst: Skip validating ports during destruction, just ref"
Lyude Paul [Wed, 28 Nov 2018 21:00:05 +0000 (16:00 -0500)]
Revert "drm/dp_mst: Skip validating ports during destruction, just ref"

This reverts commit:

c54c7374ff44 ("drm/dp_mst: Skip validating ports during destruction, just ref")

ugh.

In drm_dp_destroy_connector_work(), we have a pretty good chance of
freeing the actual struct drm_dp_mst_port. However, after destroying
things we send a hotplug through (*mgr->cbs->hotplug)(mgr) which is
where the problems start.

For i915, this calls all the way down to the fbcon probing helpers,
which start trying to access the port in a modeset.

[   45.062001] ==================================================================
[   45.062112] BUG: KASAN: use-after-free in ex_handler_refcount+0x146/0x180
[   45.062196] Write of size 4 at addr ffff8882b4b70968 by task kworker/3:1/53

[   45.062325] CPU: 3 PID: 53 Comm: kworker/3:1 Kdump: loaded Tainted: G           O      4.20.0-rc4Lyude-Test+ #3
[   45.062442] Hardware name: LENOVO 20BWS1KY00/20BWS1KY00, BIOS JBET71WW (1.35 ) 09/14/2018
[   45.062554] Workqueue: events drm_dp_destroy_connector_work [drm_kms_helper]
[   45.062641] Call Trace:
[   45.062685]  dump_stack+0xbd/0x15a
[   45.062735]  ? dump_stack_print_info.cold.0+0x1b/0x1b
[   45.062801]  ? printk+0x9f/0xc5
[   45.062847]  ? kmsg_dump_rewind_nolock+0xe4/0xe4
[   45.062909]  ? ex_handler_refcount+0x146/0x180
[   45.062970]  print_address_description+0x71/0x239
[   45.063036]  ? ex_handler_refcount+0x146/0x180
[   45.063095]  kasan_report.cold.5+0x242/0x30b
[   45.063155]  __asan_report_store4_noabort+0x1c/0x20
[   45.063313]  ex_handler_refcount+0x146/0x180
[   45.063371]  ? ex_handler_clear_fs+0xb0/0xb0
[   45.063428]  fixup_exception+0x98/0xd7
[   45.063484]  ? raw_notifier_call_chain+0x20/0x20
[   45.063548]  do_trap+0x6d/0x210
[   45.063605]  ? _GLOBAL__sub_I_65535_1_drm_dp_aux_unregister_devnode+0x2f/0x1c6 [drm_kms_helper]
[   45.063732]  do_error_trap+0xc0/0x170
[   45.063802]  ? _GLOBAL__sub_I_65535_1_drm_dp_aux_unregister_devnode+0x2f/0x1c6 [drm_kms_helper]
[   45.063929]  do_invalid_op+0x3b/0x50
[   45.063997]  ? _GLOBAL__sub_I_65535_1_drm_dp_aux_unregister_devnode+0x2f/0x1c6 [drm_kms_helper]
[   45.064103]  invalid_op+0x14/0x20
[   45.064162] RIP: 0010:_GLOBAL__sub_I_65535_1_drm_dp_aux_unregister_devnode+0x2f/0x1c6 [drm_kms_helper]
[   45.064274] Code: 00 48 c7 c7 80 fe 53 a0 48 89 e5 e8 5b 6f 26 e1 5d c3 48 8d 0e 0f 0b 48 8d 0b 0f 0b 48 8d 0f 0f 0b 48 8d 0f 0f 0b 49 8d 4d 00 <0f> 0b 49 8d 0e 0f 0b 48 8d 08 0f 0b 49 8d 4d 00 0f 0b 48 8d 0b 0f
[   45.064569] RSP: 0018:ffff8882b789ee10 EFLAGS: 00010282
[   45.064637] RAX: ffff8882af47ae70 RBX: ffff8882af47aa60 RCX: ffff8882b4b70968
[   45.064723] RDX: ffff8882af47ae70 RSI: 0000000000000008 RDI: ffff8882b788bdb8
[   45.064808] RBP: ffff8882b789ee28 R08: ffffed1056f13db4 R09: ffffed1056f13db3
[   45.064894] R10: ffffed1056f13db3 R11: ffff8882b789ed9f R12: ffff8882af47ad28
[   45.064980] R13: ffff8882b4b70968 R14: ffff8882acd86728 R15: ffff8882b4b75dc8
[   45.065084]  drm_dp_mst_reset_vcpi_slots+0x12/0x80 [drm_kms_helper]
[   45.065225]  intel_mst_disable_dp+0xda/0x180 [i915]
[   45.065361]  intel_encoders_disable.isra.107+0x197/0x310 [i915]
[   45.065498]  haswell_crtc_disable+0xbe/0x400 [i915]
[   45.065622]  ? i9xx_disable_plane+0x1c0/0x3e0 [i915]
[   45.065750]  intel_atomic_commit_tail+0x74e/0x3e60 [i915]
[   45.065884]  ? intel_pre_plane_update+0xbc0/0xbc0 [i915]
[   45.065968]  ? drm_atomic_helper_swap_state+0x88b/0x1d90 [drm_kms_helper]
[   45.066054]  ? kasan_check_write+0x14/0x20
[   45.066165]  ? i915_gem_track_fb+0x13a/0x330 [i915]
[   45.066277]  ? i915_sw_fence_complete+0xe9/0x140 [i915]
[   45.066406]  ? __i915_sw_fence_complete+0xc50/0xc50 [i915]
[   45.066540]  intel_atomic_commit+0x72e/0xef0 [i915]
[   45.066635]  ? drm_dev_dbg+0x200/0x200 [drm]
[   45.066764]  ? intel_atomic_commit_tail+0x3e60/0x3e60 [i915]
[   45.066898]  ? intel_atomic_commit_tail+0x3e60/0x3e60 [i915]
[   45.067001]  drm_atomic_commit+0xc4/0xf0 [drm]
[   45.067074]  restore_fbdev_mode_atomic+0x562/0x780 [drm_kms_helper]
[   45.067166]  ? drm_fb_helper_debug_leave+0x690/0x690 [drm_kms_helper]
[   45.067249]  ? kasan_check_read+0x11/0x20
[   45.067324]  restore_fbdev_mode+0x127/0x4b0 [drm_kms_helper]
[   45.067364]  ? kasan_check_read+0x11/0x20
[   45.067406]  drm_fb_helper_restore_fbdev_mode_unlocked+0x164/0x200 [drm_kms_helper]
[   45.067462]  ? drm_fb_helper_hotplug_event+0x30/0x30 [drm_kms_helper]
[   45.067508]  ? kasan_check_write+0x14/0x20
[   45.070360]  ? mutex_unlock+0x22/0x40
[   45.073748]  drm_fb_helper_set_par+0xb2/0xf0 [drm_kms_helper]
[   45.075846]  drm_fb_helper_hotplug_event.part.33+0x1cd/0x290 [drm_kms_helper]
[   45.078088]  drm_fb_helper_hotplug_event+0x1c/0x30 [drm_kms_helper]
[   45.082614]  intel_fbdev_output_poll_changed+0x9f/0x140 [i915]
[   45.087069]  drm_kms_helper_hotplug_event+0x67/0x90 [drm_kms_helper]
[   45.089319]  intel_dp_mst_hotplug+0x37/0x50 [i915]
[   45.091496]  drm_dp_destroy_connector_work+0x510/0x6f0 [drm_kms_helper]
[   45.093675]  ? drm_dp_update_payload_part1+0x1220/0x1220 [drm_kms_helper]
[   45.095851]  ? kasan_check_write+0x14/0x20
[   45.098473]  ? kasan_check_read+0x11/0x20
[   45.101155]  ? strscpy+0x17c/0x530
[   45.103808]  ? __switch_to_asm+0x34/0x70
[   45.106456]  ? syscall_return_via_sysret+0xf/0x7f
[   45.109711]  ? read_word_at_a_time+0x20/0x20
[   45.113138]  ? __switch_to_asm+0x40/0x70
[   45.116529]  ? __switch_to_asm+0x34/0x70
[   45.119891]  ? __switch_to_asm+0x40/0x70
[   45.123224]  ? __switch_to_asm+0x34/0x70
[   45.126540]  ? __switch_to_asm+0x34/0x70
[   45.129824]  process_one_work+0x88d/0x15d0
[   45.133172]  ? pool_mayday_timeout+0x850/0x850
[   45.136459]  ? pci_mmcfg_check_reserved+0x110/0x128
[   45.139739]  ? wake_q_add+0xb0/0xb0
[   45.143010]  ? check_preempt_wakeup+0x652/0x1050
[   45.146304]  ? worker_enter_idle+0x29e/0x740
[   45.149589]  ? __schedule+0x1ec0/0x1ec0
[   45.152937]  ? kasan_check_read+0x11/0x20
[   45.156179]  ? _raw_spin_lock_irq+0xa3/0x130
[   45.159382]  ? _raw_read_unlock_irqrestore+0x30/0x30
[   45.162542]  ? kasan_check_write+0x14/0x20
[   45.165657]  worker_thread+0x1a5/0x1470
[   45.168725]  ? set_load_weight+0x2e0/0x2e0
[   45.171755]  ? process_one_work+0x15d0/0x15d0
[   45.174806]  ? __switch_to_asm+0x34/0x70
[   45.177645]  ? __switch_to_asm+0x40/0x70
[   45.180323]  ? __switch_to_asm+0x34/0x70
[   45.182936]  ? __switch_to_asm+0x40/0x70
[   45.185539]  ? __switch_to_asm+0x34/0x70
[   45.188100]  ? __switch_to_asm+0x40/0x70
[   45.190628]  ? __schedule+0x7d4/0x1ec0
[   45.193143]  ? save_stack+0xa9/0xd0
[   45.195632]  ? kasan_check_write+0x10/0x20
[   45.198162]  ? kasan_kmalloc+0xc4/0xe0
[   45.200609]  ? kmem_cache_alloc_trace+0xdd/0x190
[   45.203046]  ? kthread+0x9f/0x3b0
[   45.205470]  ? ret_from_fork+0x35/0x40
[   45.207876]  ? unwind_next_frame+0x43/0x50
[   45.210273]  ? __save_stack_trace+0x82/0x100
[   45.212658]  ? deactivate_slab.isra.67+0x3d4/0x580
[   45.215026]  ? default_wake_function+0x35/0x50
[   45.217399]  ? kasan_check_read+0x11/0x20
[   45.219825]  ? _raw_spin_lock_irqsave+0xae/0x140
[   45.222174]  ? __lock_text_start+0x8/0x8
[   45.224521]  ? replenish_dl_entity.cold.62+0x4f/0x4f
[   45.226868]  ? __kthread_parkme+0x87/0xf0
[   45.229200]  kthread+0x2f7/0x3b0
[   45.231557]  ? process_one_work+0x15d0/0x15d0
[   45.233923]  ? kthread_park+0x120/0x120
[   45.236249]  ret_from_fork+0x35/0x40

[   45.240875] Allocated by task 242:
[   45.243136]  save_stack+0x43/0xd0
[   45.245385]  kasan_kmalloc+0xc4/0xe0
[   45.247597]  kmem_cache_alloc_trace+0xdd/0x190
[   45.249793]  drm_dp_add_port+0x1e0/0x2170 [drm_kms_helper]
[   45.252000]  drm_dp_send_link_address+0x4a7/0x740 [drm_kms_helper]
[   45.254389]  drm_dp_check_and_send_link_address+0x1a7/0x210 [drm_kms_helper]
[   45.256803]  drm_dp_mst_link_probe_work+0x6f/0xb0 [drm_kms_helper]
[   45.259200]  process_one_work+0x88d/0x15d0
[   45.261597]  worker_thread+0x1a5/0x1470
[   45.264038]  kthread+0x2f7/0x3b0
[   45.266371]  ret_from_fork+0x35/0x40

[   45.270937] Freed by task 53:
[   45.273170]  save_stack+0x43/0xd0
[   45.275382]  __kasan_slab_free+0x139/0x190
[   45.277604]  kasan_slab_free+0xe/0x10
[   45.279826]  kfree+0x99/0x1b0
[   45.282044]  drm_dp_free_mst_port+0x4a/0x60 [drm_kms_helper]
[   45.284330]  drm_dp_destroy_connector_work+0x43e/0x6f0 [drm_kms_helper]
[   45.286660]  process_one_work+0x88d/0x15d0
[   45.288934]  worker_thread+0x1a5/0x1470
[   45.291231]  kthread+0x2f7/0x3b0
[   45.293547]  ret_from_fork+0x35/0x40

[   45.298206] The buggy address belongs to the object at ffff8882b4b70968
                which belongs to the cache kmalloc-2k of size 2048
[   45.303047] The buggy address is located 0 bytes inside of
                2048-byte region [ffff8882b4b70968ffff8882b4b71168)
[   45.308010] The buggy address belongs to the page:
[   45.310477] page:ffffea000ad2dc00 count:1 mapcount:0 mapping:ffff8882c080cf40 index:0x0 compound_mapcount: 0
[   45.313051] flags: 0x8000000000010200(slab|head)
[   45.315635] raw: 8000000000010200 ffffea000aac2808 ffffea000abe8608 ffff8882c080cf40
[   45.318300] raw: 0000000000000000 00000000000d000d 00000001ffffffff 0000000000000000
[   45.320966] page dumped because: kasan: bad access detected

[   45.326312] Memory state around the buggy address:
[   45.329085]  ffff8882b4b70800: fb fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[   45.331845]  ffff8882b4b70880: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[   45.334584] >ffff8882b4b70900: fc fc fc fc fc fc fc fc fc fc fc fc fc fb fb fb
[   45.337302]                                                           ^
[   45.340061]  ffff8882b4b70980: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[   45.342910]  ffff8882b4b70a00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
[   45.345748] ==================================================================

So, this definitely isn't a fix that we want. This being said; there's
no real easy fix for this problem because of some of the catch-22's of
the MST helpers current design. For starters; we always need to validate
a port with drm_dp_get_validated_port_ref(), but validation relies on
the lifetime of the port in the actual topology. So once the port is
gone, it can't be validated again.

If we were to try to make the payload helpers not use port validation,
then we'd cause another problem: if the port isn't validated, it could
be freed and we'd just start causing more KASAN issues. There are
already hacks that attempt to workaround this in
drm_dp_mst_destroy_connector_work() by re-initializing the kref so that
it can be used again and it's memory can be freed once the VCPI helpers
finish removing the port's respective payloads. But none of these really
do anything helpful since the port still can't be validated since it's
gone from the topology. Also, that workaround is immensely confusing to
read through.

What really needs to be done in order to fix this is to teach DRM how to
track the lifetime of the structs for MST ports and branch devices
separately from their lifetime in the actual topology. Simply put; this
means having two different krefs-one that removes the port/branch device
from the topology, and one that finally calls kfree(). This would let us
simplify things, since we'd now be able to keep ports around without
having to keep them in the topology at the same time, which is exactly
what we need in order to teach our VCPI helpers to only validate ports
when it's actually necessary without running the risk of trying to use
unallocated memory.

Such a fix is on it's way, but for now let's play it safe and just
revert this. If this bug has been around for well over a year, we can
wait a little while to get an actual proper fix here.

Signed-off-by: Lyude Paul <lyude@redhat.com>
Fixes: c54c7374ff44 ("drm/dp_mst: Skip validating ports during destruction, just ref")
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: Sean Paul <sean@poorly.run>
Cc: Jerry Zuo <Jerry.Zuo@amd.com>
Cc: Harry Wentland <Harry.Wentland@amd.com>
Cc: stable@vger.kernel.org # v4.6+
Acked-by: Sean Paul <sean@poorly.run>
Link: https://patchwork.freedesktop.org/patch/msgid/20181128210005.24434-1-lyude@redhat.com
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Wed, 28 Nov 2018 20:53:48 +0000 (12:53 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) ARM64 JIT fixes for subprog handling from Daniel Borkmann.

 2) Various sparc64 JIT bug fixes (fused branch convergance, frame
    pointer usage detection logic, PSEODU call argument handling).

 3) Fix to use BH locking in nf_conncount, from Taehee Yoo.

 4) Fix race of TX skb freeing in ipheth driver, from Bernd Eckstein.

 5) Handle return value of TX NAPI completion properly in lan743x
    driver, from Bryan Whitehead.

 6) MAC filter deletion in i40e driver clears wrong state bit, from
    Lihong Yang.

 7) Fix use after free in rionet driver, from Pan Bian.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (53 commits)
  s390/qeth: fix length check in SNMP processing
  net: hisilicon: remove unexpected free_netdev
  rapidio/rionet: do not free skb before reading its length
  i40e: fix kerneldoc for xsk methods
  ixgbe: recognize 1000BaseLX SFP modules as 1Gbps
  i40e: Fix deletion of MAC filters
  igb: fix uninitialized variables
  netfilter: nf_tables: deactivate expressions in rule replecement routine
  lan743x: Enable driver to work with LAN7431
  tipc: fix lockdep warning during node delete
  lan743x: fix return value for lan743x_tx_napi_poll
  net: via: via-velocity: fix spelling mistake "alignement" -> "alignment"
  qed: fix spelling mistake "attnetion" -> "attention"
  net: thunderx: fix NULL pointer dereference in nic_remove
  sctp: increase sk_wmem_alloc when head->truesize is increased
  firestream: fix spelling mistake: "Inititing" -> "Initializing"
  net: phy: add workaround for issue where PHY driver doesn't bind to the device
  usbnet: ipheth: fix potential recvmsg bug and recvmsg bug 2
  sparc: Adjust bpf JIT prologue for PSEUDO calls.
  bpf, doc: add entries of who looks over which jits
  ...

5 years agoMerge tag 'xtensa-20181128' of git://github.com/jcmvbkbc/linux-xtensa
Linus Torvalds [Wed, 28 Nov 2018 20:51:10 +0000 (12:51 -0800)]
Merge tag 'xtensa-20181128' of git://github.com/jcmvbkbc/linux-xtensa

Pull Xtensa fixes from Max Filippov:

 - fix kernel exception on userspace access to a currently disabled
   coprocessor

 - fix coprocessor data saving/restoring in configurations with multiple
   coprocessors

 - fix ptrace access to coprocessor data on configurations with multiple
   coprocessors with high alignment requirements

* tag 'xtensa-20181128' of git://github.com/jcmvbkbc/linux-xtensa:
  xtensa: fix coprocessor part of ptrace_{get,set}xregs
  xtensa: fix coprocessor context offset definitions
  xtensa: enable coprocessors that are being flushed

5 years agodrm/amdgpu: Add delay after enable RLC ucode
shaoyunl [Thu, 22 Nov 2018 16:45:24 +0000 (11:45 -0500)]
drm/amdgpu: Add delay after enable RLC ucode

Driver shouldn't try to access any GFX registers until RLC is idle.
During the test, it took 12 seconds for RLC to clear the BUSY bit
in RLC_GPM_STAT register which is un-acceptable for driver.
As per RLC engineer, it would take RLC Ucode less than 10,000 GFXCLK
cycles to finish its critical section. In a lowest 300M enginer clock
setting(default from vbios), 50 us delay is enough.

This commit fix the hang when RLC introduce the work around for XGMI
which requires more cycles to setup more registers than normal

Signed-off-by: shaoyunl <shaoyun.liu@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amdgpu: Avoid endless loop in GPUVM fragment processing
Felix Kuehling [Sun, 25 Nov 2018 04:25:04 +0000 (23:25 -0500)]
drm/amdgpu: Avoid endless loop in GPUVM fragment processing

Don't bounce back to the root level for fragment processing, because
huge pages are not supported at that level. This is unlikely to happen
with the default VM size on Vega, but can be exposed by limiting the
VM size with the amdgpu.vm_size module parameter.

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agodrm/amdgpu: Cast to uint64_t before left shift
Felix Kuehling [Sun, 25 Nov 2018 03:46:23 +0000 (22:46 -0500)]
drm/amdgpu: Cast to uint64_t before left shift

Avoid potential integer overflows with left shift in huge-page mapping
code by casting the operand to uin64_t first.

Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agoMerge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net...
David S. Miller [Wed, 28 Nov 2018 19:33:35 +0000 (11:33 -0800)]
Merge branch '1GbE' of git://git./linux/kernel/git/jkirsher/net-queue

Jeff Kirsher says:

====================
Intel Wired LAN Driver Fixes 2018-11-28

This series contains fixes to igb, ixgbe and i40e.

Yunjian Wang from Huawei resolves a variable that could potentially be
NULL before it is used.

Lihong fixes an i40e issue which goes back to 4.17 kernels, where
deleting any of the MAC filters was causing the incorrect syncing for
the PF.

Josh Elsasser caught that there were missing enum values in the link
capabilities for x550 devices, which was preventing link for 1000BaseLX
SFP modules.

Jan fixes the function header comments for XSK methods.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agos390/qeth: fix length check in SNMP processing
Julian Wiedmann [Wed, 28 Nov 2018 15:20:50 +0000 (16:20 +0100)]
s390/qeth: fix length check in SNMP processing

The response for a SNMP request can consist of multiple parts, which
the cmd callback stages into a kernel buffer until all parts have been
received. If the callback detects that the staging buffer provides
insufficient space, it bails out with error.
This processing is buggy for the first part of the response - while it
initially checks for a length of 'data_len', it later copies an
additional amount of 'offsetof(struct qeth_snmp_cmd, data)' bytes.

Fix the calculation of 'data_len' for the first part of the response.
This also nicely cleans up the memcpy code.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Julian Wiedmann <jwi@linux.ibm.com>
Reviewed-by: Ursula Braun <ubraun@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Wed, 28 Nov 2018 19:02:45 +0000 (11:02 -0800)]
Merge git://git./pub/scm/linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains Netfilter fixes for net:

1) Disable BH while holding list spinlock in nf_conncount, from
   Taehee Yoo.

2) List corruption in nf_conncount, also from Taehee.

3) Fix race that results in leaving around an empty list node in
   nf_conncount, from Taehee Yoo.

4) Proper chain handling for inactive chains from the commit path,
   from Florian Westphal. This includes a selftest for this.

5) Do duplicate rule handles when replacing rules, also from Florian.

6) Remove net_exit path in xt_RATEEST that results in splat, from Taehee.

7) Possible use-after-free in nft_compat when releasing extensions.
   From Florian.

8) Memory leak in xt_hashlimit, from Taehee.

9) Call ip_vs_dst_notifier after ipv6_dev_notf, from Xin Long.

10) Fix cttimeout with udplite and gre, from Florian.

11) Preserve oif for IPv6 link-local generated traffic from mangle
    table, from Alin Nastac.

12) Missing error handling in masquerade notifiers, from Taehee Yoo.

13) Use mutex to protect registration/unregistration of masquerade
    extensions in order to prevent a race, from Taehee.

14) Incorrect condition check in tree_nodes_free(), also from Taehee.

15) Fix chain counter leak in rule replacement path, from Taehee.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: hisilicon: remove unexpected free_netdev
Pan Bian [Wed, 28 Nov 2018 07:30:24 +0000 (15:30 +0800)]
net: hisilicon: remove unexpected free_netdev

The net device ndev is freed via free_netdev when failing to register
the device. The control flow then jumps to the error handling code
block. ndev is used and freed again. Resulting in a use-after-free bug.

Signed-off-by: Pan Bian <bianpan2016@163.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agorapidio/rionet: do not free skb before reading its length
Pan Bian [Wed, 28 Nov 2018 06:53:19 +0000 (14:53 +0800)]
rapidio/rionet: do not free skb before reading its length

skb is freed via dev_kfree_skb_any, however, skb->len is read then. This
may result in a use-after-free bug.

Fixes: e6161d64263 ("rapidio/rionet: rework driver initialization and removal")
Signed-off-by: Pan Bian <bianpan2016@163.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoi40e: fix kerneldoc for xsk methods
Jan Sokolowski [Tue, 27 Nov 2018 17:35:35 +0000 (09:35 -0800)]
i40e: fix kerneldoc for xsk methods

One method, xsk_umem_setup, had an incorrect kernel doc
description, which has been corrected.

Also fixes small typos found in the comments.

Signed-off-by: Jan Sokolowski <jan.sokolowski@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoMerge tag 'for-4.20-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Wed, 28 Nov 2018 16:38:20 +0000 (08:38 -0800)]
Merge tag 'for-4.20-rc4-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fixes from David Sterba:
 "Some of these bugs are being hit during testing so we'd like to get
  them merged, otherwise there are usual stability fixes for stable
  trees"

* tag 'for-4.20-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: relocation: set trans to be NULL after ending transaction
  Btrfs: fix race between enabling quotas and subvolume creation
  Btrfs: send, fix infinite loop due to directory rename dependencies
  Btrfs: ensure path name is null terminated at btrfs_control_ioctl
  Btrfs: fix rare chances for data loss when doing a fast fsync
  btrfs: Always try all copies when reading extent buffers

5 years agoMerge tag 'spi-fix-v4.20-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/brooni...
Linus Torvalds [Wed, 28 Nov 2018 16:33:55 +0000 (08:33 -0800)]
Merge tag 'spi-fix-v4.20-rc4' of git://git./linux/kernel/git/broonie/spi

Pull spi fixes from Mark Brown:
 "A few driver specific fixes here, nothing big or that stands out for
  anyone other than the driver users.

  The omap2-mcspi fix is for issues that started showing up with a
  change in defconfig in this release to make cpuidle get turned on by
  default"

* tag 'spi-fix-v4.20-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: omap2-mcspi: Add missing suspend and resume calls
  spi: mediatek: use correct mata->xfer_len when in fifo transfer
  spi: uniphier: fix incorrect property items

5 years agoixgbe: recognize 1000BaseLX SFP modules as 1Gbps
Josh Elsasser [Sat, 24 Nov 2018 20:57:33 +0000 (12:57 -0800)]
ixgbe: recognize 1000BaseLX SFP modules as 1Gbps

Add the two 1000BaseLX enum values to the X550's check for 1Gbps modules,
allowing the core driver code to establish a link over this SFP type.

This is done by the out-of-tree driver but the fix wasn't in mainline.

Fixes: e23f33367882 ("ixgbe: Fix 1G and 10G link stability for X550EM_x SFP+”)
Fixes: 6a14ee0cfb19 ("ixgbe: Add X550 support function pointers")
Signed-off-by: Josh Elsasser <jelsasser@appneta.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Wed, 28 Nov 2018 16:29:18 +0000 (08:29 -0800)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "Bugfixes, many of them reported by syzkaller and mostly predating the
  merge window"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  kvm: svm: Ensure an IBPB on all affected CPUs when freeing a vmcb
  kvm: mmu: Fix race in emulated page table writes
  KVM: nVMX: vmcs12 revision_id is always VMCS12_REVISION even when copied from eVMCS
  KVM: nVMX: Verify eVMCS revision id match supported eVMCS version on eVMCS VMPTRLD
  KVM: nVMX/nSVM: Fix bug which sets vcpu->arch.tsc_offset to L1 tsc_offset
  x86/kvm/vmx: fix old-style function declaration
  KVM: x86: fix empty-body warnings
  KVM: VMX: Update shared MSRs to be saved/restored on MSR_EFER.LMA changes
  KVM: x86: Fix kernel info-leak in KVM_HC_CLOCK_PAIRING hypercall
  KVM: nVMX: Fix kernel info-leak when enabling KVM_CAP_HYPERV_ENLIGHTENED_VMCS more than once
  svm: Add mutex_lock to protect apic_access_page_done on AMD systems
  KVM: X86: Fix scan ioapic use-before-initialization
  KVM: LAPIC: Fix pv ipis use-before-initialization
  KVM: VMX: re-add ple_gap module parameter
  KVM: PPC: Book3S HV: Fix handling for interrupted H_ENTER_NESTED

5 years agoi40e: Fix deletion of MAC filters
Lihong Yang [Wed, 21 Nov 2018 17:15:37 +0000 (09:15 -0800)]
i40e: Fix deletion of MAC filters

In __i40e_del_filter function, the flag __I40E_MACVLAN_SYNC_PENDING for
the PF state is wrongly set for the VSI. Deleting any of the MAC filters
has caused the incorrect syncing for the PF. Fix it by setting this state
flag to the intended PF.

CC: stable <stable@vger.kernel.org>
Signed-off-by: Lihong Yang <lihong.yang@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoigb: fix uninitialized variables
Yunjian Wang [Tue, 6 Nov 2018 08:27:12 +0000 (16:27 +0800)]
igb: fix uninitialized variables

This patch fixes the variable 'phy_word' may be used uninitialized.

Signed-off-by: Yunjian Wang <wangyunjian@huawei.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
5 years agoblock: add io timeout to sysfs
Weiping Zhang [Wed, 28 Nov 2018 16:04:39 +0000 (00:04 +0800)]
block: add io timeout to sysfs

Give a interface to adjust io timeout(ms) by device.

Signed-off-by: Weiping Zhang <zhangweiping@didiglobal.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoblock: use rcu_work instead of call_rcu to avoid sleep in softirq
Yufen Yu [Wed, 28 Nov 2018 08:42:01 +0000 (16:42 +0800)]
block: use rcu_work instead of call_rcu to avoid sleep in softirq

We recently got a stack by syzkaller like this:

BUG: sleeping function called from invalid context at mm/slab.h:361
in_atomic(): 1, irqs_disabled(): 0, pid: 6644, name: blkid
INFO: lockdep is turned off.
CPU: 1 PID: 6644 Comm: blkid Not tainted 4.4.163-514.55.6.9.x86_64+ #76
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 04/01/2014
 0000000000000000 5ba6a6b879e50c00 ffff8801f6b07b10 ffffffff81cb2194
 0000000041b58ab3 ffffffff833c7745 ffffffff81cb2080 5ba6a6b879e50c00
 0000000000000000 0000000000000001 0000000000000004 0000000000000000
Call Trace:
 <IRQ>  [<ffffffff81cb2194>] __dump_stack lib/dump_stack.c:15 [inline]
 <IRQ>  [<ffffffff81cb2194>] dump_stack+0x114/0x1a0 lib/dump_stack.c:51
 [<ffffffff8129a981>] ___might_sleep+0x291/0x490 kernel/sched/core.c:7675
 [<ffffffff8129ac33>] __might_sleep+0xb3/0x270 kernel/sched/core.c:7637
 [<ffffffff81794c13>] slab_pre_alloc_hook mm/slab.h:361 [inline]
 [<ffffffff81794c13>] slab_alloc_node mm/slub.c:2610 [inline]
 [<ffffffff81794c13>] slab_alloc mm/slub.c:2692 [inline]
 [<ffffffff81794c13>] kmem_cache_alloc_trace+0x2c3/0x5c0 mm/slub.c:2709
 [<ffffffff81cbe9a7>] kmalloc include/linux/slab.h:479 [inline]
 [<ffffffff81cbe9a7>] kzalloc include/linux/slab.h:623 [inline]
 [<ffffffff81cbe9a7>] kobject_uevent_env+0x2c7/0x1150 lib/kobject_uevent.c:227
 [<ffffffff81cbf84f>] kobject_uevent+0x1f/0x30 lib/kobject_uevent.c:374
 [<ffffffff81cbb5b9>] kobject_cleanup lib/kobject.c:633 [inline]
 [<ffffffff81cbb5b9>] kobject_release+0x229/0x440 lib/kobject.c:675
 [<ffffffff81cbb0a2>] kref_sub include/linux/kref.h:73 [inline]
 [<ffffffff81cbb0a2>] kref_put include/linux/kref.h:98 [inline]
 [<ffffffff81cbb0a2>] kobject_put+0x72/0xd0 lib/kobject.c:692
 [<ffffffff8216f095>] put_device+0x25/0x30 drivers/base/core.c:1237
 [<ffffffff81c4cc34>] delete_partition_rcu_cb+0x1d4/0x2f0 block/partition-generic.c:232
 [<ffffffff813c08bc>] __rcu_reclaim kernel/rcu/rcu.h:118 [inline]
 [<ffffffff813c08bc>] rcu_do_batch kernel/rcu/tree.c:2705 [inline]
 [<ffffffff813c08bc>] invoke_rcu_callbacks kernel/rcu/tree.c:2973 [inline]
 [<ffffffff813c08bc>] __rcu_process_callbacks kernel/rcu/tree.c:2940 [inline]
 [<ffffffff813c08bc>] rcu_process_callbacks+0x59c/0x1c70 kernel/rcu/tree.c:2957
 [<ffffffff8120f509>] __do_softirq+0x299/0xe20 kernel/softirq.c:273
 [<ffffffff81210496>] invoke_softirq kernel/softirq.c:350 [inline]
 [<ffffffff81210496>] irq_exit+0x216/0x2c0 kernel/softirq.c:391
 [<ffffffff82c2cd7b>] exiting_irq arch/x86/include/asm/apic.h:652 [inline]
 [<ffffffff82c2cd7b>] smp_apic_timer_interrupt+0x8b/0xc0 arch/x86/kernel/apic/apic.c:926
 [<ffffffff82c2bc25>] apic_timer_interrupt+0xa5/0xb0 arch/x86/entry/entry_64.S:746
 <EOI>  [<ffffffff814cbf40>] ? audit_kill_trees+0x180/0x180
 [<ffffffff8187d2f7>] fd_install+0x57/0x80 fs/file.c:626
 [<ffffffff8180989e>] do_sys_open+0x45e/0x550 fs/open.c:1043
 [<ffffffff818099c2>] SYSC_open fs/open.c:1055 [inline]
 [<ffffffff818099c2>] SyS_open+0x32/0x40 fs/open.c:1050
 [<ffffffff82c299e1>] entry_SYSCALL_64_fastpath+0x1e/0x9a

In softirq context, we call rcu callback function delete_partition_rcu_cb(),
which may allocate memory by kzalloc with GFP_KERNEL flag. If the
allocation cannot be satisfied, it may sleep. However, That is not allowed
in softirq contex.

Although we found this problem on linux 4.4, the latest kernel version
seems to have this problem as well. And it is very similar to the
previous one:
https://lkml.org/lkml/2018/7/9/391

Fix it by using RCU workqueue, which allows sleep.

Reviewed-by: Paul E. McKenney <paulmck@linux.ibm.com>
Signed-off-by: Yufen Yu <yuyufen@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agocachefiles: Fix page leak in cachefiles_read_backing_file while vmscan is active
Kiran Kumar Modukuri [Mon, 24 Sep 2018 02:02:39 +0000 (12:02 +1000)]
cachefiles: Fix page leak in cachefiles_read_backing_file while vmscan is active

[Description]

In a heavily loaded system where the system pagecache is nearing memory
limits and fscache is enabled, pages can be leaked by fscache while trying
read pages from cachefiles backend.  This can happen because two
applications can be reading same page from a single mount, two threads can
be trying to read the backing page at same time.  This results in one of
the threads finding that a page for the backing file or netfs file is
already in the radix tree.  During the error handling cachefiles does not
clean up the reference on backing page, leading to page leak.

[Fix]
The fix is straightforward, to decrement the reference when error is
encountered.

  [dhowells: Note that I've removed the clearance and put of newpage as
   they aren't attested in the commit message and don't appear to actually
   achieve anything since a new page is only allocated is newpage!=NULL and
   any residual new page is cleared before returning.]

[Testing]
I have tested the fix using following method for 12+ hrs.

1) mkdir -p /mnt/nfs ; mount -o vers=3,fsc <server_ip>:/export /mnt/nfs
2) create 10000 files of 2.8MB in a NFS mount.
3) start a thread to simulate heavy VM presssure
   (while true ; do echo 3 > /proc/sys/vm/drop_caches ; sleep 1 ; done)&
4) start multiple parallel reader for data set at same time
   find /mnt/nfs -type f | xargs -P 80 cat > /dev/null &
   find /mnt/nfs -type f | xargs -P 80 cat > /dev/null &
   find /mnt/nfs -type f | xargs -P 80 cat > /dev/null &
   ..
   ..
   find /mnt/nfs -type f | xargs -P 80 cat > /dev/null &
   find /mnt/nfs -type f | xargs -P 80 cat > /dev/null &
5) finally check using cat /proc/fs/fscache/stats | grep -i pages ;
   free -h , cat /proc/meminfo and page-types -r -b lru
   to ensure all pages are freed.

Reviewed-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Shantanu Goel <sgoel01@yahoo.com>
Signed-off-by: Kiran Kumar Modukuri <kiran.modukuri@gmail.com>
[dja: forward ported to current upstream]
Signed-off-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agomtd: nand: Fix memory allocation in nanddev_bbt_init()
Frieder Schrempf [Tue, 27 Nov 2018 07:44:52 +0000 (07:44 +0000)]
mtd: nand: Fix memory allocation in nanddev_bbt_init()

Fix the size of the buffer allocated to store the in-memory BBT.
This bug was previously hidden by a different bug, that was fixed in
commit d098093ba06e ("mtd: nand: Fix nanddev_neraseblocks()").

Fixes: 9c3736a3de21 ("mtd: nand: Add core infrastructure to deal with NAND devices")
Cc: <stable@vger.kernel.org>
Signed-off-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Acked-by: Miquel Raynal <miquel.raynal@bootlin.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
5 years agoblk-mq: fix failure to decrement plug count on single rq removal
Jens Axboe [Wed, 28 Nov 2018 00:07:17 +0000 (17:07 -0700)]
blk-mq: fix failure to decrement plug count on single rq removal

If we yank a 'same_queue_rq' request off the plug list, we should
also decrement the cached request count.

Fixes: 5f0ed774ed29 ("block: sum requests in the plug structure")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agosunvdc: Do not spin in an infinite loop when vio_ldc_send() returns EAGAIN
Young Xiao [Wed, 28 Nov 2018 12:36:39 +0000 (12:36 +0000)]
sunvdc: Do not spin in an infinite loop when vio_ldc_send() returns EAGAIN

__vdc_tx_trigger should only loop on EAGAIN a finite
number of times.

See commit adddc32d6fde ("sunvnet: Do not spin in an
infinite loop when vio_ldc_send() returns EAGAIN") for detail.

Signed-off-by: Young Xiao <YangX92@hotmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agofscache: Fix race in fscache_op_complete() due to split atomic_sub & read
kiran.modukuri [Mon, 26 Nov 2018 15:41:48 +0000 (15:41 +0000)]
fscache: Fix race in fscache_op_complete() due to split atomic_sub & read

The code in fscache_retrieval_complete is using atomic_sub followed by an
atomic_read:

        atomic_sub(n_pages, &op->n_pages);
        if (atomic_read(&op->n_pages) <= 0)
                fscache_op_complete(&op->op, true);

This causes two threads doing a decrement of n_pages to race with each
other seeing the op->refcount 0 at same time - and they end up calling
fscache_op_complete() in both the threads leading to an assertion failure.

Fix this by using atomic_sub_return_relaxed() instead of two calls.  Note
that I'm using 'relaxed' rather than, say, 'release' as there aren't
multiple variables that appear to need ordering across the release.

The oops looks something like:

FS-Cache: Assertion failed
FS-Cache: 0 > 0 is false
...
kernel BUG at /usr/src/linux-4.4.0/fs/fscache/operation.c:449!
...
Workqueue: fscache_operation fscache_op_work_func [fscache]
...
RIP: 0010:[<ffffffffc037eacd>] fscache_op_complete+0x10d/0x180 [fscache]
...
Call Trace:
 [<ffffffffc1464cf9>] cachefiles_read_copier+0x3a9/0x410 [cachefiles]
 [<ffffffffc037e272>] fscache_op_work_func+0x22/0x50 [fscache]
 [<ffffffff81096da0>] process_one_work+0x150/0x3f0
 [<ffffffff8109751a>] worker_thread+0x11a/0x470
 [<ffffffff81808e59>] ? __schedule+0x359/0x980
 [<ffffffff81097400>] ? rescuer_thread+0x310/0x310
 [<ffffffff8109cdd6>] kthread+0xd6/0xf0
 [<ffffffff8109cd00>] ? kthread_park+0x60/0x60
 [<ffffffff8180d0cf>] ret_from_fork+0x3f/0x70
 [<ffffffff8109cd00>] ? kthread_park+0x60/0x60

This seen this in 4.4.x kernels and the same bug affects fscache in latest
upstreams kernels.

Fixes: 1bb4b7f98f36 ("FS-Cache: The retrieval remaining-pages counter needs to be atomic_t")
Signed-off-by: Kiran Kumar Modukuri <kiran.modukuri@gmail.com>
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agocachefiles: Fix an assertion failure when trying to update a failed object
David Howells [Tue, 27 Nov 2018 16:34:55 +0000 (16:34 +0000)]
cachefiles: Fix an assertion failure when trying to update a failed object

If cachefiles gets an error other then ENOENT when trying to look up an
object in the cache (in this case, EACCES), the object state machine will
eventually transition to the DROP_OBJECT state.

This state invokes fscache_drop_object() which tries to sync the auxiliary
data with the cache (this is done lazily since commit 402cb8dda949d) on an
incomplete cache object struct.

The problem comes when cachefiles_update_object_xattr() is called to
rewrite the xattr holding the data.  There's an assertion there that the
cache object points to a dentry as we're going to update its xattr.  The
assertion trips, however, as dentry didn't get set.

Fix the problem by skipping the update in cachefiles if the object doesn't
refer to a dentry.  A better way to do it could be to skip the update from
the DROP_OBJECT state handler in fscache, but that might deny the cache the
opportunity to update intermediate state.

If this error occurs, the kernel log includes lines that look like the
following:

 CacheFiles: Lookup failed error -13
 CacheFiles:
 CacheFiles: Assertion failed
 ------------[ cut here ]------------
 kernel BUG at fs/cachefiles/xattr.c:138!
 ...
 Workqueue: fscache_object fscache_object_work_func [fscache]
 RIP: 0010:cachefiles_update_object_xattr.cold.4+0x18/0x1a [cachefiles]
 ...
 Call Trace:
  cachefiles_update_object+0xdd/0x1c0 [cachefiles]
  fscache_update_aux_data+0x23/0x30 [fscache]
  fscache_drop_object+0x18e/0x1c0 [fscache]
  fscache_object_work_func+0x74/0x2b0 [fscache]
  process_one_work+0x18d/0x340
  worker_thread+0x2e/0x390
  ? pwq_unbound_release_workfn+0xd0/0xd0
  kthread+0x112/0x130
  ? kthread_bind+0x30/0x30
  ret_from_fork+0x35/0x40

Note that there are actually two issues here: (1) EACCES happened on a
cache object and (2) an oops occurred.  I think that the second is a
consequence of the first (it certainly looks like it ought to be).  This
patch only deals with the second.

Fixes: 402cb8dda949 ("fscache: Attach the index key and aux data to the cookie")
Reported-by: Zhibin Li <zhibli@redhat.com>
Signed-off-by: David Howells <dhowells@redhat.com>
5 years agox86/speculation: Provide IBPB always command line options
Thomas Gleixner [Sun, 25 Nov 2018 18:33:56 +0000 (19:33 +0100)]
x86/speculation: Provide IBPB always command line options

Provide the possibility to enable IBPB always in combination with 'prctl'
and 'seccomp'.

Add the extra command line options and rework the IBPB selection to
evaluate the command instead of the mode selected by the STIPB switch case.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185006.144047038@linutronix.de
5 years agox86/speculation: Add seccomp Spectre v2 user space protection mode
Thomas Gleixner [Sun, 25 Nov 2018 18:33:55 +0000 (19:33 +0100)]
x86/speculation: Add seccomp Spectre v2 user space protection mode

If 'prctl' mode of user space protection from spectre v2 is selected
on the kernel command-line, STIBP and IBPB are applied on tasks which
restrict their indirect branch speculation via prctl.

SECCOMP enables the SSBD mitigation for sandboxed tasks already, so it
makes sense to prevent spectre v2 user space to user space attacks as
well.

The Intel mitigation guide documents how STIPB works:

   Setting bit 1 (STIBP) of the IA32_SPEC_CTRL MSR on a logical processor
   prevents the predicted targets of indirect branches on any logical
   processor of that core from being controlled by software that executes
   (or executed previously) on another logical processor of the same core.

Ergo setting STIBP protects the task itself from being attacked from a task
running on a different hyper-thread and protects the tasks running on
different hyper-threads from being attacked.

While the document suggests that the branch predictors are shielded between
the logical processors, the observed performance regressions suggest that
STIBP simply disables the branch predictor more or less completely. Of
course the document wording is vague, but the fact that there is also no
requirement for issuing IBPB when STIBP is used points clearly in that
direction. The kernel still issues IBPB even when STIBP is used until Intel
clarifies the whole mechanism.

IBPB is issued when the task switches out, so malicious sandbox code cannot
mistrain the branch predictor for the next user space task on the same
logical processor.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185006.051663132@linutronix.de
5 years agox86/speculation: Enable prctl mode for spectre_v2_user
Thomas Gleixner [Sun, 25 Nov 2018 18:33:54 +0000 (19:33 +0100)]
x86/speculation: Enable prctl mode for spectre_v2_user

Now that all prerequisites are in place:

 - Add the prctl command line option

 - Default the 'auto' mode to 'prctl'

 - When SMT state changes, update the static key which controls the
   conditional STIBP evaluation on context switch.

 - At init update the static key which controls the conditional IBPB
   evaluation on context switch.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.958421388@linutronix.de
5 years agox86/speculation: Add prctl() control for indirect branch speculation
Thomas Gleixner [Sun, 25 Nov 2018 18:33:53 +0000 (19:33 +0100)]
x86/speculation: Add prctl() control for indirect branch speculation

Add the PR_SPEC_INDIRECT_BRANCH option for the PR_GET_SPECULATION_CTRL and
PR_SET_SPECULATION_CTRL prctls to allow fine grained per task control of
indirect branch speculation via STIBP and IBPB.

Invocations:
 Check indirect branch speculation status with
 - prctl(PR_GET_SPECULATION_CTRL, PR_SPEC_INDIRECT_BRANCH, 0, 0, 0);

 Enable indirect branch speculation with
 - prctl(PR_SET_SPECULATION_CTRL, PR_SPEC_INDIRECT_BRANCH, PR_SPEC_ENABLE, 0, 0);

 Disable indirect branch speculation with
 - prctl(PR_SET_SPECULATION_CTRL, PR_SPEC_INDIRECT_BRANCH, PR_SPEC_DISABLE, 0, 0);

 Force disable indirect branch speculation with
 - prctl(PR_SET_SPECULATION_CTRL, PR_SPEC_INDIRECT_BRANCH, PR_SPEC_FORCE_DISABLE, 0, 0);

See Documentation/userspace-api/spec_ctrl.rst.

Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.866780996@linutronix.de
5 years agox86/speculation: Prepare arch_smt_update() for PRCTL mode
Thomas Gleixner [Sun, 25 Nov 2018 18:33:52 +0000 (19:33 +0100)]
x86/speculation: Prepare arch_smt_update() for PRCTL mode

The upcoming fine grained per task STIBP control needs to be updated on CPU
hotplug as well.

Split out the code which controls the strict mode so the prctl control code
can be added later. Mark the SMP function call argument __unused while at it.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.759457117@linutronix.de
5 years agox86/speculation: Prevent stale SPEC_CTRL msr content
Thomas Gleixner [Wed, 28 Nov 2018 09:56:57 +0000 (10:56 +0100)]
x86/speculation: Prevent stale SPEC_CTRL msr content

The seccomp speculation control operates on all tasks of a process, but
only the current task of a process can update the MSR immediately. For the
other threads the update is deferred to the next context switch.

This creates the following situation with Process A and B:

Process A task 2 and Process B task 1 are pinned on CPU1. Process A task 2
does not have the speculation control TIF bit set. Process B task 1 has the
speculation control TIF bit set.

CPU0 CPU1
MSR bit is set
ProcB.T1 schedules out
ProcA.T2 schedules in
MSR bit is cleared
ProcA.T1
  seccomp_update()
  set TIF bit on ProcA.T2
ProcB.T1 schedules in
MSR is not updated  <-- FAIL

This happens because the context switch code tries to avoid the MSR update
if the speculation control TIF bits of the incoming and the outgoing task
are the same. In the worst case ProcB.T1 and ProcA.T2 are the only tasks
scheduling back and forth on CPU1, which keeps the MSR stale forever.

In theory this could be remedied by IPIs, but chasing the remote task which
could be migrated is complex and full of races.

The straight forward solution is to avoid the asychronous update of the TIF
bit and defer it to the next context switch. The speculation control state
is stored in task_struct::atomic_flags by the prctl and seccomp updates
already.

Add a new TIF_SPEC_FORCE_UPDATE bit and set this after updating the
atomic_flags. Check the bit on context switch and force a synchronous
update of the speculation control if set. Use the same mechanism for
updating the current task.

Reported-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/alpine.DEB.2.21.1811272247140.1875@nanos.tec.linutronix.de
5 years agox86/speculation: Split out TIF update
Thomas Gleixner [Sun, 25 Nov 2018 18:33:51 +0000 (19:33 +0100)]
x86/speculation: Split out TIF update

The update of the TIF_SSBD flag and the conditional speculation control MSR
update is done in the ssb_prctl_set() function directly. The upcoming prctl
support for controlling indirect branch speculation via STIBP needs the
same mechanism.

Split the code out and make it reusable. Reword the comment about updates
for other tasks.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.652305076@linutronix.de
5 years agoptrace: Remove unused ptrace_may_access_sched() and MODE_IBRS
Thomas Gleixner [Sun, 25 Nov 2018 18:33:50 +0000 (19:33 +0100)]
ptrace: Remove unused ptrace_may_access_sched() and MODE_IBRS

The IBPB control code in x86 removed the usage. Remove the functionality
which was introduced for this.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.559149393@linutronix.de
5 years agox86/speculation: Prepare for conditional IBPB in switch_mm()
Thomas Gleixner [Sun, 25 Nov 2018 18:33:49 +0000 (19:33 +0100)]
x86/speculation: Prepare for conditional IBPB in switch_mm()

The IBPB speculation barrier is issued from switch_mm() when the kernel
switches to a user space task with a different mm than the user space task
which ran last on the same CPU.

An additional optimization is to avoid IBPB when the incoming task can be
ptraced by the outgoing task. This optimization only works when switching
directly between two user space tasks. When switching from a kernel task to
a user space task the optimization fails because the previous task cannot
be accessed anymore. So for quite some scenarios the optimization is just
adding overhead.

The upcoming conditional IBPB support will issue IBPB only for user space
tasks which have the TIF_SPEC_IB bit set. This requires to handle the
following cases:

  1) Switch from a user space task (potential attacker) which has
     TIF_SPEC_IB set to a user space task (potential victim) which has
     TIF_SPEC_IB not set.

  2) Switch from a user space task (potential attacker) which has
     TIF_SPEC_IB not set to a user space task (potential victim) which has
     TIF_SPEC_IB set.

This needs to be optimized for the case where the IBPB can be avoided when
only kernel threads ran in between user space tasks which belong to the
same process.

The current check whether two tasks belong to the same context is using the
tasks context id. While correct, it's simpler to use the mm pointer because
it allows to mangle the TIF_SPEC_IB bit into it. The context id based
mechanism requires extra storage, which creates worse code.

When a task is scheduled out its TIF_SPEC_IB bit is mangled as bit 0 into
the per CPU storage which is used to track the last user space mm which was
running on a CPU. This bit can be used together with the TIF_SPEC_IB bit of
the incoming task to make the decision whether IBPB needs to be issued or
not to cover the two cases above.

As conditional IBPB is going to be the default, remove the dubious ptrace
check for the IBPB always case and simply issue IBPB always when the
process changes.

Move the storage to a different place in the struct as the original one
created a hole.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.466447057@linutronix.de
5 years agox86/speculation: Avoid __switch_to_xtra() calls
Thomas Gleixner [Sun, 25 Nov 2018 18:33:48 +0000 (19:33 +0100)]
x86/speculation: Avoid __switch_to_xtra() calls

The TIF_SPEC_IB bit does not need to be evaluated in the decision to invoke
__switch_to_xtra() when:

 - CONFIG_SMP is disabled

 - The conditional STIPB mode is disabled

The TIF_SPEC_IB bit still controls IBPB in both cases so the TIF work mask
checks might invoke __switch_to_xtra() for nothing if TIF_SPEC_IB is the
only set bit in the work masks.

Optimize it out by masking the bit at compile time for CONFIG_SMP=n and at
run time when the static key controlling the conditional STIBP mode is
disabled.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.374062201@linutronix.de
5 years agox86/process: Consolidate and simplify switch_to_xtra() code
Thomas Gleixner [Sun, 25 Nov 2018 18:33:47 +0000 (19:33 +0100)]
x86/process: Consolidate and simplify switch_to_xtra() code

Move the conditional invocation of __switch_to_xtra() into an inline
function so the logic can be shared between 32 and 64 bit.

Remove the handthrough of the TSS pointer and retrieve the pointer directly
in the bitmap handling function. Use this_cpu_ptr() instead of the
per_cpu() indirection.

This is a preparatory change so integration of conditional indirect branch
speculation optimization happens only in one place.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.280855518@linutronix.de
5 years agox86/speculation: Prepare for per task indirect branch speculation control
Tim Chen [Sun, 25 Nov 2018 18:33:46 +0000 (19:33 +0100)]
x86/speculation: Prepare for per task indirect branch speculation control

To avoid the overhead of STIBP always on, it's necessary to allow per task
control of STIBP.

Add a new task flag TIF_SPEC_IB and evaluate it during context switch if
SMT is active and flag evaluation is enabled by the speculation control
code. Add the conditional evaluation to x86_virt_spec_ctrl() as well so the
guest/host switch works properly.

This has no effect because TIF_SPEC_IB cannot be set yet and the static key
which controls evaluation is off. Preparatory patch for adding the control
code.

[ tglx: Simplify the context switch logic and make the TIF evaluation
   depend on SMP=y and on the static key controlling the conditional
   update. Rename it to TIF_SPEC_IB because it controls both STIBP and
   IBPB ]

Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.176917199@linutronix.de
5 years agox86/speculation: Add command line control for indirect branch speculation
Thomas Gleixner [Sun, 25 Nov 2018 18:33:45 +0000 (19:33 +0100)]
x86/speculation: Add command line control for indirect branch speculation

Add command line control for user space indirect branch speculation
mitigations. The new option is: spectre_v2_user=

The initial options are:

    -  on:   Unconditionally enabled
    - off:   Unconditionally disabled
    -auto:   Kernel selects mitigation (default off for now)

When the spectre_v2= command line argument is either 'on' or 'off' this
implies that the application to application control follows that state even
if a contradicting spectre_v2_user= argument is supplied.

Originally-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185005.082720373@linutronix.de
5 years agox86/speculation: Unify conditional spectre v2 print functions
Thomas Gleixner [Sun, 25 Nov 2018 18:33:44 +0000 (19:33 +0100)]
x86/speculation: Unify conditional spectre v2 print functions

There is no point in having two functions and a conditional at the call
site.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185004.986890749@linutronix.de
5 years agox86/speculataion: Mark command line parser data __initdata
Thomas Gleixner [Sun, 25 Nov 2018 18:33:43 +0000 (19:33 +0100)]
x86/speculataion: Mark command line parser data __initdata

No point to keep that around.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185004.893886356@linutronix.de
5 years agox86/speculation: Mark string arrays const correctly
Thomas Gleixner [Sun, 25 Nov 2018 18:33:42 +0000 (19:33 +0100)]
x86/speculation: Mark string arrays const correctly

checkpatch.pl muttered when reshuffling the code:
 WARNING: static const char * array should probably be static const char * const

Fix up all the string arrays.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185004.800018931@linutronix.de
5 years agox86/speculation: Reorder the spec_v2 code
Thomas Gleixner [Sun, 25 Nov 2018 18:33:41 +0000 (19:33 +0100)]
x86/speculation: Reorder the spec_v2 code

Reorder the code so it is better grouped. No functional change.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Casey Schaufler <casey.schaufler@intel.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Jon Masters <jcm@redhat.com>
Cc: Waiman Long <longman9394@gmail.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Dave Stewart <david.c.stewart@intel.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20181125185004.707122879@linutronix.de