uprobes: Shift ->readpage check from __copy_insn() to uprobe_register()

copy_insn() fails with -EIO if ->readpage == NULL, but this error
is not propagated unless uprobe_register() path finds ->mm which
already mmaps this file. In this case (say) "perf record" does not
actually install the probe, but the user can't know about this.

Move this check into uprobe_register() so that this problem can be
detected earlier and reported to user.

Note: this is still not perfect,

- copy_insn() and arch_uprobe_analyze_insn() should be called
  by uprobe_register() but this is not simple, we need vm_file
  for read_mapping_page() (although perhaps we can pass NULL),
  and we need ->mm for is_64bit_mm() (although this logic is
  broken anyway).

- uprobe_register() should be called by create_trace_uprobe(),
  not by probe_event_enable(), so that an error can be detected
  at "perf probe -x" time. This also needs more changes in the
  core uprobe code, uprobe register/unregister interface was
  poorly designed from the very beginning.

Reported-by: Denys Vlasenko <dvlasenk@redhat.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140519184054.GA6750@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86: Use common PMU interrupt disabled code

Make the x86 perf code use the new common PMU interrupt disabled code.

Typically most x86 machines have working PMU interrupts, although
some older p6-class machines had this problem.

Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1405161715560.11099@vincent-weaver-1.umelst.maine.edu
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/ARM: Use common PMU interrupt disabled code

Make the ARM perf code use the new common PMU interrupt disabled code.

This allows perf to work on ARM machines without a working PMU
interrupt (for example, raspberry pi).

Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
[peterz: applied changes suggested by Will]
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Grant Likely <grant.likely@linaro.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Cc: devicetree@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1405161712190.11099@vincent-weaver-1.umelst.maine.edu
[ Small readability tweaks to the code. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf: Disable sampled events if no PMU interrupt

Add common code to generate -ENOTSUPP at event creation time if an
architecture attempts to create a sampled event and
PERF_PMU_NO_INTERRUPT is set.

This adds a new pmu->capabilities flag.  Initially we only support
PERF_PMU_NO_INTERRUPT (to indicate a PMU has no support for generating
hardware interrupts) but there are other capabilities that can be
added later.

Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
Acked-by: Will Deacon <will.deacon@arm.com>
[peterz: rename to PERF_PMU_CAP_* and moved the pmu::capabilities word into a hole]
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1405161708060.11099@vincent-weaver-1.umelst.maine.edu
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf: Fix use after free in perf_remove_from_context()

While that mutex should guard the elements, it doesn't guard against the
use-after-free that's from list_for_each_entry_rcu().
__perf_event_exit_task() can actually free the event.

And because list addition/deletion is guarded by both ctx->mutex and
ctx->lock, holding ctx->mutex is sufficient for reading the list, so we
don't actually need the rcu list iteration.

Fixes: 3a497f48637e ("perf: Simplify perf_event_exit_task_context()")
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Tested-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Dave Jones <davej@redhat.com>
Cc: acme@ghostprotocols.net
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140529170024.GA2315@laptop.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge branch 'perf/kprobes' into perf/core

Conflicts:
arch/x86/kernel/traps.c

The kprobes enhancements are fully cooked, ship them upstream.

Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge branch 'perf/uprobes' into perf/core

These bits from Oleg are fully cooked, ship them to Linus.

Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/jolsa/perf into perf/core

Pull perf/core improvements and fixes from Jiri Olsa:

  * Warn the user when trace command is not available (Arnaldo Carvalho de Melo)

  * Add warning when disabling perl scripting support due to missing devel files (Arnaldo Carvalho de Melo)

  * Consider header files outside perf directory in tags target (Sebastian Andrzej Siewior)

  * Allow overriding sysfs and proc finding with env var (Cody P Schafer)

  * Fix "==" into "=" in ui_browser__warning assignment (zhangdianfang)

  * Factor elide bool handling in sort code (Jiri Olsa)

  * Fix poll return value propagation (Jiri Olsa)

  * Fix 'make help' message error (Jianyu Zhan)

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf tools: Fix 'make help' message error

Currently 'make help' message has such hint:

   use "make prefix=<path> <install target>" to install to a particular
       path like make prefix=/usr/local install install-doc

But this is misleading, when I specify "prefix=/usr/local", it has got no
respect at all.

This is because that, "DESTDIR" is considered first. In this case, "DESTDIR"
has an empty value, so "prefix" is honored. However, "prefix" is unconditionally
assigned to $HOME, regardless of what it is set to from command line. So our
"prefix" setting got no respect and the actual destination falls back to $HOME.

This patch fixes this issue and corrects the help message.

Signed-off-by: Jianyu Zhan <nasa4836@gmail.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1401727474-19370-1-git-send-email-nasa4836@gmail.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf record: Fix poll return value propagation

If the perf record command is interrupted in record__mmap_read_all
function, the 'done' is set and err has the latest poll return
value, which is most likely positive number (= number of pollfds
ready to read).

This 'positive err' is then propagated to the exit code, resulting
in not finishing the perf.data header properly, causing following
error in report:

  # perf record -F 50000 -a

  ---
  make the system real busy, so there's more chance
  to interrupt perf in event writing code
  ---

  ^C[ perf record: Woken up 16 times to write data ]
  [ perf record: Captured and wrote 30.292 MB perf.data (~1323468 samples) ]

  # perf report --stdio > /dev/null
  WARNING: The perf.data file's data size field is 0 which is unexpected.
  Was the 'perf record' command properly terminated?

Fixing this by checking for positive poll return value
and setting err to 0.

Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1401732126-19465-1-git-send-email-jolsa@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Move elide bool into perf_hpp_fmt struct

After output/sort fields refactoring, it's expensive
to check the elide bool in its current location inside
the 'struct sort_entry'.

The perf_hpp__should_skip function gets highly noticable in
workloads with high number of output/sort fields, like for:

  $ perf report -i perf-test.data -F overhead,sample,period,comm,pid,dso,symbol,cpu --stdio

Performance report:
   9.70%  perf  [.] perf_hpp__should_skip

Moving the elide bool into the 'struct perf_hpp_fmt', which
makes the perf_hpp__should_skip just single struct read.

Got speedup of around 22% for my test perf.data workload.
The change should not harm any other workload types.

Performance counter stats for (10 runs):
  before:
   358,319,732,626      cycles                    ( +-  0.55% )
   467,129,581,515      instructions              #    1.30  insns per cycle          ( +-  0.00% )

     150.943975206 seconds time elapsed           ( +-  0.62% )

  now:
   278,785,972,990      cycles                    ( +-  0.12% )
   370,146,797,640      instructions              #    1.33  insns per cycle          ( +-  0.00% )

     116.416670507 seconds time elapsed           ( +-  0.31% )

Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/20140601142622.GA9131@krava.brq.redhat.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Remove elide setup for SORT_MODE__MEMORY mode

There's no need to setup elide of sort_dso sort entry again
with symbol_conf.dso_list list.

The only difference were list names of memory mode data,
which does not make much sense to me.

Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1400858147-7155-2-git-send-email-jolsa@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Fix "==" into "=" in ui_browser__warning assignment

Convert "==" into "=" in ui_browser__warning assignment.

Bug description: https://bugzilla.kernel.org/show_bug.cgi?id=76751

Reported-by: David Binderman <dcb314@hotmail.com>
Signed-off-by: Dianfang Zhang <zhangdianfang@huawei.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jean Delvare <jdelvare@suse.de>
Link: http://lkml.kernel.org/r/20140530154709.GC1202@kernel.org
[ changed the changelog a bit ]
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Allow overriding sysfs and proc finding with env var

SYSFS_PATH and PROC_PATH environment variables now let the user override
the detection of sysfs and proc locations for testing purposes.

Signed-off-by: Cody P Schafer <dev@codyps.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1401236684-10579-2-git-send-email-dev@codyps.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Consider header files outside perf directory in tags target

This fixes lookups like "vi -t event_format"

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Link: http://lkml.kernel.org/r/20140528081918.GA28567@linutronix.de
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Add warning when disabling perl scripting support due to missing devel files

We were just showing "libperl: OFF", unlike other features where we
present the user with a message helping have a feature built in.

Fix it by adding the following message:

config/Makefile:450: Missing perl devel files. Disabling perl scripting support, consider installing perl-ExtUtils-Embed

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-t7yeud34ehimlfi6pklb29p7@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf trace: Warn the user when not available

When the audit-libs devel package is not found at build time we disable
the 'trace' command, as we are not able to map syscall numbers to
strings, but then the message the user is presented is cryptic:

  [root@zoo linux]# trace ls
  perf: 'ls' is not a perf-command. See 'perf --help'.

Fix it by presenting a more helpful message:

  [root@zoo linux]# trace l
  trace command not available: missing audit-libs devel package at build time.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-uxeunqetd0sgxyibusapen9a@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/jolsa/perf into perf/core

Pull perf/core improvements and fixes from Jiri Olsa:

  * Add support to accumulate hist periods (Namhyung Kim)

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf tests: Add a test case for cumulating callchains

Now it adds a new testcase to verify --children option working
correctly.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arun Sharma <asharma@fb.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-28-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tests: Define and use symbolic names for fake symbols

In various histogram test cases, fake symbols are used as raw numbers.
Define macros for each pid, map, symbols so that it can increase
readability somewhat.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arun Sharma <asharma@fb.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-27-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Reset output/sort order to default

When reset_output_field() is called, also reset field/sort order to
NULL so that it can have the default values.  It's needed for testing.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
CC: Arun Sharma <asharma@fb.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-26-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf ui/gtk: Fix callchain display

With current output field change, GTK browser cannot display callchain
information correctly since it couldn't determine where the symbol
column is.  This is a problem - just for now I changed to use the last
column since it'll work for most cases.

Also it has a same problem of the percentage as stdio code.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arun Sharma <asharma@fb.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-25-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf ui/stdio: Fix invalid percentage value of cumulated hist entries

On stdio, there's a problem that it shows invalid values for
callchains in cumulated hist entries.  It's because it only cares
about the self period.  But with --children behavior, we always add
callchain info to the cumulated entries so it should use the value in
that case.

Before:

  # Children      Self  Command      Shared Object            Symbol
  # ........  ........  .......  .................  ................
  #
      61.22%     0.32%  swapper  [kernel.kallsyms]      [k] cpu_idle
                    |
                    --- cpu_idle
                       |
                       |--16530.76%-- start_secondary
                       |
                       |--2758.70%-- rest_init
                       |          start_kernel
                       |          x86_64_start_reservations
                       |          x86_64_start_kernel
                        --6837850969203030.00%-- [...]

After:

  # Children      Self  Command      Shared Object            Symbol
  # ........  ........  .......  .................  ................
  #
      61.22%     0.32%  swapper  [kernel.kallsyms]      [k] cpu_idle
                    |
                    --- cpu_idle
                       |
                       |--85.70%-- start_secondary
                       |
                        --14.30%-- rest_init
                                  start_kernel
                                  x86_64_start_reservations
                                  x86_64_start_kernel

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arun Sharma <asharma@fb.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-24-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Enable --children option by default

Now perf top and perf report will show children column by default if
it has callchain information.

Requested-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Tested-by: Arun Sharma <asharma@fb.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-23-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf top: Add top.children config option

Add top.children config option for setting default value of
callchain accumulation.  It affects the output only if one of
-g or --call-graph option is given as well.

A user can write .perfconfig file like below to enable accumulation
by default:

  $ cat ~/.perfconfig
  [top]
  children = true

And it can be disabled through command line:

  $ perf top --no-children

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-22-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf top: Add --children option

The --children option is for showing accumulated overhead (period)
value as well as self overhead.  It should be used with one of -g or
--call-graph option.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-21-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf top: Convert to hist_entry_iter

Reuse hist_entry_iter__add() function to share the similar code with
perf report.  Note that it needs to be called with hists.lock so tweak
some internal functions not to deadlock or hold the lock too long.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Link: http://lkml.kernel.org/r/1401335910-16832-20-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Add callback function to hist_entry_iter

The new ->add_entry_cb() will be called after an entry was added to
the histogram.  It's used for code sharing between perf report and
perf top.  Note that ops->add_*_entry() should set iter->he properly
in order to call the ->add_entry_cb.

Also pass @arg to the callback function.  It'll be used by perf top
later.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/87k393g999.fsf@sejong.aot.lge.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Do not auto-remove Children column if --fields given

Depending on the configuration perf inserts/removes the Children
column in the output automatically.  But it might not be what user
wants if [s]he give --fields option explicitly.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Arun Sharma <asharma@fb.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-18-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf report: Add report.children config option

Add report.children config option for setting default value of
callchain accumulation.  It affects the report output only if
perf.data contains callchain info.

A user can write .perfconfig file like below to enable accumulation
by default:

  $ cat ~/.perfconfig
  [report]
  children = true

And it can be disabled through command line:

  $ perf report --no-children

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-17-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf report: Add --children option

The --children option is for showing accumulated overhead (period)
value as well as self overhead.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-16-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Add more hpp helper functions

Sometimes it needs to disable some columns at runtime.  Add help
functions to support that.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-15-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Apply percent-limit to cumulative percentage

If -g cumulative option is given, it needs to show entries which don't
have self overhead.  So apply percent-limit to accumulated overhead
percentage in this case.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-14-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf ui/gtk: Add support to accumulated hist stat

Print accumulated stat of a hist entry if requested.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-13-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf ui/browser: Add support to accumulated hist stat

Print accumulated stat of a hist entry if requested.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-12-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf ui/hist: Add support to accumulated hist stat

Print accumulated stat of a hist entry if requested.

To do that, add new HPP_PERCENT_ACC_FNS macro and generate a
perf_hpp_fmt using it.  The __hpp__sort_acc() function sorts entries
by accumulated period value.  When accumulated periods of two entries
are same (i.e. single path callchain) put the caller above since
accumulation tends to put callers on higher position for obvious
reason.

Also add "overhead_children" output field to be selected by user.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-11-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Save callchain info for each cumulative entry

When accumulating callchain entry, also save current snapshot of the
chain so that it can show the rest of the chain.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-10-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf callchain: Add callchain_cursor_snapshot()

The callchain_cursor_snapshot() is for saving current status of the
callchain.  It'll be used to accumulate callchain information for each node.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-9-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf report: Cache cumulative callchains

It is possble that a callchain has cycles or recursive calls.  In that
case it'll end up having entries more than 100% overhead in the
output.  In order to prevent such entries, cache each callchain node
and skip if same entry already cumulated.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-8-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Update cpumode for each cumulative entry

The cpumode and level in struct addr_localtion was set for a sample
and but updated as cumulative callchains were added.  This led to have
non-matching symbol and cpumode in the output.

Update it accordingly based on the fact whether the map is a part of
the kernel or not.  This is a reverse of what thread__find_addr_map()
does.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-7-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf hists: Accumulate hist entry stat based on the callchain

Call __hists__add_entry() for each callchain node to get an
accumulated stat for an entry.  Introduce new cumulative_iter ops to
process them properly.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-6-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf hists: Check if accumulated when adding a hist entry

To support callchain accumulation, @entry should be recognized if it's
accumulated or not when add_hist_entry() called.  The period of an
accumulated entry should be added to ->stat_acc but not ->stat. Add
@sample_self arg for that.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-5-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf hists: Add support for accumulated stat of hist entry

Maintain accumulated stat information in hist_entry->stat_acc if
symbol_conf.cumulate_callchain is set.  Fields in ->stat_acc have same
vaules initially, and will be updated as callchain is processed later.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/r/1401335910-16832-4-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Introduce struct hist_entry_iter

There're some duplicate code when adding hist entries.  They are
different in that some have branch info or mem info but generally do
same thing.  So introduce new struct hist_entry_iter and add callbacks
to customize each case in general way.

The new perf_evsel__add_entry() function will look like:

  iter->prepare_entry();
  iter->add_single_entry();

  while (iter->next_entry())
    iter->add_next_entry();

  iter->finish_entry();

This will help further work like the cumulative callchain patchset.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Tested-by: Arun Sharma <asharma@fb.com>
Tested-by: Rodrigo Campos <rodrigo@sdfg.com.ar>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1401335910-16832-3-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Introduce hists__inc_nr_samples()

There're some duplicate code for counting number of samples.  Add
hists__inc_nr_samples() and reuse it.

Suggested-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1401335910-16832-2-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

Merge branch 'uprobes/core' of git://git./linux/kernel/git/oleg/misc into perf/uprobes

Pull uprobes fixes and changes from Oleg Nesterov:

" Denys found another nasty old bug in uprobes/x86: div, mul, shifts with
  count in CL, and cmpxchg are not handled correctly.

  Plus a couple of other minor fixes. Nobody acked the changes in x86/traps,
  hopefully they are simple enough, and I believe that they make sense anyway
  and allow to do more cleanups."

Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/jolsa/perf into perf/core

Pull perf/core improvements and fixes from Jiri Olsa:

  * Android related fixes for pager and map dso resolving (Michael Lentine)

  * Add -F option for specifying output fields (Namhyung Kim)

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf tools: Add automatic remapping of Android libraries

This patch automatically adjusts the path of MMAP records
associated with Android system libraries.

The Android system is organized with system libraries found in
/system/lib and user libraries in /data/app-lib. On the host system
(not running Android), system libraries can be found in the downloaded
NDK directory under ${NDK_ROOT}/platforms/${APP_PLATFORM}/arch-${ARCH}/usr/lib
and the user libraries are installed under libs/${APP_ABI} within
the apk build directory. This patch makes running the reporting
tools possible on the host system using the libraries from the NDK.

Signed-off-by: Michael Lentine <mlentine@google.com>
Reviewed-by: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1400579330-5043-3-git-send-email-eranian@google.com
[ fixed 'space required before the open parenthesis' checkpatch.pl errors ]
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Add cat as fallback pager

This patch adds a fallback to cat for the pager. This is useful
on environments, such as Android, where less does not exist.
It is better to default to cat than to abort.

Signed-off-by: Michael Lentine <mlentine@google.com>
Reviewed-by: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1400579330-5043-2-git-send-email-eranian@google.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tests: Add a testcase for histogram output sorting

With new output fields option, its internal implementation was changed
so add a new testcase to verify whether it breaks things.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-21-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tests: Factor out print_hists_*()

Those print helper functions can be reused by later hist test cases so
factor them out to a common location.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-20-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Introduce reset_output_field()

The reset_output_field() function is for clearing output field
settings and will be used for test code in later patch.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-19-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Get rid of obsolete hist_entry__sort_list

Now we moved to the perf_hpp_[_sort]_list so no need to keep the old
hist_entry__sort_list and sort__first_dimension.  Also the
hist_entry__sort_snprintf() can be gone as hist_entry__snprintf()
provides the functionality.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-18-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf hists: Reset width of output fields with header length

Some fields missed to set default column length so it broke align in
--stdio output.  Add perf_hpp__reset_width() to set it to a sane
default value.

Note that this change will ignore -w/--column-widths option for now.

Before:
  $ perf report -F cpu,comm,overhead --stdio
  ...
  # CPU          Command  Overhead
  #   ...............  ........
  #
    0          firefox     2.65%
    0      kworker/0:0     1.45%
    0          swapper     5.52%
    0         synergys     0.92%
    1          firefox     4.54%

After:
  # CPU          Command  Overhead
  # ...  ...............  ........
  #
      0          firefox     2.65%
      0      kworker/0:0     1.45%
      0          swapper     5.52%
      0         synergys     0.92%
      1          firefox     4.54%

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-17-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Skip elided sort entries

When it converted sort entries to hpp formats, it missed se->elide
handling, so add it for compatibility.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-16-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf top: Add --fields option to specify output fields

The --fields option is to allow user setup output field in any order.
It can receive any sort keys and following (hpp) fields:

  overhead, overhead_sys, overhead_us, sample and period

If guest profiling is enabled, overhead_guest_{sys,us} will be
available too.

More more information, please see previous patch "perf report:
Add -F option to specify output fields"

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-15-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf report/tui: Fix a bug when --fields/sort is given

The hists__filter_entries() function is called when down arrow key is
pressed for navigating through the entries in TUI.  It has a check for
filtering out entries that have very small overhead (under min_pcnt).

However it just assumed the entries are sorted by the overhead so when
it saw such a small overheaded entry, it just stopped navigating as an
optimization.  But it's not true anymore due to new --fields and
--sort optoin behavior and this case users cannot go down to a next
entry if ther's an entry with small overhead in-between.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-14-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Add ->sort() member to struct sort_entry

Currently, what the sort_entry does is just identifying hist entries
so that they can be grouped properly.  However, with -F option
support, it indeed needs to sort entries appropriately to be shown to
users.  So add ->sort() member to do it.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-13-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf report: Add -F option to specify output fields

The -F/--fields option is to allow user setup output field in any
order.  It can receive any sort keys and following (hpp) fields:

  overhead, overhead_sys, overhead_us, sample and period

If guest profiling is enabled, overhead_guest_{sys,us} will be
available too.

The output fields also affect sort order unless you give -s/--sort
option.  And any keys specified on -s option, will also be added to
the output field list automatically.

  $ perf report -F sym,sample,overhead
  ...
  #                     Symbol       Samples  Overhead
  # ..........................  ............  ........
  #
    [.] __cxa_atexit                       2     2.50%
    [.] __libc_csu_init                    4     5.00%
    [.] __new_exitfn                       3     3.75%
    [.] _dl_check_map_versions             1     1.25%
    [.] _dl_name_match_p                   4     5.00%
    [.] _dl_setup_hash                     1     1.25%
    [.] _dl_sysdep_start                   1     1.25%
    [.] _init                              5     6.25%
    [.] _setjmp                            6     7.50%
    [.] a                                  8    10.00%
    [.] b                                  8    10.00%
    [.] brk                                1     1.25%
    [.] c                                  8    10.00%

Note that, the example output above is captured after applying next
patch which fixes sort/comparing behavior.

Requested-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-12-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Call perf_hpp__init() before setting up GUI browsers

So that it can be set properly prior to set up output fields.  That
makes easy to handle/warn errors during the setup since it doesn't
need to be bothered with the GUI.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-11-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Consolidate management of default sort orders

The perf uses different default sort orders for different use-cases,
and this was scattered throughout the code.  Add get_default_sort_
order() function to handle this and change initial value of sort_order
to NULL to distinguish it from user-given one.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1400480762-22852-10-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Allow hpp fields to be sort keys

Add overhead{,_sys,_us,_guest_sys,_guest_us}, sample and period sort
keys so that they can be selected with --sort/-s option.

  $ perf report -s period,comm --stdio
  ...
  # Overhead        Period          Command
  # ........  ............  ...............
  #
      47.06%           152          swapper
      13.93%            45  qemu-system-arm
      12.38%            40         synergys
       3.72%            12          firefox
       2.48%             8            xchat

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-9-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf ui: Get rid of callback from __hpp__fmt()

The callback was used by TUI for determining color of folded sign
using percent of first field/column. But it cannot be used anymore
since it now support dynamic reordering of output field.

So move the logic to the hist_browser__show_entry().

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-8-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Consolidate output field handling to hpp format routines

Until now the hpp and sort functions do similar jobs different ways.
Since the sort functions converted/wrapped to hpp formats it can do
the job in a uniform way.

The perf_hpp__sort_list has a list of hpp formats to sort entries and
the perf_hpp__list has a list of hpp formats to print output result.

To have a backward compatibility, it automatically adds 'overhead'
field in front of sort list.  And then all of fields in sort list
added to the output list (if it's not already there).

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/n/tip-7g3h86woz2sckg3h1lj42ygj@git.kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Use hpp formats to sort final output

Convert output sorting function to use ->sort hpp functions.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-6-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Support event grouping in hpp ->sort()

Move logic of hist_entry__sort_on_period to __hpp__sort() in order to
support event group report.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-5-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Use hpp formats to sort hist entries

It wrapped sort entries to hpp functions, so using the hpp sort list
to sort entries.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-4-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Convert sort entries to hpp formats

This is a preparation of consolidating management of output field and
sort keys.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-3-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Add ->cmp(), ->collapse() and ->sort() to perf_hpp_fmt

Those function pointers will be used to sort report output based on
the selected fields.  This is a preparation of later change.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1400480762-22852-2-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/jolsa/perf into perf/core

Pull perf/core improvements and fixes from Jiri Olsa:

  * Add libdw DWARF post unwind support for ARM (Jean Pihet)

  * Consolidate types.h for ARM and ARM64 (Jean Pihet)

  * Fix possible null pointer dereference in session.c (Masanari Iida)

  * Cleanup, remove unused variables in map_switch_event() (Dongsheng Yang)

  * Remove nr_state_machine_bugs in perf latency (Dongsheng Yang)

  * Remove usage of trace_sched_wakeup(.success) (Peter Zijlstra)

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

fix Haswell precise store data source encoding

This patch fixes a bug in  precise_store_data_hsw() whereby
it would set the data source memory level to the wrong value.

As per the the SDM Vol 3b Table 18-41 (Layout of Data Linear
Address Information in PEBS Record), when status bit 0 is set
this is a L1 hit, otherwise this is a L1 miss.

This patch encodes the memory level according to the specification.

In V2, we added the filtering on the store events.
Only the following events produce L1 information:
 * MEM_UOPS_RETIRED.STLB_MISS_STORES
 * MEM_UOPS_RETIRED.LOCK_STORES
 * MEM_UOPS_RETIRED.SPLIT_STORES
 * MEM_UOPS_RETIRED.ALL_STORES

Cc: mingo@elte.hu
Cc: acme@ghostprotocols.net
Cc: jolsa@redhat.com
Cc: jmario@redhat.com
Cc: ak@linux.intel.com
Tested-and-Reviewed-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140515155644.GA3884@quad
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

perf: Fix perf_event_open(.flags) test

Vince noticed that we test the (unsigned long) flags field against an
(unsigned int) constant. This would allow setting the high bits on 64bit
platforms and not get an error.

There is nothing that uses the high bits, so it should be entirely
harmless, but we don't want userspace to accidentally set them anyway,
so fix the constants.

Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Tested-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140423102254.GL11096@twins.programming.kicks-ass.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

perf/events/core: Drop unused variable after cleanup

... in 3a497f48637 ("perf: Simplify perf_event_exit_task_context()")

Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1399720259-28275-1-git-send-email-bp@alien8.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

perf tools: Add libdw DWARF post unwind support for ARM

Adding libdw DWARF post unwind support, which is part
of elfutils-devel/libdw-dev package from version 0.158.

The new code is contained in unwin-libdw.c object, and
implements unwind__get_entries unwind interface function.

Signed-off-by: Jean Pihet <jean.pihet@linaro.org>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1400229672-16104-4-git-send-email-jean.pihet@linaro.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tests: Add dwarf unwind test on ARM

Adding dwarf unwind test, that setups live machine data over
the perf test thread and does the remote unwind.

Need to use -fno-optimize-sibling-calls for test compilation,
otherwise 'krava_*' function calls are optimized into jumps
and omitted from the stack unwind.

So far it was enabled only for x86.

Signed-off-by: Jean Pihet <jean.pihet@linaro.org>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1400229672-16104-3-git-send-email-jean.pihet@linaro.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tests: Introduce perf_regs_load function on ARM

Introducing perf_regs_load function, which is going
to be used for dwarf unwind test in following patches.

It takes single argument as a pointer to the regs dump
buffer and populates it with current registers values.

Signed-off-by: Jean Pihet <jean.pihet@linaro.org>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1400229672-16104-2-git-send-email-jean.pihet@linaro.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Consolidate types.h for ARM and ARM64

Prevents a build breakage since commit d944c4eebcf4c0d5e5d9728fec110cbf0047ad7f
'tools: Consolidate types.h'

Signed-off-by: Jean Pihet <jean.pihet@linaro.org>
Link: http://lkml.kernel.org/r/1400229672-16104-1-git-send-email-jean.pihet@linaro.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf session: Fix possible null pointer dereference in session.c

cppcheck detected following warning:
[tools/perf/util/session.c:1628] -> [tools/perf/util/session.c:1632]:
 (warning) Possible null pointer dereference: session - otherwise it
 is redundant to check it against null.

In order to avoide null pointer, check the pointer before use.

Signed-off-by: Masanari Iida <standby24x7@gmail.com>
Link: http://lkml.kernel.org/r/1400087618-13628-1-git-send-email-standby24x7@gmail.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf sched: Cleanup, remove unused variables in map_switch_event()

In map_switch_event(), we don't care the previous process currently,
this patch remove the infomation we get but not used.

Signed-off-by: Dongsheng Yang <yangds.fnst@cn.fujitsu.com>
Link: http://lkml.kernel.org/r/1400218625-14613-1-git-send-email-yangds.fnst@cn.fujitsu.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf sched: Remove nr_state_machine_bugs in perf latency

As we do not use .success in sched_wakeup event any more, then
we can not guarantee that the task when wakeup event happen is
out of run queue. So the message of nr_state_machine_bugs is
not correct.

Signed-off-by: Dongsheng Yang <yangds.fnst@cn.fujitsu.com>
Link: http://lkml.kernel.org/r/1399945101-21736-1-git-send-email-yangds.fnst@cn.fujitsu.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

uprobes/x86: Fix the wrong ->si_addr when xol triggers a trap

If the probed insn triggers a trap, ->si_addr = regs->ip is technically
correct, but this is not what the signal handler wants; we need to pass
the address of the probed insn, not the address of xol slot.

Add the new arch-agnostic helper, uprobe_get_trap_addr(), and change
fill_trap_info() and math_error() to use it. !CONFIG_UPROBES case in
uprobes.h uses a macro to avoid include hell and ensure that it can be
compiled even if an architecture doesn't define instruction_pointer().

Test-case:

#include <signal.h>
#include <stdio.h>
#include <unistd.h>

extern void probe_div(void);

void sigh(int sig, siginfo_t *info, void *c)
{
int passed = (info->si_addr == probe_div);
printf(passed ? "PASS\n" : "FAIL\n");
_exit(!passed);
}

int main(void)
{
struct sigaction sa = {
.sa_sigaction = sigh,
.sa_flags = SA_SIGINFO,
};

sigaction(SIGFPE, &sa, NULL);

asm (
"xor %ecx,%ecx\n"
".globl probe_div; probe_div:\n"
"idiv %ecx\n"
);

return 0;
}

it fails if probe_div() is probed.

Note: show_unhandled_signals users should probably use this helper too,
but we need to cleanup them first.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>

x86/traps: Kill DO_ERROR_INFO()

Now that DO_ERROR_INFO() doesn't differ from DO_ERROR() we can remove
it and use DO_ERROR() instead.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>

x86/traps: Shift fill_trap_info() from DO_ERROR_INFO() to do_error_trap()

Move the callsite of fill_trap_info() into do_error_trap() and remove
the "siginfo_t *info" argument.

This obviously breaks DO_ERROR() which passed info == NULL, we simply
change fill_trap_info() to return "siginfo_t *" and add the "default"
case which returns SEND_SIG_PRIV.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>

x86/traps: Introduce fill_trap_info(), simplify DO_ERROR_INFO()

Extract the fill-siginfo code from DO_ERROR_INFO() into the new helper,
fill_trap_info().

It can calculate si_code and si_addr looking at trapnr, so we can remove
these arguments from DO_ERROR_INFO() and simplify the source code. The
generated code is the same, __builtin_constant_p(trapnr) == T.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>

x86/traps: Introduce do_error_trap()

Move the common code from DO_ERROR() and DO_ERROR_INFO() into the new
helper, do_error_trap(). This simplifies define's and shaves 527 bytes
from traps.o.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>

x86/traps: Use SEND_SIG_PRIV instead of force_sig()

force_sig() is just force_sig_info(SEND_SIG_PRIV). Imho it should die,
we have too many ugly "send signal" helpers.

And do_trap() looks just ugly because it uses force_sig_info() or
force_sig() depending on info != NULL.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>

x86/traps: Make math_error() static

Trivial, make math_error() static.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>

uprobes/x86: Fix scratch register selection for rip-relative fixups

Before this patch, instructions such as div, mul, shifts with count
in CL, cmpxchg are mishandled.

This patch adds vex prefix handling. In particular, it avoids colliding
with register operand encoded in vex.vvvv field.

Since we need to avoid two possible register operands, the selection of
scratch register needs to be from at least three registers.

After looking through a lot of CPU docs, it looks like the safest choice
is SI,DI,BX. Selecting BX needs care to not collide with implicit use of
BX by cmpxchg8b.

Test-case:

#include <stdio.h>

static const char *const pass[] = { "FAIL", "pass" };

long two = 2;
void test1(void)
{
long ax = 0, dx = 0;
asm volatile("\n"
" xor %%edx,%%edx\n"
" lea 2(%%edx),%%eax\n"
// We divide 2 by 2. Result (in eax) should be 1:
" probe1: .globl probe1\n"
" divl two(%%rip)\n"
// If we have a bug (eax mangled on entry) the result will be 2,
// because eax gets restored by probe machinery.
: "=a" (ax), "=d" (dx) /*out*/
: "0" (ax), "1" (dx) /*in*/
: "memory" /*clobber*/
);
dprintf(2, "%s: %s\n", __func__,
pass[ax == 1]
);
}

long val2 = 0;
void test2(void)
{
long old_val = val2;
long ax = 0, dx = 0;
asm volatile("\n"
" mov val2,%%eax\n"     // eax := val2
" lea 1(%%eax),%%edx\n" // edx := eax+1
// eax is equal to val2. cmpxchg should store edx to val2:
" probe2: .globl  probe2\n"
" cmpxchg %%edx,val2(%%rip)\n"
// If we have a bug (eax mangled on entry), val2 will stay unchanged
: "=a" (ax), "=d" (dx) /*out*/
: "0" (ax), "1" (dx) /*in*/
: "memory" /*clobber*/
);
dprintf(2, "%s: %s\n", __func__,
pass[val2 == old_val + 1]
);
}

long val3[2] = {0,0};
void test3(void)
{
long old_val = val3[0];
long ax = 0, dx = 0;
asm volatile("\n"
" mov val3,%%eax\n"  // edx:eax := val3
" mov val3+4,%%edx\n"
" mov %%eax,%%ebx\n" // ecx:ebx := edx:eax + 1
" mov %%edx,%%ecx\n"
" add $1,%%ebx\n"
" adc $0,%%ecx\n"
// edx:eax is equal to val3. cmpxchg8b should store ecx:ebx to val3:
" probe3: .globl  probe3\n"
" cmpxchg8b val3(%%rip)\n"
// If we have a bug (edx:eax mangled on entry), val3 will stay unchanged.
// If ecx:edx in mangled, val3 will get wrong value.
: "=a" (ax), "=d" (dx) /*out*/
: "0" (ax), "1" (dx) /*in*/
: "cx", "bx", "memory" /*clobber*/
);
dprintf(2, "%s: %s\n", __func__,
pass[val3[0] == old_val + 1 && val3[1] == 0]
);
}

int main(int argc, char **argv)
{
test1();
test2();
test3();
return 0;
}

Before this change all tests fail if probe{1,2,3} are probed.

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Reviewed-by: Jim Keniston <jkenisto@us.ibm.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>

uprobes/x86: Simplify rip-relative handling

It is possible to replace rip-relative addressing mode with addressing
mode of the same length: (reg+disp32). This eliminates the need to fix
up immediate and correct for changing instruction length.

And we can kill arch_uprobe->def.riprel_target.

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
Reviewed-by: Jim Keniston <jkenisto@us.ibm.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>

uprobes: Add mem_cgroup_charge_anon() into uprobe_write_opcode()

Hugh says:

    The one I noticed was that it forgets all about memcg (because
    it was copied from KSM, and there the replacement page has already
    been charged to a memcg). See how mm/memory.c do_anonymous_page()
    does a mem_cgroup_charge_anon().

Hopefully not a big problem, uprobes is a system-wide thing and only
root can insert the probes. But I agree, should be fixed anyway.

Add mem_cgroup_{un,}charge_anon() into uprobe_write_opcode(). To simplify
the error handling (and avoid the new "uncharge" label) the patch also
moves anon_vma_prepare() up before we alloc/charge the new page.

While at it fix the comment about ->mmap_sem, it is held for write.

Suggested-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>

perf tools: Remove usage of trace_sched_wakeup(.success)

trace_sched_wakeup(.success) is a dead argument and has been for ages,
the only reason its still there is because of brain dead software, which
apparently includes perf tools

There's a few more instances in pearly snake shit, but that's not
supported as far as I care anyhow, so let that bitrot.

Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140512181946.GG13467@laptop.programming.kicks-ass.net
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/jolsa/perf into perf/core

Pull perf/core improvements and fixes from Jiri Olsa:

  * Propagate exit status of a command line workload for
    record command (Namhyung Kim)

  * Use tid for finding thread (Namhyung Kim)

  * Clarify the output of perf sched map plus small sched
    command fixies (Dongsheng Yang)

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf tools: Use tid for finding thread

I believe that passing pid (instead of tid) as the 3rd arg of the
machine__find*_thread() was to find a main thread so that it can
search proper map group for symbols.  However with the map sharing
patch applied, it now can do it in any thread.

It fixes a bug when each thread has different name, it only reports a
main thread for samples in other threads.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: David Ahern <dsahern@gmail.com>
Acked-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/1399856202-26221-1-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Get rid of on_exit() feature test

The on_exit() function was only used in perf record but it's gone in
previous patch.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Stephane Eranian <eranian@google.com>
Cc: Bernhard Rosenkraenzer <Bernhard.Rosenkranzer@linaro.org>
Cc: Irina Tirdea <irina.tirdea@intel.com>
Link: http://lkml.kernel.org/r/1399855645-25815-2-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf record: Propagate exit status of a command line workload

Currently perf record doesn't propagate the exit status of a workload
given by the command line.  But sometimes it'd useful if it's
propagated so that a monitoring script can handle errors
appropriately.

To do that, it moves most of logic out of the exit handlers and run
them directly in the __cmd_record().  The only thing needs to be done
in the handler is propagating terminating signal so that the shell can
terminate its loop properly when Ctrl-C was pressed.  Also it cleaned
up the resource management code in record__exit().

With this change, perf record returns the child exit status in case of
normal termination and send signal to itself when terminated by signal.

Example run of Stephane's case:

  $ perf record true && echo yes || echo no
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.013 MB perf.data (~589 samples) ]
  yes

  $ perf record false && echo yes || echo no
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.013 MB perf.data (~589 samples) ]
  no

Jiri's case (error in parent):

  $ perf record -m 10G true && echo yes || echo no
  rounding mmap pages size to 17179869184 bytes (4194304 pages)
  failed to mmap with 12 (Cannot allocate memory)
  no

  $ ulimit -n 6
  $ perf record sleep 1 && echo yes || echo no
  failed to create 'go' pipe: Too many open files
  Couldn't run the workload!
  no

And Peter's case (interrupted by signal):

  $ while :; do perf record sleep 1; done
  ^C[ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.014 MB perf.data (~593 samples) ]

Reported-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Stephane Eranian <eranian@google.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1399855645-25815-1-git-send-email-namhyung@kernel.org
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Clarify the output of perf sched map.

In output of perf sched map, any shortname of thread will be explained
at the first time when it appear.

Example:
              *A0       228836.978985 secs A0 => perf:23032
          *.   A0       228836.979016 secs B0 => swapper:0
           .  *C0       228836.979099 secs C0 => migration/3:22
  *A0      .   C0       228836.979115 secs
   A0      .  *.        228836.979115 secs

But B0, which is explained as swapper:0 did not appear in the
left part of output. Instead, we use '.' as the shortname of
swapper:0. So the comment of "B0 => swapper:0" is not easy to
understand.

This patch clarify the output of perf sched map with not allocating
one letter-number shortname for swapper:0 and print ". => swapper:0"
as the explanation for swapper:0.

Example:
              *A0       228836.978985 secs A0 => perf:23032
          * .  A0       228836.979016 secs .  => swapper:0
            . *B0       228836.979099 secs B0 => migration/3:22
  *A0       .  B0       228836.979115 secs
   A0       . * .       228836.979115 secs
   A0     *C0   .       228836.979225 secs C0 => ksoftirqd/2:18
   A0     *D0   .       228836.979236 secs D0 => rcu_sched:7

Signed-off-by: Dongsheng <yangds.fnst@cn.fujitsu.com>
Acked-by: Ingo Molnar <mingo@kernel.org>
Link: http://lkml.kernel.org/r/1399354741-19522-1-git-send-email-yangds.fnst@cn.fujitsu.com
[ small style fixes to make checkpatch happy ]
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Adapt the TASK_STATE_TO_CHAR_STR to new value in kernel space.

Currently, TASK_STATE_TO_CHAR_STR in kernel space is already expanded to RSDTtZXxKWP,
but it is still RSDTtZX in perf sched tool.

This patch update TASK_STATE_TO_CHAR_STR to the new value in kernel space.

Signed-off-by: Dongsheng <yangds.fnst@cn.fujitsu.com>
Link: http://lkml.kernel.org/r/6d2f55dc1e02c1e29a5d70bfeb9d6e8863caf2aa.1399273302.git.yangds.fnst@cn.fujitsu.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

perf tools: Add missing event for perf sched record.

We should record and process sched:sched_wakeup_new event in
perf sched tool, but currently, there is the process function
for it, without recording it in record subcommand.

This patch add -e sched:sched_wakeup_new to perf sched record.

Signed-off-by: Dongsheng <yangds.fnst@cn.fujitsu.com>
Link: http://lkml.kernel.org/r/710c6edd2162b2cea1711443f54de47c0210d9fd.1399273302.git.yangds.fnst@cn.fujitsu.com
Signed-off-by: Jiri Olsa <jolsa@kernel.org>

kprobes: Ensure blacklist data is aligned

ARC Linux (not supporting native unaligned access) was failing
to boot because __start_kprobe_blacklist was not aligned.

This was because per generated vmlinux.lds it was emitted right
next to .rodata with strings etc hence could be randomly
unaligned.

Fix that by ensuring a word alignment. While 4 would suffice for
32bit arches and problem at hand, it is probably better to put 8.

| Path: (null) CPU: 0 PID: 1 Comm: swapper Not tainted
| 3.15.0-rc3-next-20140430 #2
| task: 8f044000 ti: 8f01e000 task.ti: 8f01e000
|
| [ECR   ]: 0x00230400 => Misaligned r/w from 0x800fb0d3
| [EFA   ]: 0x800fb0d3
| [BLINK ]: do_one_initcall+0x86/0x1bc
| [ERET  ]: init_kprobes+0x52/0x120

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Cc: <torvalds@linux-foundation.org>
Cc: <rusty@rustcorp.com.au>
Cc: <rdunlap@infradead.org>
Cc: <jeremy@goop.org>
Cc: <arnd@arndb.de>
Cc: <dl9pf@gmx.de>
Cc: <sparse@chrisli.org>
Cc: <anil.s.keshavamurthy@intel.com>
Cc: <davem@davemloft.net>
Cc: <ananth@in.ibm.com>
Cc: <masami.hiramatsu.pt@hitachi.com>
Cc: <chrisw@sous-sol.org>
Cc: <akataria@vmware.com>
Cc: anton Kolesov <Anton.Kolesov@synopsys.com>
Link: http://lkml.kernel.org/r/5361DB14.7010406@synopsys.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf: Simplify perf_event_exit_task_context()

Instead of jumping through hoops to make sure to find (and exit) each
event, do it the simple straight fwd way.

Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-tij931199thfkys8vbnokdpf@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>