coresight: etb10: fixing the right amount of words to read

This patch rectifies the amount of words to read when the internal
buffer is deemed bigger than the amount of space available in the
perf ring buffer.

The amount to read is set to the amount of space in the perf ring
buffer rather than being subtracted by it.

Reported-by: Suzuki K Poulose <Suzuki.Poulose@arm.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit b5af0a26da84b75376706a92c7a58036a0bf3541)

coresight: stm: adding driver for CoreSight STM component

This driver adds support for the STM CoreSight IP block, allowing any
system compoment (HW or SW) to log and aggregate messages via a
single entity.

The CoreSight STM exposes an application defined number of channels
called stimulus port.  Configuration is done using entries in sysfs
and channels made available to userspace via configfs.

Signed-off-by: Pratik Patel <pratikp@codeaurora.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Michael Williams <michael.williams@arm.com>
Signed-off-by: Chunyan Zhang <zhang.chunyan@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 237483aa5cf43105d148d3f03b29eed47c3e6cf9)

coresight: adding path for STM device

>From a core framework point of view an STM device is a source that is
treated the same way as any other tracers.  Unlike tracers though STM
devices are not associated with a CPU.  As such it doesn't make sense
to associate the path from an STM device to its sink with a per-cpu
variable as it is done for tracers.

This patch simply adds another global variable to keep STM paths and the
processing in coresight_enable/disable() is updated to deal with STM
devices properly.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Chunyan Zhang <zhang.chunyan@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit a685d68328f14579b2e68d6a3a2066089cffbf98)

coresight: etm4x: modify q_support type

Because this operation exceed the range of boolean,
so we should modify q_support to unit8 bit.
drvdata->q_support = BMVAL(etmidr0, 15, 16)

Signed-off-by: Li Pengcheng <lipengcheng8@huawei.com>
Signed-off-by: Li Zhong <lizhong11@hisilicon.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 6327a454a8ab0dcab24a647367d216c1b84020c6)

coresight: no need to do the forced type conversion

activated and enable are already unsigned type,
no need to change them to unsigned.

Signed-off-by: Li Pengcheng <lipengcheng8@huawei.com>
Signed-off-by: Li Zhong <lizhong11@hisilicon.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit e8dc27d0ee458f9622b50e2d9476719b3a0e686b)

coresight: removing gratuitous boot time log messages

Removing boot time log for drivers that don't report useful information
other than they came up properly.  The same information can be found in
sysFS once the system has booted and as such doesn't provide any value
in the boot log.

Reported-by: Mark Brown <broonie@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit ef0fd640e3312b8164ec43e1eff24769a7c08b7f)

coresight: etb10: splitting sysFS "status" entry

The sysFS "status" entry conveys a wealth of information about
the status of the HW but goes agains the sysFS rule of one topic
per file.

This patch rectify the situation by adding read-only entries for
each of the field formaly displayed by "status".  The ABI
documentation is kept up to date.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit ad352acbb9d606a5facff31fd96b05d0346726b1)

coresight: moving coresight_simple_func() to header file

Macro "coresight_simple_func()" can be used by several drivers.
As such making the structure type generic and moving to a
globally available header file.  That way individual drivers
can use the functionality by simply specifying the structure
they need to work with.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 154f3520fe1cdef9009909dc62828eb2d7635631)

coresight: etm4x: implementing the perf PMU API

Adding a set of API allowing the Perf core to treat ETMv4
tracers like other PMUs.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 37fbbdbde9ad3722a7a18beab936825a6ff322bf)

coresight: etm4x: implementing user/kernel mode tracing

Adding new mode to limit tracing to kernel or user space.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 4f6fce54528e0382281cf199635d098e4b108357)

coresight: etm4x: moving etm_drvdata::enable to atomic field

Similarly to ETMv3, moving etmv4_drvdata::enable to an atomic
type that gives the 'mode' of a tracer and prevents multiple,
simultanious access by different subsystems.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit c38a9ec2b2c12c38abca0b7954ed793f26969835)

coresight: etm4x: unlocking tracers in default arch init

As with the ETMv3.x driver, calling 'smp_call_function_single()'
twice in a row is highly ineffective.  As such moving function
'etm4_os_unlock()' before the default initialisation takes
place, which results in the same outcome.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 66bbbb77540e846b9aac4c9467aca936128951bf)

coresight: etm4x: splitting etmv4 default configuration

Splitting and updating the default initialisation for each etmv4
configuration so that it can be called at the beginning of each
session rather than initialisation time only.

Since the trace ID isn't expected to change with every session,
moving it with the default tracer initialisation.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit fc208abef39279903887bea955139f64bf0bbb12)

coresight: etm4x: splitting struct etmv4_drvdata

Similar to what was done on etm3x, splitting driver structure
etmv4_drvdata in two.  One half is concerned with the HW
characteristics that are generally static in nature.  The other
half deals with user configuration and will change from one
trace session to another.

No gain/loss of functionality is incurred from this patch.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 54ff892b76c68ea3fa0ba53a0cdc4508b35aee6f)

coresight: etm4x: adding config and traceid registers

Adding new sysFS management interface to query the configuration
and the traceid registers.  Both are required to convey information
to the perf cmd line tools when using ETMv4 tracers as PMU.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 7c38aa4b03b3fc6ce17e5a00327f8c0be18daf8a)

coresight: etm4x: moving sysFS entries to a dedicated file

As with the etm3x driver, sysFS entries are big enough to justify
their own file. As such moving all sysFS related declarations to
a dedicated location.

No gain/loss of functionality is incurred from this patch.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit a77de2637c9eb4794c6234b40cee2a243c548875)

stm class: Support devices that override software assigned masters

Some STM devices adjust software assigned master numbers depending on
the trace source and its runtime state and whatnot. This patch adds
a sysfs attribute to inform the trace-side software that master numbers
assigned to software sources will not match those in the STP stream,
so that, for example, master/channel allocation policy can be adjusted
accordingly.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 8e996a2874bbbed30e8dfe881453825fc6b7654e)

stm class: Remove unnecessary pointer increment

Readability: a postfix increment is used on a pointer which is not
used anywhere afterwards, which may send the reader looking through
the function one extra time. Drop the unnecessary increment.

Reported-by: Alan Cox <alan.cox@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reviewed-by: Laurent Fert <laurent.fert@intel.com>
(cherry picked from commit fb0801904bbbc7b109d4009520c7fa34bcfb7450)

stm class: Fix stm device initialization order

Currently, stm_register_device() makes the device visible and then
proceeds to initializing spinlocks and other properties, which leaves
a window when the device can already be opened but is not yet fully
operational.

Fix this by reversing the initialization order.

Reported-by: Alan Cox <alan.cox@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reviewed-by: Laurent Fert <laurent.fert@intel.com>
(cherry picked from commit 389b6699a2aa0b457aa69986e9ddf39f3b4030fd)

stm class: Do not leak the chrdev in error path

Currently, the error path of stm_register_device() forgets to unregister
the chrdev. Fix this.

Reported-by: Alan Cox <alan.cox@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reviewed-by: Laurent Fert <laurent.fert@intel.com>
(cherry picked from commit cbe4a61d1ddc4790d950ca8c33ef79ee68ef5e2b)

stm class: Remove a pointless line

No point in explicitly setting something to zero right after we
explicitly checked that it is zero. Fix this.

Reported-by: Alan Cox <alan.cox@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reviewed-by: Laurent Fert <laurent.fert@intel.com>
(cherry picked from commit 8fa11d1c1322f3de40a0e3f3f3e57436a204fcc4)

stm class: stm_heartbeat: Make nr_devs parameter read-only

Changing nr_devs after the module has been loaded doesn't actually
change anything, so just make it read-only.

Reported-by: Alan Cox <alan.cox@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reviewed-by: Laurent Fert <laurent.fert@intel.com>
(cherry picked from commit c8be4899449c0b27bc5daf71742cd601b2b3b4e3)

stm class: dummy_stm: Make nr_dummies parameter read-only

Changing nr_dummies after the module has been loaded doesn't actually
change anything, so just make it read-only.

Reported-by: Alan Cox <alan.cox@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reviewed-by: Laurent Fert <laurent.fert@intel.com>
(cherry picked from commit 118b4515aa6916ee7751f29c8b2a3af95abc9783)

MAINTAINERS: Add a git tree for the stm class

So that people know where their patches go.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Reviewed-by: Laurent Fert <laurent.fert@intel.com>
(cherry picked from commit e787bc463cc1fe3f51b0cd7bf540236318f69cf1)

perf/ring_buffer: Document AUX API usage

In order to ensure safe AUX buffer management, we rely on the assumption
that pmu::stop() stops its ongoing AUX transaction and not just the hw.

This patch documents this requirement for the perf_aux_output_{begin,end}()
APIs.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1457098969-21595-4-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit af5bb4ed1254a378b6028c09e58bdcc1cd9bf5b3)

perf/core: Free AUX pages in unmap path

Now that we can ensure that when ring buffer's AUX area is on the way
to getting unmapped new transactions won't start, we only need to stop
all events that can potentially be writing aux data to our ring buffer.

Having done that, we can safely free the AUX pages and corresponding
PMU data, as this time it is guaranteed to be the last aux reference
holder.

This partially reverts:

  57ffc5ca679 ("perf: Fix AUX buffer refcounting")

... which was made to defer deallocation that was otherwise possible
from an NMI context. Now it is no longer the case; the last call to
rb_free_aux() that drops the last AUX reference has to happen in
perf_mmap_close() on that AUX area.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/87d1qtz23d.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit 95ff4ca26c492fc1ed7751f5dd7ab7674b54f4e0)

perf/ring_buffer: Refuse to begin AUX transaction after rb->aux_mmap_count drops

When ring buffer's AUX area is unmapped and rb->aux_mmap_count drops to
zero, new AUX transactions into this buffer can still be started,
even though the buffer in en route to deallocation.

This patch adds a check to perf_aux_output_begin() for rb->aux_mmap_count
being zero, in which case there is no point starting new transactions,
in other words, the ring buffers that pass a certain point in
perf_mmap_close will not have their events sending new data, which
clears path for freeing those buffers' pages right there and then,
provided that no active transactions are holding the AUX reference.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1457098969-21595-2-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit dcb10a967ce82d5ad20570693091139ae716ff76)

perf auxtrace: Add perf_evlist pointer to *info_priv_size()

On some architecture the size of the private header may be dependent on
the number of tracers used in the session.  As such adding a "struct
perf_evlist *" parameter, which should contain all the required
information.

Also adjusting the existing client of the interface to take the new
parameter into account.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Al Grant <al.grant@arm.com>
Cc: Chunyan Zhang <zhang.chunyan@linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-doc@vger.kernel.org
Cc: Mike Leach <mike.leach@arm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Rabin Vincent <rabin@rab.in>
Cc: Tor Jeremiassen <tor@ti.com>
Link: http://lkml.kernel.org/r/1452807977-8069-22-git-send-email-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
(cherry picked from commit 14a05e13a044c1cd6aaa3eb1a5fcdad7b4f6c990)

perf session: Simplify tool stubs

Some of the stubs are identical so just have one function for them.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1457005856-6143-3-git-send-email-adrian.hunter@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit 5fb0ac16c5091f48eecf1a77e461f6957a463d61)

perf inject: Hit all DSOs for AUX data in JIT and other cases

Currently, when injecting build ids, if there is AUX data then 'perf
inject' hits all DSOs because it is not known which DSOs the trace data
would hit.

That needs to be done for JIT injection also, and in fact there is no
reason to distinguish what kind of injection is being done.  That is,
any time there is AUX data and the HEADER_BUID_ID feature flag is set,
and the AUX data is not being processed, then hit all DSOs.  This patch
does that.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1457005856-6143-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit 640dad47988ec4b734d71934be103bb6e931279f)

perf tools: tracepoint_error() can receive e=NULL, robustify it

Fixes segmentation fault using, for instance:

  (gdb) run record -I -e intel_pt/tsc=1,noretcomp=1/u /bin/ls
  Starting program: /home/acme/bin/perf record -I -e intel_pt/tsc=1,noretcomp=1/u /bin/ls
  Missing separate debuginfos, use: dnf debuginfo-install glibc-2.22-7.fc23.x86_64
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib64/libthread_db.so.1".

 Program received signal SIGSEGV, Segmentation fault.
  0 x00000000004b9ea5 in tracepoint_error (e=0x0, err=13, sys=0x19b1370 "sched", name=0x19a5d00 "sched_switch") at util/parse-events.c:410
  (gdb) bt
  #0  0x00000000004b9ea5 in tracepoint_error (e=0x0, err=13, sys=0x19b1370 "sched", name=0x19a5d00 "sched_switch") at util/parse-events.c:410
  #1  0x00000000004b9fc5 in add_tracepoint (list=0x19a5d20, idx=0x7fffffffb8c0, sys_name=0x19b1370 "sched", evt_name=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:433
  #2  0x00000000004ba334 in add_tracepoint_event (list=0x19a5d20, idx=0x7fffffffb8c0, sys_name=0x19b1370 "sched", evt_name=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:498
  #3  0x00000000004bb699 in parse_events_add_tracepoint (list=0x19a5d20, idx=0x7fffffffb8c0, sys=0x19b1370 "sched", event=0x19a5d00 "sched_switch", err=0x0, head_config=0x0)
      at util/parse-events.c:936
  #4  0x00000000004f6eda in parse_events_parse (_data=0x7fffffffb8b0, scanner=0x19a49d0) at util/parse-events.y:391
  #5  0x00000000004bc8e5 in parse_events__scanner (str=0x663ff2 "sched:sched_switch", data=0x7fffffffb8b0, start_token=258) at util/parse-events.c:1361
  #6  0x00000000004bca57 in parse_events (evlist=0x19a5220, str=0x663ff2 "sched:sched_switch", err=0x0) at util/parse-events.c:1401
  #7  0x0000000000518d5f in perf_evlist__can_select_event (evlist=0x19a3b90, str=0x663ff2 "sched:sched_switch") at util/record.c:253
  #8  0x0000000000553c42 in intel_pt_track_switches (evlist=0x19a3b90) at arch/x86/util/intel-pt.c:364
  #9  0x00000000005549d1 in intel_pt_recording_options (itr=0x19a2c40, evlist=0x19a3b90, opts=0x8edf68 <record+232>) at arch/x86/util/intel-pt.c:664
  #10 0x000000000051e076 in auxtrace_record__options (itr=0x19a2c40, evlist=0x19a3b90, opts=0x8edf68 <record+232>) at util/auxtrace.c:539
  #11 0x0000000000433368 in cmd_record (argc=1, argv=0x7fffffffde60, prefix=0x0) at builtin-record.c:1264
  #12 0x000000000049bec2 in run_builtin (p=0x8fa2a8 <commands+168>, argc=5, argv=0x7fffffffde60) at perf.c:390
  #13 0x000000000049c12a in handle_internal_command (argc=5, argv=0x7fffffffde60) at perf.c:451
  #14 0x000000000049c278 in run_argv (argcp=0x7fffffffdcbc, argv=0x7fffffffdcb0) at perf.c:495
  #15 0x000000000049c60a in main (argc=5, argv=0x7fffffffde60) at perf.c:618
(gdb)

Intel PT attempts to find the sched:sched_switch tracepoint but that seg
faults if tracefs is not readable, because the error reporting structure
is null, as errors are not reported when automatically adding
tracepoints.  Fix by checking before using.

Committer note:

This doesn't take place in a kernel that supports
perf_event_attr.context_switch, that is the default way that will be
used for tracking context switches, only in older kernels, like 4.2, in
a machine with Intel PT (e.g. Broadwell) for non-priviledged users.

Further info from a similar patch by Wang:

The error is in tracepoint_error: it assumes the 'e' parameter is valid.

However, there are many situation a parse_event() can be called without
parse_events_error. See result of

  $ grep 'parse_events(.*NULL)' ./tools/perf/ -r'

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Tong Zhang <ztong@vt.edu>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: stable@vger.kernel.org # v4.4+
Fixes: 196581717d85 ("perf tools: Enhance parsing events tracepoint error output")
Link: http://lkml.kernel.org/r/1453809921-24596-2-git-send-email-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
(cherry picked from commit ec183d22cc284a7a1e17f0341219d8ec8ca070cc)

perf evlist: Make perf_evlist__open() open evsels with their cpus and threads (like perf record does)

'perf record' uses perf_evsel__open() to open events and passes the
evsel->cpus and evsel->threads.  Many tests and some tools instead use
perf_evlist__open() which passes instead evlist->cpus and
evlist->threads.

Make perf_evlist__open() follow the 'perf record' behaviour so that a
consistent approach is taken.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Noel Grandin <noelgrandin@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1452158050-28061-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
(cherry picked from commit 23df7f798435796aff07d641456326b81cb34a77)

perf evsel: Introduce disable() method

Adding perf_evsel__disable function to have complement for
perf_evsel__enable function. Both will be used in following patch to
factor perf_evlist__(enable|disable).

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1449133606-14429-3-git-send-email-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
(cherry picked from commit e98a4cbb01e0ba1110eba5166a425b3eab9b2244)

perf cpumap: Auto initialize cpu__max_{node,cpu}

Since it was always checking if the initialization was done, use that
branch to do the initialization if not done already.

With this we reduce the number of exported globals from these files.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/20160125212955.GG22501@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
(cherry picked from commit 5ac76283b32b116c58e362e99542182ddcfc8262)

drivers/hwtracing: make coresight-etm-perf.c explicitly non-modular

In commit 941943cf519f7cacbbcecee5c4ef4b77b466bd5c ("drivers/hwtracing:
make coresight-* explicitly non-modular") we removed all uses of
modular functions/macros in favour of their built-in equivlents in
this subsystem.

However that commit and commit 0bcbf2e30ff2271b54f54c8697a185f7d86ec6e4
("coresight: etm-perf: new PMU driver for ETM tracers") were in flight
at the same time, and hence one new non-modular user of module_init
crept back in.  Fix it up like we did all the others.

Since module_init translates to device_initcall in the non-modular
case, the init ordering remains unchanged with this commit.

Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Acked-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit ca48fa22c3ed3b7b062bc6fa7b72493c00571e33)

drivers/hwtracing: make coresight-* explicitly non-modular

None of the Kconfig currently controlling compilation of any of
the files here are tristate, meaning that none of it currently
is being built as a module by anyone.

We need not be concerned about .remove functions and blocking the
unbind sysfs operations, since that was already done in a recent
commit.

Lets remove any remaining modular references, so that when reading the
drivers there is no doubt they are builtin-only.

All drivers get mostly the same changes, so they are handled in batch.
Changes are (1) convert to builtin_amba_driver, (2) delete module.h
include where unused, and (3) relocate the description into the
comments so we don't need MODULE_DESCRIPTION and associated tags.

The etm3x and etm4x use module_param_named, and have been adjusted
to just include moduleparam.h for that purpose.

In commit f309d4443130bf814e991f836e919dca22df37ae ("platform_device:
better support builtin boilerplate avoidance") we introduced the
builtin_driver macro.

Here we use that support and extend it to amba driver registration,
so where a driver is clearly non-modular and builtin-only, we can
update with the simple mapping of

     module_amba_driver(...)  ---> builtin_amba_driver(...)

Since module_amba_driver() uses the same init level priority as
builtin_amba_driver() the init ordering remains unchanged with
this commit.

Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 941943cf519f7cacbbcecee5c4ef4b77b466bd5c)

coresight: introducing a global trace ID function

TraceID values have to be unique for all tracers and
consistent between drivers and user space.  As such
introducing a central function to be used whenever a
traceID value is required.

The patch also account for data traceIDs, which are usually
I(N) + 1.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 17534ceb835a1a96eb921a2a80df168723d6570a)

coresight: etm-perf: new PMU driver for ETM tracers

Perf is a well known and used tool for performance monitoring
and much more. A such it is an ideal candidate for integration
with coresight based HW tracing.

This patch introduces a PMU that represent a coresight tracer to
the Perf core.

Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 0bcbf2e30ff2271b54f54c8697a185f7d86ec6e4)

coresight: etb10: implementing AUX API

Adding an ETB10 specific AUX area operations to be used
by the perf framework when events are initialised.

Part of this operation involves modeling the mmap'ed area
based on the specific ways a sink buffer gathers information.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 2997aa4063d97fdb39450c6078bd81a7b0504f22)

coresight: etb10: adding operation mode for sink->enable()

Adding an operation mode to the sink->enable() API in order
to prevent simultaneous access from different callers.

TPIU and TMC won't be supplemented with the AUX area
API immediately and as such ignore the new mode.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit e827d4550aa3225b8965ce4c266208cfe0297509)

coresight: etb10: moving to local atomic operations

Moving to use local atomic operations to take advantage of the
lockless implementation, something that will come handy when
the ETB is accessed from the Perf subsystem. Also changing the
name of the variable to something more meaningful.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 27b10da8fff27d74b755707e61637f6ab488c617)

coresight: etm3x: implementing perf_enable/disable() API

That way traces can be enabled and disabled automatically
from the Perf subystem using the PMU abstraction.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 882d5e112491c875ab7c8c336b8beaeec54d0509)

coresight: etm3x: implementing user/kernel mode tracing

Adding new mode to limit tracing to kernel or user space.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 2127154d115d4fe8f18300e5ef6f566581359d56)

coresight: etm3x: consolidating initial config

There is really no point in having two functions to take care
of doing the initial tracer configuration.  As such moving
everything to 'etm_set_default()'.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit c528a25ac7c4dacba9e4d98d5f06846939c5966f)

coresight: etm3x: changing default trace configuration

Changing default configuration to include the entire address
range rather than just the kernel.  That way traces are more
inclusive and it is easier to narrow down if needed.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit e19217299caf1a54c55081ab6339b3baccec63b0)

coresight: etm3x: set progbit to stop trace collection

There is no need to use the event enable's "always false" event to
stop trace collection.  For that purpose setting the programming bit
(ETMCR:10) is enough.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 47cd066cd00a65902ee3bd57da5bd395cb83aff9)

coresight: etm3x: adding operation mode for etm_enable()

Adding a new mode to source API enable() in order to
distinguish where the request comes from.  That way it is
possible to perform different operations based on where
the request was issued from.

The ETM4x driver is also modified to keep in sync with the
new interface.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 22fd532eaa0c24d86e23d8e9e3b7feac4a8cac80)

coresight: etm3x: splitting struct etm_drvdata

Splitting "etm_drvdata" in two sections, one for the HW specific
data and another for user configuration.

That way it is easier to manipulate and zero out the configuration
data when more than one concurrent tracing session configuration
is active.

Also taking care of up-lifting all the code affected by this new
arrangement.  No loss or gain of functionality (other than what is
mentioned above) is introduced by this patch.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 1925a470ce69cdfa2b82ac1565d58dfd39cd877d)

coresight: etm3x: unlocking tracers in default arch init

Calling function 'smp_call_function_single()' to unlock a
tracer and calling it again right after to perform the
default initialisation doesn't make sense.

Moving 'etm_os_unlock()' just before making the default
initialisation results in the same outcome while saving
one call to 'smp_call_function_single()'.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit ae69a1da399fccaed406932f5cbee55a6f9d4425)

coresight: etm3x: moving sysFS entries to dedicated file

SysFS entries are big enough to justify their own file.
As such moving all sysFS related declarations to a dedicated
location.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit c04148e708c0d8d7eabc447946b712a66b468b47)

coresight: etm3x: moving etm_readl/writel to header file

Moving functions etm_readl/writel to file "coresight-etm.h"
so that the main ETM3x driver can be split in more than one
file.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit c1f8e57c9e6692f6e8c6c1f9eab7a46264ac4245)

coresight: moving PM runtime operations to core framework

Moving PM runtime operations in Coresight devices enable() and
disable() API to the framework core when a path is setup.  That
way the runtime core doesn't have to be involved everytime a
path is enabled.  It also avoids calling runtime PM operations
in IRQ context.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 5da5325fa85658ee793792b5285dd5fdb76ccfb7)

coresight: add API to get sink from path

Add an API allowing external code to quickly get a handle on the
sink within a path.  The sink is always last, but adding an API allows
to keep the path's node structure private and remove redundant checks.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit b6404e21f023e4aa208a0ba03d55a9c8a57cb940)

coresight: associating path with session rather than tracer

When using the Coresight framework from the sysFS interface a
tracer is always handling a single session and as such, a path
can be associated with a tracer.  But when supporting multiple
session per tracer there is no guarantee that sessions will always
have the same path from source to sink.

This patch is removing the automatic association between path and
tracers.  The building of a path and enablement of the components
in the path are decoupled, allowing for the association of a path
with a session rather than a tracer.

To keep backward functionality with the current sysFS access methods
a per-cpu place holder is used to keep a handle on the path built when
tracers are enabled.  Lastly APIs to build paths and enable tracers are
made public so that other subsystem can interact with the Coresight
framework.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit b3e94405941e6916d5e365454d74560c2bea47ca)

coresight: etm4x: Check every parameter used by dma_xx_coherent.

The dma_alloc_coherent return an "void *" not an "void __iomen *".
It uses the wrong parameters when calls dma_free_coherent function.

The sparse tool output logs as the following:
coresight-tmc.c:199:23:    expected void *<noident>
coresight-tmc.c:199:23:    got void [noderef] <asn:2>*vaddr
coresight-tmc.c:336:30: warning: incorrect type in assignment
(different address spaces)
coresight-tmc.c:336:30:    expected char *buf
coresight-tmc.c:336:30:    got void [noderef] <asn:2>*
coresight-tmc.c:769:50: warning: incorrect type in argument 4
(different base types)
coresight-tmc.c:769:50:    expected unsigned long long
[unsigned] [usertype] dma_handle
coresight-tmc.c:769:50:    got restricted gfp_t

Signed-off-by: Eric Long <eric.long@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 61390593f72377c3a8f41ef998462e2d3985adac)

coresight: "DEVICE_ATTR_RO" should defined as static.

"DEVICE_ATTR_RO(name)" should be defined as static. And
there is an unnecessary space at the front of the code.

The sparse tool output logs as the following:
coresight-etm4x.c:2224:1: warning: symbol 'dev_attr_trcoslsr' was
not declared. Should it be static?
coresight-etm4x.c:2225:1: warning: symbol 'dev_attr_trcpdcr' was
not declared. Should it be static?
coresight-etm4x.c:2226:1: warning: symbol 'dev_attr_trcpdsr' was
not declared. Should it be static?
And the smatch tool output logs as the following:
of_coresight.c:89 of_coresight_alloc_memory() warn:
inconsistent indenting

Signed-off-by: Eric Long <eric.long@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit bf16e5b8cdeabc1fe6565af0be475bb2084dc388)

coresight: implementing 'cpu_id()' API

Other than plainly parsing the device tree there is no way to
know which CPU a tracer is affined to.  As such adding an
interface to lookup the CPU field enclosed in the etm_drvdata
structure that was initialised at boot time.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 52210c8745e418f82f3f0aeeee01d7bc4858812a)

coresight: removing bind/unbind options from sysfs

The coresight drivers have absolutely no control over bind and unbind
operations triggered from sysfs. The operations simply can't be
cancelled or denied event when one or several tracing sessions are
under way.  Since the memory associated to individual device is
invariably freed, the end result is a kernel crash when the path from
source to sink is travelled again as demonstrated here[1].

One solution could be to keep track of all the path (i.e tracing
session) that get created and iterate through the elements of those path
looking for the coresight device that is being removed.  This proposition
doesn't scale well since there is no upper bound on the amount of
concurrent trace session that can be created.

With the above in mind, this patch prevent devices from being unbounded
from their driver by using the driver->suppress_bind_attr option.  That way
trace sessions can be managed without fearing to loose devices.

Since device can't be removed anymore the xyz_remove() functions found in
each driver is also removed.

[1]. http://www.spinics.net/lists/arm-kernel/msg474952.html

Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit b15f0fb657e040401d875d11ae13b269af8a16e0)

coresight: remove csdev's link from topology

In function 'coresight_unregister()', all references to the csdev that
is being taken away need to be removed from the topology.  Otherwise
building the next coresight path from source to sink may use memory
that has been released.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit ad725aee070caf8fa93d84d6fb78321f9642db18)

coresight: release reference taken by 'bus_find_device()'

The reference count taken by function bus_find_device() needs
to be released if a child device is found, something this patch
is adding.

Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit f2dfab3568fc32afeac8b698481e80e7ab2dc658)

coresight: coresight_unregister() function cleanup

In its current form the code never frees csdev->refcnt allocated
in coresight_register().  There is also a problem with csdev->conns
that is freed before device_unregister() rather than in the device
release function.

This patch addresses both issues by moving kfree(csdev->conns) to
coresight_device_release() and freeing csdev->refcnt, also in
the same function.

Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit fae54158792aec705620bdc3938d342879204f0c)

coresight: fixing lockdep error

On some platform the following lockdep error occurs when doing simple
manipulations:

    [   23.197021]
    [   23.198608] ======================================================
    [   23.205078] [ INFO: possible circular locking dependency detected ]
    [   23.211639] 4.4.0-rc8-00025-gbbf360b #172 Not tainted
    [   23.216918] -------------------------------------------------------
    [   23.223480] sh/858 is trying to acquire lock:
    [   23.228057]  (coresight_mutex){+.+.+.}, at: [<c0415d40>] coresight_enable+0x1c/0x1b4
    [   23.236206]
    [   23.236206] but task is already holding lock:
    [   23.242309]  (s_active#52){++++.+}, at: [<c01d4b40>] kernfs_fop_write+0x5c/0x1c0
    [   23.250122]
    [   23.250122] which lock already depends on the new lock.
    [   23.250122]
    [   23.258697]
    [   23.258697] the existing dependency chain (in reverse order) is:
    [   23.266510]
    -> #1 (s_active#52){++++.+}:
    [   23.270843]        [<c01d30ec>] __kernfs_remove+0x294/0x35c
    [   23.276672]        [<c01d3e44>] kernfs_remove_by_name_ns+0x44/0x8c
    [   23.283172]        [<c01d6318>] remove_files+0x3c/0x84
    [   23.288543]        [<c01d66b4>] sysfs_remove_group+0x48/0x9c
    [   23.294494]        [<c01d6734>] sysfs_remove_groups+0x2c/0x3c
    [   23.300506]        [<c030b658>] device_remove_attrs+0x5c/0x74
    [   23.306549]        [<c030c290>] device_del+0x110/0x218
    [   23.311950]        [<c030c3c4>] device_unregister+0x2c/0x6c
    [   23.317779]        [<c04156d8>] coresight_unregister+0x30/0x40
    [   23.323883]        [<c041a290>] etm_probe+0x228/0x2e8
    [   23.329193]        [<c02bc760>] amba_probe+0xe4/0x160
    [   23.334503]        [<c0310540>] driver_probe_device+0x23c/0x480
    [   23.340728]        [<c0310820>] __driver_attach+0x9c/0xa0
    [   23.346374]        [<c030e400>] bus_for_each_dev+0x70/0xa4
    [   23.352142]        [<c030fcf4>] driver_attach+0x24/0x28
    [   23.357604]        [<c030f86c>] bus_add_driver+0x1e0/0x278
    [   23.363372]        [<c0310d48>] driver_register+0x80/0x100
    [   23.369110]        [<c02bc508>] amba_driver_register+0x58/0x5c
    [   23.375244]        [<c0749514>] etm_driver_init+0x18/0x1c
    [   23.380889]        [<c0009918>] do_one_initcall+0xc4/0x20c
    [   23.386657]        [<c0715e7c>] kernel_init_freeable+0x160/0x208
    [   23.392974]        [<c052d7fc>] kernel_init+0x18/0xf0
    [   23.398254]        [<c0010850>] ret_from_fork+0x14/0x24
    [   23.403747]
    -> #0 (coresight_mutex){+.+.+.}:
    [   23.408447]        [<c008ed60>] lock_acquire+0xe4/0x210
    [   23.413909]        [<c0530a30>] mutex_lock_nested+0x74/0x450
    [   23.419860]        [<c0415d40>] coresight_enable+0x1c/0x1b4
    [   23.425689]        [<c0416030>] enable_source_store+0x58/0x68
    [   23.431732]        [<c030b358>] dev_attr_store+0x20/0x2c
    [   23.437286]        [<c01d55e8>] sysfs_kf_write+0x50/0x54
    [   23.442871]        [<c01d4ba8>] kernfs_fop_write+0xc4/0x1c0
    [   23.448699]        [<c015b60c>] __vfs_write+0x34/0xe4
    [   23.454040]        [<c015bf38>] vfs_write+0x98/0x174
    [   23.459228]        [<c015c7a8>] SyS_write+0x4c/0xa8
    [   23.464355]        [<c00107c0>] ret_fast_syscall+0x0/0x1c
    [   23.470031]
    [   23.470031] other info that might help us debug this:
    [   23.470031]
    [   23.478393]  Possible unsafe locking scenario:
    [   23.478393]
    [   23.484619]        CPU0                    CPU1
    [   23.489349]        ----                    ----
    [   23.494079]   lock(s_active#52);
    [   23.497497]                                lock(coresight_mutex);
    [   23.503906]                                lock(s_active#52);
    [   23.509918]   lock(coresight_mutex);
    [   23.513702]
    [   23.513702]  *** DEADLOCK ***
    [   23.513702]
    [   23.519897] 3 locks held by sh/858:
    [   23.523529]  #0:  (sb_writers#7){.+.+.+}, at: [<c015ec38>] __sb_start_write+0xa8/0xd4
    [   23.531799]  #1:  (&of->mutex){+.+...}, at: [<c01d4b38>] kernfs_fop_write+0x54/0x1c0
    [   23.539916]  #2:  (s_active#52){++++.+}, at: [<c01d4b40>] kernfs_fop_write+0x5c/0x1c0
    [   23.548156]
    [   23.548156] stack backtrace:
    [   23.552734] CPU: 0 PID: 858 Comm: sh Not tainted 4.4.0-rc8-00025-gbbf360b #172
    [   23.560302] Hardware name: Generic OMAP4 (Flattened Device Tree)
    [   23.566589] Backtrace:
    [   23.569152] [<c00154d4>] (dump_backtrace) from [<c00156d0>] (show_stack+0x18/0x1c)
    [   23.577087]  r7:ed4b8570 r6:c0936400 r5:c07ae71c r4:00000000
    [   23.583038] [<c00156b8>] (show_stack) from [<c027e69c>] (dump_stack+0x98/0xc0)
    [   23.590606] [<c027e604>] (dump_stack) from [<c008a750>] (print_circular_bug+0x21c/0x33c)
    [   23.599090]  r5:c0939d60 r4:c0936400
    [   23.602874] [<c008a534>] (print_circular_bug) from [<c008e370>] (__lock_acquire+0x1c98/0x1d88)
    [   23.611877]  r10:00000003 r9:c0fd7a5c r8:ed4b8550 r7:ed4b8570 r6:ed4b8000 r5:c0ff69e4
    [   23.620117]  r4:c0936400 r3:ed4b8550
    [   23.623901] [<c008c6d8>] (__lock_acquire) from [<c008ed60>] (lock_acquire+0xe4/0x210)
    [   23.632080]  r10:00000000 r9:00000000 r8:60000013 r7:c07cb7b4 r6:00000001 r5:00000000
    [   23.640350]  r4:00000000
    [   23.643005] [<c008ec7c>] (lock_acquire) from [<c0530a30>] (mutex_lock_nested+0x74/0x450)
    [   23.651458]  r10:ecc0bf80 r9:edbe7dcc r8:ed4b8000 r7:c0fd7a5c r6:c0415d40 r5:00000000
    [   23.659729]  r4:c07cb780
    [   23.662384] [<c05309bc>] (mutex_lock_nested) from [<c0415d40>] (coresight_enable+0x1c/0x1b4)
    [   23.671234]  r10:ecc0bf80 r9:edbe7dcc r8:ed733c00 r7:00000000 r6:ed733c00 r5:00000002
    [   23.679473]  r4:ed762140
    [   23.682128] [<c0415d24>] (coresight_enable) from [<c0416030>] (enable_source_store+0x58/0x68)
    [   23.691070]  r7:00000000 r6:ed733c00 r5:00000002 r4:ed762160
    [   23.697052] [<c0415fd8>] (enable_source_store) from [<c030b358>] (dev_attr_store+0x20/0x2c)
    [   23.705780]  r5:edbe7dc0 r4:c0415fd8
    [   23.709533] [<c030b338>] (dev_attr_store) from [<c01d55e8>] (sysfs_kf_write+0x50/0x54)
    [   23.717834]  r5:edbe7dc0 r4:c030b338
    [   23.721618] [<c01d5598>] (sysfs_kf_write) from [<c01d4ba8>] (kernfs_fop_write+0xc4/0x1c0)
    [   23.730163]  r7:00000000 r6:00000000 r5:00000002 r4:edbe7dc0
    [   23.736145] [<c01d4ae4>] (kernfs_fop_write) from [<c015b60c>] (__vfs_write+0x34/0xe4)
    [   23.744323]  r10:00000000 r9:ecc0a000 r8:c0010964 r7:ecc0bf80 r6:00000002 r5:c01d4ae4
    [   23.752593]  r4:ee385a40
    [   23.755249] [<c015b5d8>] (__vfs_write) from [<c015bf38>] (vfs_write+0x98/0x174)
    [   23.762908]  r9:ecc0a000 r8:c0010964 r7:ecc0bf80 r6:000ab0d8 r5:00000002 r4:ee385a40
    [   23.771057] [<c015bea0>] (vfs_write) from [<c015c7a8>] (SyS_write+0x4c/0xa8)
    [   23.778442]  r8:c0010964 r7:00000002 r6:000ab0d8 r5:ee385a40 r4:ee385a40
    [   23.785522] [<c015c75c>] (SyS_write) from [<c00107c0>] (ret_fast_syscall+0x0/0x1c)
    [   23.793457]  r7:00000004 r6:00000001 r5:000ab0d8 r4:00000002
    [   23.799652] coresight-etb10 54162000.etb: ETB enabled
    [   23.805084] coresight-funnel 54164000.funnel: FUNNEL inport 0 enabled
    [   23.811859] coresight-replicator 44000000.ocp:replicator: REPLICATOR enabled
    [   23.819335] coresight-funnel 54158000.funnel: FUNNEL inport 0 enabled
    [   23.826110] coresight-etm3x 5414c000.ptm: ETM tracing enabled

The locking in coresight_unregister() is not required as the only customers of
the function are drivers themselves when an initialisation failure has been
encoutered.

Reported-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit a9ddc71f5840c2711e530f2e055b278f79948b29)

coresight: fixing indentation problem

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit ff8742279ff163e36ee50bb4e75090af1d7d6e3b)

coresight: Fix a typo in Kconfig

Signed-off-by: Andrew F. Davis <afd@ti.com>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit b9884d3b79f60846cdf04be262d13bca8996f99a)

coresight: checking for NULL string in coresight_name_match()

Connection child names associated to ports can sometimes be NULL,
which is the case when booting a system on QEMU or when the Coresight
power domain isn't switched on.

This patch is adding a check to make sure a NULL string isn't fed
to strcmp(), something that avoid crashing the system.

Cc: <stable@vger.kernel.org> # v3.18+
Reported-by: Tyler Baker <tyler.baker@linaro.org>
Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit fadf3a44e974b030e7145218ad1ab25e3ef91738)

perf/core: Disable the event on a truncated AUX record

When the PMU driver reports a truncated AUX record, it effectively means
that there is no more usable room in the event's AUX buffer (even though
there may still be some room, so that perf_aux_output_begin() doesn't take
action). At this point the consumer still has to be woken up and the event
has to be disabled, otherwise the event will just keep spinning between
perf_aux_output_begin() and perf_aux_output_end() until its context gets
unscheduled.

Again, for cpu-wide events this means never, so once in this condition,
they will be forever losing data.

Fix this by disabling the event and waking up the consumer in case of a
truncated AUX record.

Reported-by: Markus Metzger <markus.t.metzger@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1462886313-13660-3-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit 9f448cd3cbcec8995935e60b27802ae56aac8cc0)

perf/core: Don't leak event in the syscall error path

In the error path, event_file not being NULL is used to determine
whether the event itself still needs to be free'd, so fix it up to
avoid leaking.

Reported-by: Leon Yu <chianglungyu@gmail.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Fixes: 130056275ade ("perf: Do not double free")
Link: http://lkml.kernel.org/r/87twk06yxp.fsf@ashishki-desk.ger.corp.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit 201c2f85bd0bc13b712d9c0b3d11251b182e06ae)

perf/core: Fix perf_sched_count derailment

The error path in perf_event_open() is such that asking for a sampling
event on a PMU that doesn't generate interrupts will end up in dropping
the perf_sched_count even though it hasn't been incremented for this
event yet.

Given a sufficient amount of these calls, we'll end up disabling
scheduler's jump label even though we'd still have active events in the
system, thereby facilitating the arrival of the infernal regions upon us.

I'm fixing this by moving account_event() inside perf_event_alloc().

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1456917854-29427-1-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit 927a5570855836e5d5859a80ce7e91e963545e8f)

stm class: dummy_stm: Add link callback for fault injection

STM device's link callback has the power to abort master/channel
assignment by returning a negative error code. Use this in dummy
stm device to optionally abort assigning certain channel IDs.
This is useful as fault injection into the stm class core, for
testing purposes.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit adcde635f5d6b77e9f11087af09c6f62da2bf108)

stm class: Plug stm device's unlink callback

STM device's unlink callback is never actually called from anywhere in
the stm class code.

This patch adds calls to stm driver's unlink method after the unlinking
has succeeded.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit cc8424074e51355e0c6ba717d8edc50d408f2802)

stm class: Fix a race in unlinking

There is a window in stm_source_link_drop(), during which the source's
link may change before locks are acquired. When this happens, it throws
a warning, since this is not an expected scenario.

This patch handles the race in such a way that if the link appears to
have changed by the time we took the locks, it will release them and
repeat the whole unlinking procedure from the beginning, unless the
other contender beat us to it.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit b4ca34aaf78ed0cdfc15956d377064104257a437)

stm class: Fix unbalanced module/device refcounting

STM code takes references to the stm device and its module for the
duration of the character device's existence or the stm_source link.
Dropping these references is not well balanced everywhere, which may
lead to leaks.

This patch balances the acquisition and releasing of these two
references and annotates each site so that it's easier to verify
correctness by reading the code.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit f7c81c7176c72c7899390754b4b038a64b296e4d)

stm class: Guard output assignment against concurrency

It is possible to concurrently assign the same output (a character
device writer or an stm_source device) to different stm devices,
which sets off a strategically placed warning in stm_output_assign().

To avoid this, use a spinlock to serialize (un)assignments between
outputs and stm devices.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit cde4ad8368840e414ecf67db258fe1dabaa5fd2e)

stm class: Fix unlocking braino in the error path

If an illegal attempt is made to unlink stm source device from an
stm device, the stm device's link spinlock mistakenly remains locked.
While this really shouldn't happen (there's a warning in place), the
locking should remain in order so that we can still recover from this
situation if it indeed does happen.

This patch unifies the unlocking in the exit path of
__stm_source_link_drop() to fix this.

Reported-by: Laurent Fert <laurent.fert@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 1810f2c44817c74ca3d05d1e3981e3a2e2ceb6f5)

stm class: Add heartbeat stm source device

Heartbeat stm source may have multiple instances (for connecting to
different stm devices). Each instance will send a periodic test message
over its stm device when it is linked. This can be used for testing stm
class framework, stm device drivers or as a heartbeat over the stm link.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 1192918530381b5cfc0e5da51233fa94f783b221)

stm class: dummy_stm: Create multiple devices

STM framework should be able to handle multiple STM devices at a time,
each one with its own master allocation policy.

This patch changes dummy_stm driver to create multiple STM sinks to
help testing the framework.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit bcfdf8afdebe63a2217fa632ae94f8aeecf9126f)

stm class: Support devices with multiple instances

By convention, the name of the stm policy directory in configfs consists of
the device name to which it applies and the actual policy name, separated
by a dot. Now, some devices already have dots in their names that separate
name of the actual device from its instance identifier. Such devices will
result in two (or more, who can tell) dots in the policy directory name.

Existing policy code, however, will treat the first dot as the one that
separates device name from policy name, therefore failing the above case.

This patch makes the last dot in the directory name be the separator, thus
prohibiting dots from being used in policy names.

Suggested-by: Chunyan Zhang <zhang.chunyan@linaro.org>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 59be422e4ce10e3d49d4c9407a80fab8a9b7bc84)

stm class: Use driver's packet callback return value

STM drivers provide a callback to generate/send individual STP packets;
it also tells the stm core how many bytes of payload it has consumed.
However, we would also need to use the negative space of this return
value to communicate errors that occur during the packet generation,
in which case the stm core will have to take appropriate action.

For now, we need to account for the possibility that the stm driver may
not support certain combinations of packet type/flags, in which case
it is expected to signal an error.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit f8560a9bc76b2cd5c06fa412cb7b5481d70fcf34)

stm class: Prevent user-controllable allocations

Currently, the character device write method allocates a temporary buffer
for user's data, but the user's data size is not sanitized and can cause
arbitrarily large allocations via kzalloc() or an integer overflow that
will then result in overwriting kernel memory.

This patch trims the input buffer size to avoid these issues.

Reported-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit f08b18266c7116e2ec6885dd53a928f580060a71)

stm class: Fix link list locking

Currently, the list of stm_sources linked to an stm device is protected by
a spinlock, which also means that sources' .unlink() method is called under
this spinlock. However, this method may (and does) sleep, which means
trouble.

This patch slightly reworks locking around stm::link_list so that bits that
might_sleep() are called with a mutex held instead. Modification of this
list requires both mutex and spinlock to be held, while looking at the list
can be done under either mutex or spinlock.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit c74f7e8281add80bdfa0ad2998b8df287b13df73)

stm class: Fix locking in unbinding policy path

Right now, if stm device removal has to unbind from a policy (that is,
an stm device that has STP policy, gets removed), it will trigger a
nested lock on the stm device's policy mutex.

This patch fixes the problem by moving the locking from the policy
unbinding to policy removal (configfs path), where it's actually needed;
the other caller of the policy unbinding function already takes the
mutex around the call.

Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 4c127fd16e6b33ecb7badc091480c84ea9aebeb6)

stm class: Select CONFIG_SRCU

The newly added STM code uses SRCU, but does not ensure that
this code is part of the kernel:

drivers/built-in.o: In function `stm_source_link_show':
include/linux/srcu.h:221: undefined reference to `__srcu_read_lock'
include/linux/srcu.h:238: undefined reference to `__srcu_read_unlock'
drivers/built-in.o: In function `stm_source_link_drop':
include/linux/srcu.h:221: undefined reference to `__srcu_read_lock'
include/linux/srcu.h:238: undefined reference to `__srcu_read_unlock'

This adds a Kconfig 'select' statement like all the other SRCU using
drivers have.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Fixes: 7bd1d4093c2f ("stm class: Introduce an abstraction for System Trace Module devices")
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 042d4460b5b4379a12f375045ff9065cf6758735)

stm class: Hide STM-specific options if STM is disabled

If STM=n, it doesn't make sense to ask about STM_DUMMY and
STM_SOURCE_CONSOLE support, which are not even built when enabled
anyway. Hence hide these options if STM=n.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit 4a2e2b19f96acfc037a9773c7729d133ce1e7e3b)

perf: Synchronously free aux pages in case of allocation failure

We are currently using asynchronous deallocation in the error path in
AUX mmap code, which is unnecessary and also presents a problem for users
that wish to probe for the biggest possible buffer size they can get:
they'll get -EINVAL on all subsequent attemts to allocate a smaller
buffer before the asynchronous deallocation callback frees up the pages
from the previous unsuccessful attempt.

Currently, gdb does that for allocating AUX buffers for Intel PT traces.
More specifically, overwrite mode of AUX pmus that don't support hardware
sg (some implementations of Intel PT, for instance) is limited to only
one contiguous high order allocation for its buffer and there is no way
of knowing its size without trying.

This patch changes error path freeing to be synchronous as there won't
be any contenders for the AUX pages at that point.

Reported-by: Markus Metzger <markus.t.metzger@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1453216469-9509-1-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
(cherry picked from commit 45c815f06b80031659c63d7b93e580015d6024dd)

Linux 4.4

Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fix from James Bottomley:
 "A single fix for machines with pages > 4k (PPC mostly).

  There's a bug in our optimal transfer size code where we don't account
  for pages > 4k and can set the transfer size to be less than the page
  size causing nasty failures"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  sd: Reject optimal transfer length smaller than page size

Merge tag 'pci-v4.4-fixes-4' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fixlet from Bjorn Helgaas:
 "This marks the TI DRA7xx host bridge driver as broken.  Apparently it
  has never worked without some additional out-of-tree code, so I'm
  going to mark it broken now and remove it completely next cycle unless
  it's fixed"

* tag 'pci-v4.4-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI: dra7xx: Mark driver as broken

vmstat: allocate vmstat_wq before it is used

kernel test robot has reported the following crash:

  BUG: unable to handle kernel NULL pointer dereference at 00000100
  IP: [<c1074df6>] __queue_work+0x26/0x390
  *pdpt = 0000000000000000 *pde = f000ff53f000ff53 *pde = f000ff53f000ff53
  Oops: 0000 [#1] PREEMPT PREEMPT SMP SMP
  CPU: 0 PID: 24 Comm: kworker/0:1 Not tainted 4.4.0-rc4-00139-g373ccbe #1
  Workqueue: events vmstat_shepherd
  task: cb684600 ti: cb7ba000 task.ti: cb7ba000
  EIP: 0060:[<c1074df6>] EFLAGS: 00010046 CPU: 0
  EIP is at __queue_work+0x26/0x390
  EAX: 00000046 EBX: cbb37800 ECX: cbb37800 EDX: 00000000
  ESI: 00000000 EDI: 00000000 EBP: cb7bbe68 ESP: cb7bbe38
   DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
  CR0: 8005003b CR2: 00000100 CR3: 01fd5000 CR4: 000006b0
  Stack:
  Call Trace:
    __queue_delayed_work+0xa1/0x160
    queue_delayed_work_on+0x36/0x60
    vmstat_shepherd+0xad/0xf0
    process_one_work+0x1aa/0x4c0
    worker_thread+0x41/0x440
    kthread+0xb0/0xd0
    ret_from_kernel_thread+0x21/0x40

The reason is that start_shepherd_timer schedules the shepherd work item
which uses vmstat_wq (vmstat_shepherd) before setup_vmstat allocates
that workqueue so if the further initialization takes more than HZ we
might end up scheduling on a NULL vmstat_wq.  This is really unlikely
but not impossible.

Fixes: 373ccbe59270 ("mm, vmstat: allow WQ concurrency to discover memory reclaim doesn't make any progress")
Reported-by: kernel test robot <ying.huang@linux.intel.com>
Signed-off-by: Michal Hocko <mhocko@suse.com>
Tested-by: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: stable@vger.kernel.org
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge tag 'fixes-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Arnd Bergmann:
 "This is the final small set of ARM SoC bug fixes for linux-4.4, almost
  all regressions:

  OMAP:
   - data corruption on the Nokia N900 flash

  Allwinner:
   - Two defconfig change to get USB working again

  ARM Versatile:
   - Interrupt numbers gone bad after an older bug fix

  Nomadik:
   - Crashes from incorrect L2 cache settings

  VIA vt8500:
   - SD/MMC support on WM8650 never worked"

* tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  dts: vt8500: Add SDHC node to DTS file for WM8650
  ARM: Fix broken USB support in multi_v7_defconfig for sunxi devices
  ARM: versatile: fix MMC/SD interrupt assignment
  ARM: nomadik: set latencies to 8 cycles
  ARM: OMAP2+: Fix onenand rate detection to avoid filesystem corruption
  ARM: Fix broken USB support in sunxi_defconfig

Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fix from Paolo Bonzini:
 "A simple fix.  I'm sending it before the merge window, because it
  refines a patch found in your master branch but not yet in the
  kvm/next branch that is destined for 4.5"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  kvm: x86: only channel 0 of the i8254 is linked to the HPET

Merge tag 'pm+acpi-4.4-final' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI fix from Rafael Wysocki:
 "Just one obvious fix that adds a missing function argument in ACPI
  code introduced recently (Kees Cook)"

* tag 'pm+acpi-4.4-final' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI / property: avoid leaking format string into kobject name

Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:
 "A handful of x86 fixes:

   - a syscall ABI fix, fixing an Android breakage
   - a Xen PV guest fix relating to the RTC device, causing a
     non-working console
   - a Xen guest syscall stack frame fix
   - an MCE hotplug CPU crash fix"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/numachip: Fix NumaConnect2 MMCFG PCI access
  x86/entry: Restore traditional SYSENTER calling convention
  x86/entry: Fix some comments
  x86/paravirt: Prevent rtc_cmos platform device init on PV guests
  x86/xen: Avoid fast syscall path for Xen PV guests
  x86/mce: Ensure offline CPUs don't participate in rendezvous process

Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fixes from Ingo Molnar:
 "Misc scheduler fixes"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/core: Reset task's lockless wake-queues on fork()
  sched/core: Fix unserialized r-m-w scribbling stuff
  sched/core: Check tgid in is_global_init()
  sched/fair: Fix multiplication overflow on 32-bit systems

Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "Two core subsystem fixes, plus a handful of tooling fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf: Fix race in swevent hash
  perf: Fix race in perf_event_exec()
  perf list: Robustify event printing routine
  perf list: Add support for PERF_COUNT_SW_BPF_OUT
  perf hists browser: Fix segfault if use symbol filter in cmdline
  perf hists browser: Reset selection when refresh
  perf hists browser: Add NULL pointer check to prevent crash
  perf buildid-list: Fix return value of perf buildid-list -k
  perf buildid-list: Show running kernel build id fix

Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq fix from Ingo Molnar:
 "Fixes a core IRQ subsystem deadlock"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Prevent chip buslock deadlock

Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block revert from Jens Axboe:
 "The previous pull request had a split fix for NVMe, however there are
  corner cases where that ends up blowing up.

  So let's revert it for 4.4.  The regression isn't introduced in this
  cycle, and it's "just" a performance regression, not a
  stability/integrity issue"

* 'for-linus' of git://git.kernel.dk/linux-block:
  Revert "block: Split bios on chunk boundaries"

Merge tag 'dmaengine-fix-4.4' of git://git.infradead.org/users/vkoul/slave-dma

Pull dmaengine fixes from Vinod Koul:
 "Late fixes for 4.4 are three fixes for drivers which include a revert
  of mic-x100 fix which is causing regression, xgene fix for double IRQ
  and async_tx fix to use GFP_NOWAIT"

* tag 'dmaengine-fix-4.4' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: xgene-dma: Fix double IRQ issue by setting IRQ_DISABLE_UNLAZY flag
  async_tx: use GFP_NOWAIT rather than GFP_IO
  dmaengine: Revert "dmaengine: mic_x100: add missing spin_unlock"

Merge branch 'dmi-for-linus' of git://git./linux/kernel/git/jdelvare/staging

Pull dmi fix from Jean Delvare.

* 'dmi-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jdelvare/staging:
  firmware: dmi_scan: Fix UUID endianness for SMBIOS >= 2.6

Merge tag 'sound-4.4' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A slightly higher volume than a new year's wish, but not too
  worrisome: a large LOC is only for HD-audio device-specific quirks, so
  fairly safe to apply.  The rest ASoC fixes are all trivial and small;
  a simple replacement of mutex call with nested lock version, a few
  Arizona and Realtek codec fixes, and a regression fix for Skylake
  firmware handling"

* tag 'sound-4.4' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ASoC: Intel: Skylake: Fix the memory leak
  ASoC: Intel: Skylake: Revert previous broken fix memory leak fix
  ASoC: Use nested lock for snd_soc_dapm_mutex_lock
  ASoC: rt5645: add sys clk detection
  ALSA: hda - Add keycode map for alc input device
  ALSA: hda - Add mic mute hotkey quirk for Lenovo ThinkCentre AIO
  ASoC: arizona: Fix bclk for sample rates that are multiple of 4kHz

Merge tag 'omap-for-v4.4/onenand-corruption' of git://git./linux/kernel/git/tmlind/linux-omap into fixes

Pull "urgent onenand file system corruption fix for n900" from Tony Lindgren:

Last minute urgent pull request to prevent file system corruption
on Nokia N900.

Looks like we have a GPMC bus timing bug that has gone unnoticed
because of bootloader configured registers until few days ago. We
are not detecting the onenand clock rate properly unless we have
CONFIG_OMAP_GPMC_DEBUG set and this causes onenand corruption
that can be easily be reproduced.

There seems to be also an additional bug still lurking around for
onenand corruption. But that is still being investigated and
it does not seem to be GPMC timings related.

Meanwhile, it would be good to get this fix into v4.4 to prevent
wrong timings from corrupting onenand.

* tag 'omap-for-v4.4/onenand-corruption' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap:
  ARM: OMAP2+: Fix onenand rate detection to avoid filesystem corruption