perf tools: Convert to NEWT_SUPPORT

For building perf without libnewt, we can set NO_NEWT=1 as a argument of
make.  It then defines NO_NEWT_SUPPORT macro for C code to do the proper
handling.  However it usually used in a negative semantics - e.g.  #ifndef -
so we saw double negations which can be misleading.  Convert it to a
positive form to make it more readable.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348824728-14025-7-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Convert to LIBAUDIT_SUPPORT

For building perf without libaudit, we can set NO_LIBAUDIT=1 as a
argument of make.  It then defines NO_LIBAUDIT_SUPPORT macro for C code
to do the proper handling.  However it usually used in a negative
semantics - e.g. #ifndef - so we saw double negations which can be
misleading.  Convert it to a positive form to make it more readable.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348824728-14025-6-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Convert to LIBUNWIND_SUPPORT

For building perf without libunwind, we can set NO_LIBUNWIND=1 as a
argument of make.  It then defines NO_LIBUNWIND_SUPPORT macro for C code
to do the proper handling.  However it usually used in a negative
semantics - e.g. #ifndef - so we saw double negations which can be
misleading.  Convert it to a positive form to make it more readable.

Also change NO_PERF_REGS macro to HAVE_PERF_REGS for the same reason.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348824728-14025-5-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Convert to LIBELF_SUPPORT

For building perf without libelf, we can set NO_LIBELF=1 as a argument
of make.  It then defines NO_LIBELF_SUPPORT macro for C code to do the
proper handling.  However it usually used in a negative semantics -
e.g. #ifndef - so we saw double negations which can be misleading.
Convert it to a positive form to make it more readable.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348824728-14025-4-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Remove unused PYRF_OBJS variable on Makefile

It seems that the PYRF_OBJS variable is not used anymore or has no
effect at least.  The util/setup.py tracks its dependency using
util/python-ext-sources file and resulting objects are saved under
python_ext_build/tmp/.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348824728-14025-3-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Move libdw availability check before arch Makefile

Since NO_DWARF is used in arch/$(ARCH)/Makefiles, it should be checked
before including those files.  It was moved by mistake during libelf
dependency removal work by me, sorry.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348824728-14025-2-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

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

Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:

 * Improve listing of accessible enum perf probe variables, from Hyeoncheol Lee.

 * Don't stop the build if the audit libraries are not installed, fix from Namhyung Kim.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf trace: Add aliases for some syscalls

What we get from audit_syscall_to_name isn't what we find in the
syscalls: tracepoint events, so add the alias that allows the tool to
find prctl, fstat, fstatat and stat.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-3m9su7jhwnxvepnr3ne1du5k@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Print an enum type variable in "enum variable-name" format when showing accessible variables

When showing accessible variables, an enum type variable was printed in
"variable-name" format. Change this format into "enum variable-name".

Signed-off-by: Hyeoncheol Lee <hyc.lee@gmail.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Link: http://lkml.kernel.org/r/1348713399-4541-1-git-send-email-hyc.lee@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Check libaudit availability for perf-trace builtin

The newly added trace command requires an external audit library.

However it can cause a build error because it's not checked whether the
libaudit is installed on system:

    CC builtin-trace.o
  builtin-trace.c:7:22: fatal error: libaudit.h: No such file or directory
  compilation terminated.
  make: *** [builtin-trace.o] Error 1

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348745018-21744-1-git-send-email-namhyung@kernel.org
[ committer note: Added ", disables 'trace tool' to the feature warning msg ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf hists: Add missing period_* fields when collapsing a hist entry

So that the perf report won't lost the cpu utilization information.

For example, if there're two process that have same name.

  $ perf report --stdio --showcpuutilization -s pid
  [SNIP]
  #   Overhead       sys        us  Command:  Pid
  #   ........  ........  ........  .............
  #
        55.12%     0.01%    55.10%  noploop:28781
        44.88%     0.06%    44.83%  noploop:28782

Before:
  $ perf report --stdio --showcpuutilization -s comm
  [SNIP]
  #   Overhead       sys        us
  #   ........  ........  ........
  #
       100.00%     0.06%    44.83%

After:
  $ perf report --stdio --showcpuutilization -s comm
  [SNIP]
  #   Overhead       sys        us
  #   ........  ........  ........
  #
       100.00%     0.07%    99.93%

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Arun Sharma <asharma@fb.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
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/1348645663-25303-2-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: New tool

Initially should look loosely like the venerable 'strace' tool, but
using the infrastructure in the perf tools to allow tracing extra
targets:

  [acme@sandy linux]$ perf trace --hell
  Error: unknown option `hell'

   usage: perf trace <PID>

      -p, --pid <pid>       trace events on existing process id
          --tid <tid>       trace events on existing thread id
          --all-cpus        system-wide collection from all CPUs
          --cpu <cpu>       list of cpus to monitor
          --no-inherit      child tasks do not inherit counters
          --mmap-pages <n>  number of mmap data pages
          --uid <user>      user to profile

  [acme@sandy linux]$

Those should have the same semantics as when using with 'perf record'.

It gets stuck sometimes, but hey, it works sometimes too!

In time it should support perf.data based workloads, i.e. it should have
a:
-o filename

Command line option that will produce a perf.data file that can then be
used with 'perf trace' or any of the other perf tools (script, report,
etc).

It will also eventually have the set of functionalities described in the
previous 'trace' prototype by Thomas Gleixner:

   "Announcing a new utility: 'trace'"
   http://lwn.net/Articles/415728/

Also planned is to have some of the features suggested in the comments
of that LWN article.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/n/tip-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Export the event_format constructor

It'll be needed in the next patches, where it'll be not associated
directly to an evsel.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Introduce rawptr() method

Will be used for things like the args field in the raw_syscalls:sys_enter
tracepoint.

Implement strval with it, its basicaly strval returning void *.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Use perf_evsel__newtp in the event parser

Elliminating code duplication.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-9v4zl7ldlp8v6azrpsu5lupk@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: The tracepoint constructor should store sys:name

Not event_format->name, that doesn't contains the sys: part.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Introduce set_filter() method

To apply a filter to all the evsels in an evlist.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evlist: Renane set_filters method to apply_filters

Because that is what it really does, i.e. it applies the filters that
were parsed from the command line and stashed into the evsels they refer
to.

We'll need the set_filter method name to actually apply a filter to all
the evsels in an evlist, for instance, to ask that a syswide tracer
doesn't trace itself.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test: Add test to check we correctly parse and match syscall open parms

It will set up a syscall open tracepoint event, generate an open with
invalid flags, then check those flags were the ones reported in the
tracepoint fired.

For the filename we need vfs:getname, but that will go thru some more
iterations as the vfs getname codebase is going thru changes lately.

When that is in I'll just check that the perf_evsel__newtp constructor
is not bailing out and then add it to the evlist, catch the event and
check the filename against the one used in the 'open' call used to
trigger the event.

Cc: David Ahern <dsahern@gmail.com>
Cc: Eric Paris <eparis@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jeff Layton <jlayton@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-p5w9aq0jcbb91ghzqomowm16@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Handle endianity in intval method

We were relying on the info in pevent, but since we have it in
perf_evsel, set up by the perf_session routine if read from a perf.data
file or by whoever creates the evsels, use it.

New 'perf test' entries will use it to parse locally generated events,
in a non perf.data centered workflow.

As well as use byteswap.h to get per arch optimized swap routines, like
other parts of perf (header, perf_evsel__parse_sample, symbol, etc)
already do.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-8tjuxk09mlsfmh7macgkxsip@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Know if byte swap is needed

Instead of passing it around for parsing as an explicit parameter, will
help with reading tracepoint fields when not using a perf session or
pevent structure, i.e. for non perf.data centered workflows.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-qa67ikv2sm49cwa7dyjhhp6g@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Allow handling a NULL cpu_map as meaning "all cpus"

Or one with cpu_map->map[0] == -1.

Reducing the boilerplate in setting up an evlist by nor requiring a
cpu_map to be created at all.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-rnaqn3dtnsfo1wlbbf3fhx00@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Improve tracepoint constructor setup

It needs to properly set the sample_type, sample_period and the KVM
related perf_event_attr fields.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge branch 'tip/perf/core' of git://git./linux/kernel/git/rostedt/linux-trace into perf/core

Pull tracing updates from Steve Rostedt.

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

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

Pull perf/core fixes from Arnaldo Carvalho de Melo:

 * The new perf_evsel__tp_sched_test 'perf test' broke the build by setting the
   'ret' variable but not using it, caught by newer gcc
   -Werror=unused-but-set-variable, fix from Namhyung Kim.

 * pevent_parse_event should return a proper PEVENT_ERRNO__ and call
   pevent_free_format on its failure path, fixes from Namhyung Kim.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

tools lib traceevent: Fix error path on pevent_parse_event

If __pevent_parse_format() succeeded but add_event() failed, 'ret' didn't
have a proper error code.  Set it to PEVENT_ERRNO__MEM_ALLOC_FAILED.

In addition, at that point 'event' also has fields and format
information and they all need to be freed.  Call pevent_free_format() to
handle it.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1348575919-4954-1-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test: Fix build failure

The commit 6a6cd11d4e57 ("perf test: Add test for the sched tracepoint
format fields") added following build error:

    CC builtin-test.o
  builtin-test.c: In function ‘perf_evsel__test_field’:
  builtin-test.c:1216:6: error: variable ‘ret’ set but not used [-Werror=unused-but-set-variable]
  builtin-test.c: In function ‘perf_evsel__tp_sched_test’:
  builtin-test.c:1242:6: error: variable ‘ret’ set but not used [-Werror=unused-but-set-variable]
  cc1: all warnings being treated as errors
  make: *** [builtin-test.o] Error 1

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348539628-3821-1-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

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

Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:

 * Convert the trace builtins to use the growing evsel/evlist
   tracepoint infrastructure, removing several open coded constructs
   like switch like series of strcmp to dispatch events, etc.
   Basically what had already been showcased in 'perf sched'.

 * Add evsel constructor for tracepoints, that uses libtraceevent
   just to parse the /format events file, use it in a new 'perf test'
   to make sure the libtraceevent format parsing regressions can
   be more readily caught.

 * Some strange errors were happening in some builds, but not on the
   next, reported by several people, problem was some parser related
   files, generated during the build, didn't had proper make deps,
   fix from Eric Sandeen.

 * Fix some compiling errors on 32-bit, from Feng Tang.

 * Don't use sscanf extension %as, not available on bionic, reimplementation
   by Irina Tirdea.

 * Fix bfd.h/libbfd detection with recent binutils, from Markus Trippelsdorf.

 * Introduce struct and cache information about the environment where a
   perf.data file was captured, from Namhyung Kim.

 * Fix several error paths in libtraceevent, from Namhyung Kim.

   Print event causing perf_event_open() to fail in 'perf record',
   from Stephane Eranian.

 * New 'kvm' analysis tool, from Xiao Guangrong.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

trace: Move trace event enable from fs_initcall to core_initcall

This patch splits trace event initialization in two stages:
 * ftrace enable
 * sysfs event entry creation

This allows to capture trace events from an earlier point
by using 'trace_event' kernel parameter and is important
to trace boot-up allocations.

Note that, in order to enable events at core_initcall,
it's necessary to move init_ftrace_syscalls() from
core_initcall to early_initcall.

Link: http://lkml.kernel.org/r/1347461277-25302-1-git-send-email-elezegarcia@gmail.com
Signed-off-by: Ezequiel Garcia <elezegarcia@gmail.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

tracing: Add an option for disabling markers

In our application, we have trace markers spread through user-space.
We have markers in GL, X, etc. These are super handy for Chrome's
about:tracing feature (Chrome + system + kernel trace view), but
can be very distracting when you're trying to debug a kernel issue.

I normally, use "grep -v tracing_mark_write" but it would be nice
if I could just temporarily disable markers all together.

Link: http://lkml.kernel.org/r/1347066739-26285-1-git-send-email-msb@chromium.org
CC: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Mandeep Singh Baines <msb@chromium.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

tools lib traceevent: Handle alloc_arg failure

Now alloc_arg returns NULL if memory allocation failed, it should be
handled on callsites properly.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/87k3vpzbqo.fsf_-_@sejong.aot.lge.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test: Add test for the sched tracepoint format fields

So that we make sure the routines that do event format parsing are
working on at least two well know scheduler tracepoints.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-g3rm9b3wtim4djx3z8dkftrj@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Provide a new constructor for tracepoints

The existing constructor receives a perf_event_attr filled with the
event type and the config.

To reduce the boilerplate for tracepoints, provide a new constructor,
perf_evsel__newtp() that receives the tracepoint name and will open
the debugfs file, call into libtraceevent new pevent_parse_format file
to fill its ->tp_format member, so that users can then just call
perf_evsel__field() to access its fields.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-6du8dl1hz0y5l4cybodye7hn@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Carve out events format parsing routine

The pevent_parse_event() routine will parse a events/sys/tp/format file
and add an event_format instance to the pevent struct.

This patch introduces a pevent_parse_format() routine with just the bits
needed to parse the event/sys/tp/format file and just return the
event_format instance, useful for when all we want is to parse the
format file, without requiring the pevent struct.

Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-lge0afl47arh86om0m6a5bqr@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Remove some die() calls

Cleaned event-parse.c this time, just propagate the errors and in handle
them the call sites.

Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-9ebpr2vgfk2qs2841i99sa8y@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Fix afterlife gotos

Instead of dying, just use do_warning and let the goto that is there to
take place.

Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-aoaus46ngnt9oc2pt7ckot5d@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Use calloc were applicable

Replacing the equivalent open coded malloc + memset bits.

Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-598fjtjbzal4wxh7fp0yv0q1@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Use asprintf were applicable

Replacing the equivalent open coded malloc + sprintf bits.

Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-ghokwtdw2hgmmmn7oa9s03r4@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: bfd.h/libbfd detection fails with recent binutils

With recent binutils I get:

 perf % make
Makefile:668: No bfd.h/libbfd found, install binutils-dev[el]/zlib-static to gain symbol demanglin

That happens because bfd.h now contains:

I've reopened a bug in the hope that this check will be deleted:
http://sourceware.org/bugzilla/show_bug.cgi?id=14243

But in the meantime, the following patch fixes the problem

Signed-off-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Mike Frysinger <vapier@gentoo.org>
Cc: Paul Mackerras <paulus@samba.org>
Link: http://lkml.kernel.org/r/20120919072902.GA262@x4
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Free field if an error occurs on process_flags/symbols

The field should be freed on error paths.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1348037924-17568-5-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Free field if an error occurs on process_fields

The field should be freed on error paths.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1348037924-17568-4-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Make sure that arg->op.right is set properly

When process_op failed, @arg will be freed on a caller with type of
PRINT_OP.  Thus free_arg() will try to free ->op.right field which can
have stale value if something bad happens in the middle.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1348037924-17568-3-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Fix error path on process_array()

free_token() under out_free should be called with 'token' and no need
to set *tok to NULL since it's set already.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1348037924-17568-2-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: remove sscanf extension %as

perf uses sscanf extension %as to read and allocate a string in the same
step.  This is a non-standard extension only present in new versions of
glibc.

Replacing the use of sscanf and %as with strtok_r calls in order to
parse a given string into its components.  This is needed in Android
since bionic does not support
%as extension for sscanf.

Reviewed-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Tested-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Irina Tirdea <irina.tirdea@intel.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1348173470-4936-1-git-send-email-irina.tirdea@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf header: Remove perf_header__read_feature

Because its only user builtin-kvm::get_cpu_isa() has gone, It can be
removed safely.  In general, we have the feature information in
perf_session_env already, no need to read it again.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Dong Hao <haodong@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1348474503-15070-7-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf kvm: Use perf_session_env for reading cpuid

We have processed and saved cpuid information to perf_session_env so
reuse it for get_cpu_isa().

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Dong Hao <haodong@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1348474503-15070-6-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf header: Remove unused @feat arg from ->process callback

As the @feat arg is not used anywhere, get rid of it from the signature.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1348474503-15070-5-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf header: Use pre-processed session env when printing

From now on each feature information is processed and saved in perf
header so that it can be used for printing.  The event desc and branch
stack features are not touched since they're not saved.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1348474503-15070-4-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf header: Add ->process callbacks to most of features

From now on each feature information is processed and saved in perf
header so that it can be used wherever needed.  The BRANCH_STACK feature
is an exception since it needs nothing to be done.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1348474503-15070-3-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf header: Add struct perf_session_env

The struct perf_session_env will preserve environment information at the
time of perf record.  It can be accessed anytime after parsing a
perf.data file if needed.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@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/1348474503-15070-2-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf timechart: Use zalloc and fix a couple leaks

Use zalloc for the malloc+memset open coded sequence.

Fix leak on the #ifdef'ed C state handling and when detecting invalid
data in p_state_change().

Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf lock: Use perf_evsel__intval and perf_session__set_tracepoints_handlers

Following the model of 'perf sched':

. raw_field_value searches first on the common fields, that are unused
  in this tool

. Leave using perf_evsel__intval to the actual handlers, some may not
  need to incur some of the cost because they may not need all the
  fields values.

. Using perf_session__set_tracepoints_handlers will save all those
  strcmp to find the right handler at sample processing time, do it just
  once and get the handler from evsel->handler.func.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf kmem: Use perf_evsel__intval and perf_session__set_tracepoints_handlers

Following the model of 'perf sched':

. raw_field_value searches first on the common fields, that are unused
  in this tool

. Using perf_session__set_tracepoints_handlers will save all those
  strcmp to find the right handler at sample processing time, do it just
  once and get the handler from evsel->handler.func.

Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
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-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf kvm: Use perf_evsel__intval

Using plain raw_field_value(evsel->tp_format) will look at the common
fields as well, and since this tool doesn't need those, speed it up a
bit by looking at just the event specific fields.

Also in general use just evsel and sample, just like was done in 'perf
sched'.

v2: Fixed up test against evsel->name, that contains the subsys name
too, by David Ahern.

Cc: David Ahern <dsahern@gmail.com>
Cc: Dong Hao <haodong@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Runzhen Wang <runzhen@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/n/tip-v9x3q9rv4caxtox7wtjpchq5@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf kvm: Events analysis tool

Add 'perf kvm stat' support to analyze kvm vmexit/mmio/ioport smartly

Usage:
- kvm stat
  run a command and gather performance counter statistics, it is the alias of
  perf stat

- trace kvm events:
  perf kvm stat record, or, if other tracepoints are interesting as well, we
  can append the events like this:
  perf kvm stat record -e timer:* -a

  If many guests are running, we can track the specified guest by using -p or
  --pid, -a is used to track events generated by all guests.

- show the result:
  perf kvm stat report

The output example is following:
13005
13059

total 2 guests are running on the host

Then, track the guest whose pid is 13059:
^C[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.253 MB perf.data.guest (~11065 samples) ]

See the vmexit events:

Analyze events for all VCPUs:

             VM-EXIT    Samples  Samples%     Time%         Avg time

         APIC_ACCESS        460    70.55%     0.01%     22.44us ( +-   1.75% )
                 HLT         93    14.26%    99.98% 832077.26us ( +-  10.42% )
  EXTERNAL_INTERRUPT         64     9.82%     0.00%     35.35us ( +-  14.21% )
   PENDING_INTERRUPT         24     3.68%     0.00%      9.29us ( +-  31.39% )
           CR_ACCESS          7     1.07%     0.00%      8.12us ( +-   5.76% )
      IO_INSTRUCTION          3     0.46%     0.00%     18.00us ( +-  11.79% )
       EXCEPTION_NMI          1     0.15%     0.00%      5.83us ( +-   -nan% )

Total Samples:652, Total events handled time:77396109.80us.

See the mmio events:

Analyze events for all VCPUs:

         MMIO Access    Samples  Samples%     Time%         Avg time

        0xfee00380:W        387    84.31%    79.28%      8.29us ( +-   3.32% )
        0xfee00300:W         24     5.23%     9.96%     16.79us ( +-   1.97% )
        0xfee00300:R         24     5.23%     7.83%     13.20us ( +-   3.00% )
        0xfee00310:W         24     5.23%     2.93%      4.94us ( +-   3.84% )

Total Samples:459, Total events handled time:4044.59us.

See the ioport event:

Analyze events for all VCPUs:

      IO Port Access    Samples  Samples%     Time%         Avg time

         0xc050:POUT          3   100.00%   100.00%     13.75us ( +-  10.83% )

Total Samples:3, Total events handled time:41.26us.

And, --vcpu is used to track the specified vcpu and --key is used to sort the
result:

Analyze events for VCPU 0:

             VM-EXIT    Samples  Samples%     Time%         Avg time

                 HLT         27    13.85%    99.97% 405790.24us ( +-  12.70% )
  EXTERNAL_INTERRUPT         13     6.67%     0.00%     27.94us ( +-  22.26% )
         APIC_ACCESS        146    74.87%     0.03%     21.69us ( +-   2.91% )
      IO_INSTRUCTION          2     1.03%     0.00%     17.77us ( +-  20.56% )
           CR_ACCESS          2     1.03%     0.00%      8.55us ( +-   6.47% )
   PENDING_INTERRUPT          5     2.56%     0.00%      6.27us ( +-   3.94% )

Total Samples:195, Total events handled time:10959950.90us.

Signed-off-by: Dong Hao <haodong@linux.vnet.ibm.com>
Signed-off-by: Runzhen Wang <runzhen@linux.vnet.ibm.com>
[ Dong Hao <haodong@linux.vnet.ibm.com>
  Runzhen Wang <runzhen@linux.vnet.ibm.com>:
     - rebase it on current acme's tree
     - fix the compiling-error on i386 ]
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Avi Kivity <avi@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: kvm@vger.kernel.org
Cc: Runzhen Wang <runzhen@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1347870675-31495-4-git-send-email-haodong@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

KVM: x86: Export svm/vmx exit code and vector code to userspace

Exporting KVM exit information to userspace to be consumed by perf.

Signed-off-by: Dong Hao <haodong@linux.vnet.ibm.com>
[ Dong Hao <haodong@linux.vnet.ibm.com>: rebase it on acme's git tree ]
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Acked-by: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Avi Kivity <avi@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: kvm@vger.kernel.org
Cc: Runzhen Wang <runzhen@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/1347870675-31495-2-git-send-email-haodong@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Fix parallel build

Parallel builds of perf were failing for me on a 32p box, with:

    * new build flags or prefix
util/pmu.l:7:23: error: pmu-bison.h: No such file or directory

...

make: *** [util/pmu-flex.o] Error 1
make: *** Waiting for unfinished jobs....

This can pretty quickly be seen by adding a sleep in front of the bison
calls in tools/perf/Makefile and running make -j4 on a smaller box i.e.:

sleep 10; $(QUIET_BISON)$(BISON) -v util/pmu.y -d -o $(OUTPUT)util/pmu-bison.c

Adding the following dependencies fixes it for me.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Namhyung Kim <namhyung@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: http://lkml.kernel.org/r/505BD190.40707@redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Print event causing perf_event_open() to fail

Got tired of not getting the event that caused the perf_event_open()
syscall to fail. So I fixed the error message. This is very useful when
monitoring lots of events in a single run.

Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20120920161945.GA7064@quad
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

kprobes/x86: Move skip_singlestep up

I get this warning:

  arch/x86/kernel/kprobes.c:544:23: warning: ‘skip_singlestep’ declared ‘static’ but never defined

on tip/auto-latest.

Put the skip_singlestep function declaration up, in
KPROBES_CAN_USE_FTRACE and drop the superfluous forward
declaration.

Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1348145034-16603-1-git-send-email-bp@amd64.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf tools: Fix a compiling error in util/map.c

This patch fix a compile warning taken as error:

CC util/map.o
cc1: warnings being treated as errors
util/map.c: In function ‘map__fprintf_dsoname’:
util/map.c:240: error: ‘dsoname’ may be used uninitialized in this function
make: *** [util/map.o] Error 1

Signed-off-by: Feng Tang <feng.tang@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1346053107-11946-3-git-send-email-feng.tang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Fix a compiling error in trace-event-perl.c for 32 bits machine

On my x86_32 mahcine, there is a compile error:

        CC util/scripting-engines/trace-event-perl.o
        cc1: warnings being treated as errors
        util/scripting-engines/trace-event-perl.c: In function
perl_process_tracepoint:
        util/scripting-engines/trace-event-perl.c:285: error: format
expects type 'int', but argument 2 has type '__u64'
        make: *** [util/scripting-engines/trace-event-perl.o] Error 1

Fix it by using the "%PRIu64" for __u64.

v2: use PRIu64 as suggested by Arnaldo.

Signed-off-by: Feng Tang <feng.tang@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/20120828101730.6b2fd97e@feng-i7
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf: Fix off by one test in perf_reg_value()

The test should be >= ARRAY_SIZE() instead of > ARRAY_SIZE().

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Link: http://lkml.kernel.org/r/20120905123126.GC6128@elgon.mountain
Signed-off-by: Ingo Molnar <mingo@kernel.org>

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

Pull uprobes fixes + cleanups from Oleg Nesterov.

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

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

Pull perf/core improvements and fixes from  Arnaldo Carvalho de Melo:

 * Fix handling of unresolved samples when --symbols is used in 'report',
   from Feng Tang.

 * Add --symbols to 'script', similar to the one in 'report', from Feng Tang.

 * Add union member access support to 'probe', from Hyeoncheol Lee.

 * Make 'archive' work on Android, tweaking some of the utility parameters
   used (tar, rm), from Irina Tirdea.

 * Fixups to die() removal, from Namhyung Kim.

 * Render fixes for the TUI, from Namhyung Kim.

 * Don't enable annotation in non symbolic view, from Namhyung Kim.

 * Fix pipe mode in 'report', from Namhyung Kim.

 * Move related stats code from stat to util/, will be used by the 'stat'
   kvm tool, from Xiao Guangrong.

 * Add cpumask for uncore pmu, use it in 'stat', from Yan, Zheng.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf stat: Check PMU cpumask file

If user doesn't explicitly specify CPU list, perf-stat only collects
events on CPUs listed in the PMU cpumask file.

Signed-off-by: "Yah, Zheng" <zheng.z.yan@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1347263631-23175-3-git-send-email-zheng.z.yan@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf/x86: Add cpumask for uncore pmu

This patch adds a cpumask file to the uncore pmu sysfs directory.  The
cpumask file contains one active cpu for every socket.

Signed-off-by: "Yan, Zheng" <zheng.z.yan@intel.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Ingo Molnar <mingo@kernel.org>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: "Yan, Zheng" <zheng.z.yan@intel.com>
Link: http://lkml.kernel.org/r/1347263631-23175-2-git-send-email-zheng.z.yan@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf report: Add missing perf_hpp__init for pipe-mode

The perf_hpp__init() function was only called from setup_browser() so
that the pipe-mode missed the initialization thus didn't respond to
related options.  Fix it.

Reported-by: Robert Richter <robert.richter@amd.com>
Tested-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-tip-commits@vger.kernel.org
Link: http://lkml.kernel.org/r/87txv28spl.fsf_-_@sejong.aot.lge.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts: Export a find_scripts() function

So that other perf commands/browser has a way to dig out the available
scripts info in system, this is a preparation for the script browser.

Signed-off-by: Feng Tang <feng.tang@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347007349-3102-5-git-send-email-feng.tang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts: Add event_analyzing_sample-record/report

So that event_analyzing_sample.py can be shown by "perf script -l"

Signed-off-by: Feng Tang <feng.tang@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347007349-3102-4-git-send-email-feng.tang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts: Add --symbols option to handle specific symbols

Since perf script no longer only handle the trace points, we can add the
symbol filter option so that scripts can handle specified samples.

Signed-off-by: Feng Tang <feng.tang@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347007349-3102-3-git-send-email-feng.tang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf symbols: Filter samples with unresolved symbol when "--symbols" option is used

Report/top commands support to only handle specific symbols with
"--symbols" option, but current code will keep those samples whose
symbol can't be resolved, which should actually be filtered.

If we run following commands:
$perf record -a tree
$perf report --symbols intel_idle -n
the output will be:

Without the patch:
==================
    46.27%        156     sshd  [unknown]
    26.05%         48  swapper  [kernel.kallsyms]
    17.26%         38     tree  libc-2.12.1.so
     7.69%         17     tree  tree
     2.73%          6     tree  ld-2.12.1.so

With the patch:
===============
   100.00%         48  swapper  [kernel.kallsyms]

Signed-off-by: Feng Tang <feng.tang@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347007349-3102-2-git-send-email-feng.tang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf archive: Make 'f' the last parameter for tar

On some systems, tar needs to specify the name of the archive immediately
after the -f parameter.

Change the order of the parameters so tar can run properly.

Signed-off-by: Irina Tirdea <irina.tirdea@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1347574063-22521-5-git-send-email-irina.tirdea@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf archive: Remove -f from the rm command

In Android, rm does not support the -f parameter.

Remove -f from rm and make sure rm does not fail even if the files to be
removed are not found.

Signed-off-by: Irina Tirdea <irina.tirdea@intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Namhyung Kim <namhyung.kim@lge.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1347574063-22521-4-git-send-email-irina.tirdea@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf stat: Move stats related code to util/stat.c

Then, the code can be shared between kvm events and perf stat.

Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
[ Dong Hao <haodong@linux.vnet.ibm.com>: rebase it on acme's git tree ]
Signed-off-by: Dong Hao <haodong@linux.vnet.ibm.com>
Cc: Avi Kivity <avi@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: kvm@vger.kernel.org
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Runzhen Wang <runzhen@linux.vnet.ibm.com>
Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com
Link: http://lkml.kernel.org/r/1347870675-31495-3-git-send-email-haodong@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf report: Enable integrated annotation only if possible

The integrated annotation feature is supported only in TUI mode.  Also
it should be enabled with 'symbol' sort key otherwise resulting hist
entry doesn't need to have same symbol as of a sample so that it can
fail on hist_entry__inc_addr_samples with -ERANGE.

You can easily see it when start perf report TUI without symbol* sort
key.  This patch fixes the problem.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347611729-16994-2-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add sort__has_sym

The sort__has_sym variable is for checking whether the sort_list
includes 'symbol' as a sort key.  It will be used for later patch.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347611729-16994-1-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

uprobes: Make arch_uprobe_task->saved_trap_nr "unsigned int"

Make arch_uprobe_task->saved_trap_nr "unsigned int" and move it down
after ->saved_scratch_register, this changes sizeof() from 24 to 16.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes/x86: Fix arch_uprobe_disable_step() && UTASK_SSTEP_TRAPPED interaction

arch_uprobe_disable_step() should also take UTASK_SSTEP_TRAPPED into
account. In this case the probed insn was not executed, we need to
clear X86_EFLAGS_TF if it was set by us and that is all.

Again, this code will look more clean when we move it into
arch_uprobe_post_xol() and arch_uprobe_abort_xol().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes/x86: Xol should send SIGTRAP if X86_EFLAGS_TF was set

arch_uprobe_disable_step() correctly preserves X86_EFLAGS_TF and
returns to user-mode. But this means the application gets SIGTRAP
only after the next insn.

This means that UPROBE_CLEAR_TF logic is not really right. _enable
should only record the state of X86_EFLAGS_TF, and _disable should
check it separately from UPROBE_FIX_SETF.

Remove arch_uprobe_task->restore_flags, add ->saved_tf instead, and
change enable/disable accordingly. This assumes that the probed insn
was not trapped, see the next patch.

arch_uprobe_skip_sstep() logic has the same problem, change it to
check X86_EFLAGS_TF and send SIGTRAP as well. We will cleanup this
all after we fold enable/disable_step into pre/post_hol hooks.

Note: send_sig(SIGTRAP) is not actually right, we need send_sigtrap().
But this needs more changes, handle_swbp() does the same and this is
equally wrong.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes/x86: Do not (ab)use TIF_SINGLESTEP/user_*_single_step() for single-stepping

user_enable/disable_single_step() was designed for ptrace, it assumes
a single user and does unnecessary and wrong things for uprobes. For
example:

- arch_uprobe_enable_step() can't trust TIF_SINGLESTEP, an
  application itself can set X86_EFLAGS_TF which must be
  preserved after arch_uprobe_disable_step().

- we do not want to set TIF_SINGLESTEP/TIF_FORCED_TF in
  arch_uprobe_enable_step(), this only makes sense for ptrace.

- otoh we leak TIF_SINGLESTEP if arch_uprobe_disable_step()
  doesn't do user_disable_single_step(), the application will
  be killed after the next syscall.

- arch_uprobe_enable_step() does access_process_vm() we do
  not need/want.

Change arch_uprobe_enable/disable_step() to set/clear X86_EFLAGS_TF
directly, this is much simpler and more correct. However, we need to
clear TIF_BLOCKSTEP/DEBUGCTLMSR_BTF before executing the probed insn,
add set_task_blockstep(false).

Note: with or without this patch, there is another (hopefully minor)
problem. A probed "pushf" insn can see the wrong X86_EFLAGS_TF set by
uprobes. Perhaps we should change _disable to update the stack, or
teach arch_uprobe_skip_sstep() to emulate this insn.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

ptrace/x86: Partly fix set_task_blockstep()->update_debugctlmsr() logic

Afaics the usage of update_debugctlmsr() and TIF_BLOCKSTEP in
step.c was always very wrong.

1. update_debugctlmsr() was simply unneeded. The child sleeps
   TASK_TRACED, __switch_to_xtra(next_p => child) should notice
   TIF_BLOCKSTEP and set/clear DEBUGCTLMSR_BTF after resume if
   needed.

2. It is wrong. The state of DEBUGCTLMSR_BTF bit in CPU register
   should always match the state of current's TIF_BLOCKSTEP bit.

3. Even get_debugctlmsr() + update_debugctlmsr() itself does not
   look right. Irq can change other bits in MSR_IA32_DEBUGCTLMSR
   register or the caller can be preempted in between.

4. It is not safe to play with TIF_BLOCKSTEP if task != current.
   DEBUGCTLMSR_BTF and TIF_BLOCKSTEP should always match each
   other if the task is running. The tracee is stopped but it
   can be SIGKILL'ed right before set/clear_tsk_thread_flag().

However, now that uprobes uses user_enable_single_step(current)
we can't simply remove update_debugctlmsr(). So this patch adds
the additional "task == current" check and disables irqs to avoid
the race with interrupts/preemption.

Unfortunately this patch doesn't solve the last problem, we need
another fix. Probably we should teach ptrace_stop() to set/clear
single/block stepping after resume.

And afaics there is yet another problem: perf can play with
MSR_IA32_DEBUGCTLMSR from nmi, this obviously means that even
__switch_to_xtra() has problems.

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

ptrace/x86: Introduce set_task_blockstep() helper

No functional changes, preparation for the next fix and for uprobes
single-step fixes.

Move the code playing with TIF_BLOCKSTEP/DEBUGCTLMSR_BTF into the
new helper, set_task_blockstep().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes/x86: Implement x86 specific arch_uprobe_*_step

The arch specific implementation behaves like user_enable_single_step()
except that it does not disable single stepping if it was already
enabled by ptrace. This allows the debugger to single step over an
uprobe. The state of block stepping is not restored. It makes only sense
together with TF and if that was enabled then the debugger is notified.

Note: this is still not correct. For example, TIF_SINGLESTEP check
is not right, the application itself can set X86_EFLAGS_TF. And otoh
we leak TIF_SINGLESTEP (set by enable) if the probed insn is "popf".
See the next patches, we need the changes in arch/x86/kernel/step.c
first.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes: Introduce arch_uprobe_enable/disable_step()

As Oleg pointed out in [0] uprobe should not use the ptrace interface
for enabling/disabling single stepping.

[0] http://lkml.kernel.org/r/20120730141638.GA5306@redhat.com

Add the new "__weak arch" helpers which simply call user_*_single_step()
as a preparation. This is only needed to not break the powerpc port, we
will fold this logic into arch_uprobe_pre/post_xol() hooks later.

We should also change handle_singlestep(), _disable_step(&uprobe->arch)
should be called before put_uprobe().

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes: Teach find_active_uprobe() to clear MMF_HAS_UPROBES

The wrong MMF_HAS_UPROBES doesn't really hurt, just it triggers
the "slow" and unnecessary handle_swbp() path if the task hits
the non-uprobe breakpoint.

So this patch changes find_active_uprobe() to check every valid
vma and clear MMF_HAS_UPROBES if no uprobes were found. This is
adds the slow O(n) path, but it is only called in unlikely case
when the task hits the normal breakpoint first time after
uprobe_unregister().

Note the "not strictly accurate" comment in mmf_recalc_uprobes().
We can fix this, we only need to teach vma_has_uprobes() to return
a bit more more info, but I am not sure this worth the trouble.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes: Introduce MMF_RECALC_UPROBES

Add the new MMF_RECALC_UPROBES flag, it means that MMF_HAS_UPROBES
can be false positive after remove_breakpoint() or uprobe_munmap().
It is also set by uprobe_dup_mmap(), this is not optimal but simple.
We could add the new hook, uprobe_dup_vma(), to set MMF_HAS_UPROBES
only if the new mm actually has uprobes, but I don't think this
makes sense.

The next patch will use this flag to clear MMF_HAS_UPROBES.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes: uprobes_treelock should not disable irqs

Nobody plays with uprobes_tree/uprobes_treelock in interrupt context,
no need to disable irqs.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

uprobes: Don't put NULL pointer in uprobe_register()

alloc_uprobe() might return a NULL pointer, put_uprobe() can't deal with
this.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>

perf hists browser: Fix first column printing

As a side effect of commit f5951d56a2ab ("perf hists browser: Use
perf_hpp__format functions") perf report TUI got a problem of not
refreshing the first character.

Since the previous patch restores the column width of "overhead" to 7
we can start at column 0 now.

Reported-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347431706-7839-2-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf hists browser: Fix output for 100.00%

Current hpp format functions assume that the output will fit to 6
character including % sign (XX.YY%) so used "%5.2f%%" as a format
string.  However it might be the case if collapsing resulted in a single
entry which has 100.00% (7 character) of period. In this case the output
will be shifted by 1 character.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347431706-7839-1-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools lib traceevent: Define _GNU_SOURCE in Makefile

For the reasons stated on:

  commit 0a84f00
  Author: David Daney <david.daney@cavium.com>

    perf tools: Fix broken build by defining _GNU_SOURCE in Makefile

Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: David Daney <david.daney@cavium.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-e2nofbmj4uf0ykgsytxvt9pu@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf sched: Fixup for the die() removal

The commit a116e05dcf61 ("perf sched: Remove die() calls") replaced
die() call to pr_debug + return -1, but it should be pr_err otherwise
it'll not show up unless -v option is given.  Fix it.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347415866-303-2-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test: Fixup for the die() removal

The commit 32c7f7383a09 ("perf test: Remove die() calls") replaced die()
call to pr_debug + return -1, but it should be pr_err otherwise it'll
not show up unless -v option is given.  Fix it.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1347415866-303-1-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf probe: Add union member access support

Union members can be accessed with '.' or '->' like data structure
member access

Signed-off-by: Hyunchul Lee <hyc.lee@gmail.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/CANFS6baeuSBxPGQ8SUZWZErJ2bWs-Nojg+FSo138E1QK8bJJig@mail.gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge branch 'tip/perf/core' of git://git./linux/kernel/git/rostedt/linux-trace into perf/core

Pull tracing updates from Steve Rostedt.

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

kprobes/x86: Fix to support jprobes on ftrace-based kprobe

Fix kprobes/x86 to support jprobes on ftrace-based kprobes.
Because of -mfentry support of ftrace, ftrace is now put
on the beginning of function where jprobes are put.

Originally ftrace-based kprobes doesn't support jprobe
because it will change regs->ip and ftrace doesn't support
changing IP and ftrace itself doesn't conflict jprobe.
However, ftrace -mfentry support moves mcount call on the
top of functions where jprobes are put. This means that
jprobe always conflicts with ftrace-based kprobe and fails.

This patch allows ftrace-based kprobes to support jprobes
by allowing to modify regs->ip and kprobes breakpoint
handler also allows to skip singlestepping because there
is a ftrace call (not an original instruction).

Link: http://lkml.kernel.org/r/20120905143125.10329.90836.stgit@localhost.localdomain
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

ftrace/x86-64: Allow to change RIP in handlers

Allow ftrace handlers to change RIP register (regs->ip)
in handlers. This will allow handlers to call another
function instead of original function.

Link: http://lkml.kernel.org/r/20120905143118.10329.5078.stgit@localhost.localdomain
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

kprobes/x86: Fix kprobes to collectly handle IP on ftrace

Current kprobe_ftrace_handler expects regs->ip == ip, but it is
incorrect (originally on x86-64). Actually, ftrace handler sets
regs->ip = ip + MCOUNT_INSN_SIZE.
kprobe_ftrace_handler must take care for that.

Link: http://lkml.kernel.org/r/20120905143112.10329.72069.stgit@localhost.localdomain
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

ftrace/x86: Adjust x86 regs.ip as like as x86-64

Adjust x86 regs.ip to ip + MCOUNT_INSN_SIZE as like as
on x86-64. This helps us to consolidate codes which use
regs->ip on both of x86/x86-64.

Link: http://lkml.kernel.org/r/20120905143100.10329.60109.stgit@localhost.localdomain
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

trace: Stop compiling in trace_clock unconditionally

Commit 56449f437 "tracing: make the trace clocks available generally",
in April 2009, made trace_clock available unconditionally, since
CONFIG_X86_DS used it too.

Commit faa4602e47 "x86, perf, bts, mm: Delete the never used BTS-ptrace code",
in March 2010, removed CONFIG_X86_DS, and now only CONFIG_RING_BUFFER (split
out from CONFIG_TRACING for general use) has a dependency on trace_clock. So,
only compile in trace_clock with CONFIG_RING_BUFFER or CONFIG_TRACING
enabled.

Link: http://lkml.kernel.org/r/20120903024513.GA19583@leaf
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>