perf list: Display metric expressions for --details option

Display metric expression itself when --details is specified.

Current list with no details:

  # perf list metrics
  ...
  TopDownL1:
    IPC
         [Instructions Per Cycle (per logical thread)]
    SLOTS
         [Total issue-pipeline slots]
  ...

Detailed output with metric formula:

  # perf list --details metrics
  ...
  TopDownL1:
    IPC
         [Instructions Per Cycle (per logical thread)]
         [inst_retired.any / cpu_clk_unhalted.thread]
    SLOTS
         [Total issue-pipeline slots]
         [4*(( cpu_clk_unhalted.thread_any / 2 ) if #smt_on else cycles)]
  ...

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190213123246.4015-6-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Fix legacy events symbol separator parsing

Fixing legacy symbol events parsing. We can't support single slash
separator, like 'cycles/u', because it conflicts with non empty terms,
like 'cycles/period/u'.

Keeping only '//' and ':' separator for these events:
  cycles//u
  cycles:k

And removing '/' separator support, which is not working
anymore. Also adding automated tests for above events.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190213123246.4015-5-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Rename build libperf to perf

Rename build libperf to perf, because it's used to build perf.

The libperf build object name will be used for libperf library.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190213123246.4015-4-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Rename LIB_FILE to LIBPERF_A

Simple rename, no functional change.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190213123246.4015-3-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Compile perf with libperf-in.o instead of libperf.a

There's no need for perf build to use libperf.a,
we can use directly libperf-in.o.

The libperf.a stays as a target if needed:

  $ make libperf.a
  ...
    CC       util/pmu.o
    CC       util/pmu-flex.o
    LD       util/libperf-in.o
    LD       libperf-in.o
    AR       libperf.a

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190213123246.4015-2-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Modularize auxtrace_buffer fetch function

Making the auxtrace_buffer fetch function modular so that it can be
called from different decoding context (timeless vs. non-timeless),
avoiding to repeat code.

No change in functionality is introduced by this patch.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-14-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Modularize main packet processing loop

Making the main packet processing loop modular so that it can be called
from different decoding context (timeless vs. non-timless), avoiding to
repeat code.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-13-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Modularize main decoder function

Making the main decoder block modular so that it can be called from
different decoding context (timeless vs. non-timeless), avoiding
to repeat code.

No change in functionality is introduced by this patch.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-12-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Make cs_etm__run_decoder() queue independent

This patch makes decoding of auxtrace buffer centered around a struct
cs_etm_queue.  This eliminates surperflous variables and is a precursor
for work that simplifies the main decoder loop.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-11-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Rethink kernel address initialisation

Moving initialisation of the kernel start address to function
cs_etm__setup_queues(), considered to be the common denominator for
queue initialisation.  That way we don't have to repeat the same code
at different places.

No change of functionatlity is introduced by this patch.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-10-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Cleaning up function cs_etm__alloc_queue()

Function cs_etm__alloc_queue() should only be concerned with the allocation
of memory for the etmq and accompanying decoder.  Everything else should
be done in the calling function.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-9-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Fix erroneous comment

The comment just before initialising the decoder is plane wrong since it
is part of the decoding queue setup function and the operation code
specifically mention that trace data is to be decoded rather than printed
out.

This patch simply fix the comment to prevent people from getting really
confused.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-8-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Introducing function cs_etm__init_trace_params()

The trace parameter initialisation code is repeated in two different
places, something that bloats the file and can lead to errors.  This
is fixed by introducing a helper function and calling the right
protocol initialisation code when required.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-7-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Fix memory leak in error path

Memory allocated for variable 't_params' isn't released properly in the
error path of function cs_etm_queue *cs_etm__alloc_queue() and
cs_etm__dump_event(), something this patch addresses.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-6-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Introducing function cs_etm_decoder__init_dparams()

Introducing function cs_etm_decoder__init_dparams() to avoid repeating
code at two different places.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-5-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Fix wrong return values in error path

Function cs_etm__mem_access() is supposed to return a u32 but the error
path returns negative values at a couple of places, something that really
throws off the clients using it.  Fix the situation by return '0'.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-4-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Remove unused structure field "time" and "timestamp"

Field "time" and "timestamp" in structure cs_etm_queue are no longer
used and need to be removed.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-3-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Remove unused structure field "state"

Field "state" in structure cs_etm_queue is no longer used and needs
to be removed.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190212171618.25355-2-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf build: Add missing FEATURE_CHECK_LDFLAGS-libcrypto

When the libcrypto feature test was added we forgot to add its
FEATURE_CHECK_LDFLAGS pointing to the library needed to link with the
test-all.bin feature test fast path binary, so even when it was
introduced we got this:

  $ cat /tmp/build/perf/feature/test-all.make.output
  /usr/bin/ld: /tmp/ccjKeJJU.o: in function `main_test_libcrypto':
  /home/acme/git/perf/tools/build/feature/test-libcrypto.c:10: undefined reference to `MD5_Init'
  /usr/bin/ld: /home/acme/git/perf/tools/build/feature/test-libcrypto.c:11: undefined reference to `MD5_Update'
  /usr/bin/ld: /home/acme/git/perf/tools/build/feature/test-libcrypto.c:12: undefined reference to `MD5_Final'
  /usr/bin/ld: /home/acme/git/perf/tools/build/feature/test-libcrypto.c:14: undefined reference to `SHA1'
  collect2: error: ld returned 1 exit status
  $ cat /tmp/build/perf/feature/test-libcrypto.
  test-libcrypto.bin          test-libcrypto.d            test-libcrypto.make.output
  $ cat /tmp/build/perf/feature/test-libcrypto.make.output
  $

Fix it, so that we keep the fast path, which, at this point, will fail
with the unwind-ARCH feature tests, that will be fixed in a followup
patch:

  $ make -C tools/perf O=/tmp/build/perf
  ...                     libcrypto: [ on  ]
   <SNIP>
  $ cat /tmp/build/perf/feature/test-all.make.output
  $ ldd /tmp/build/perf/feature/test-all.bin | grep libcrypto
libcrypto.so.1.1 => /lib64/libcrypto.so.1.1 (0x00007f9892805000)
  $
  $ grep libcrypto /tmp/build/perf/FEATURE-DUMP
  feature-libcrypto=1
  $

With the unwind-ARCH tests fixed, we now finally manage to get
test-all.bin built and linked with the features it tests, among them the
ones fixed in this patchkit:

  $ ldd /tmp/build/perf/feature/test-all.bin  | egrep 'unwind|crypto'
libcrypto.so.1.1 => /lib64/libcrypto.so.1.1 (0x00007f95cf2b8000)
libunwind-x86_64.so.8 => /lib64/libunwind-x86_64.so.8 (0x00007f95cf294000)
libunwind.so.8 => /lib64/libunwind.so.8 (0x00007f95cf278000)
  $

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: John McCutchan <johnmccutchan@google.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sonny Rao <sonnyrao@chromium.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Fixes: 8ee4646038e4 ("perf build: Add libcrypto feature detection")
Link: https://lkml.kernel.org/n/tip-rexc248jorf5b4l3qjn888cz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools build: Add test-reallocarray.c to test-all.c to fix the build

When a test is in the FEATURE_TESTS_BASIC list in tools/build/Makefile.feature
must be added to tools/build/feature/test-all.c, because the successfull
compilation and linking of that test-all.bin file means that all the
features listed in FEATURE_TESTS_BASIC are present in the system, so we
don't have to go on feature by feature test building them.

Since reallocarray() is expected to be present in modern systems, it has
a place in FEATURE_TESTS_BASIC, so that we speed up the build process
building just that file.

For older systems, such as ubuntu:16.04 (build failure reported by Jin
Yao) debian:8, and for the current flagship RHEL distro, RHEL7, the
build will fail as test-all.bin (without test-reallocarray.c included)
passes but reallocarray() isn't present, making the build fail with:

    CC       /tmp/build/perf/libbpf.o
    MKDIR    /tmp/build/perf/fs/
    CC       /tmp/build/perf/fs/tracing_path.o
    LD       /tmp/build/perf/fd/libapi-in.o
    CC       /tmp/build/perf/bpf.o
  libbpf.c: In function 'bpf_object__add_program':
  libbpf.c:367:10: error: implicit declaration of function 'reallocarray' [-Werror=implicit-function-declaration]
    progs = reallocarray(progs, nr_progs + 1, sizeof(progs[0]));
            ^
  libbpf.c:367:2: error: nested extern declaration of 'reallocarray' [-Werror=nested-externs]
    progs = reallocarray(progs, nr_progs + 1, sizeof(progs[0]));
    ^
  libbpf.c:367:8: error: assignment makes pointer from integer without a cast [-Werror=int-conversion]
    progs = reallocarray(progs, nr_progs + 1, sizeof(progs[0]));
          ^
  libbpf.c: In function 'bpf_object__elf_collect':
  libbpf.c:887:10: error: assignment makes pointer from integer without a cast [-Werror=int-conversion]
      reloc = reallocarray(reloc, nr_reloc,
            ^
  libbpf.c: In function 'bpf_program__reloc_text':
  libbpf.c:1394:12: error: assignment makes pointer from integer without a cast [-Werror=int-conversion]
     new_insn = reallocarray(prog->insns, new_cnt, sizeof(*insn));
              ^
    CC       /tmp/build/perf/nlattr.o

Even with:

  $ grep reallocarray /tmp/build/perf/FEATURE-DUMP
  feature-reallocarray=1
  $

Which ubuntu:16.04.5 LTS doesn't have:

  perfbuilder@38a153a1bba8:/$ head -2 /etc/os-release
  NAME="Ubuntu"
  VERSION="16.04.5 LTS (Xenial Xerus)"
  perfbuilder@38a153a1bba8:/$ find /usr/include/ -name "*.h" | xargs grep -w reallocarray
  perfbuilder@38a153a1bba8:/$

Fix it by including it to test-all.c, which ends up forcing the
individual tests to be triggered and for the build process to notice
that indeed reallocarray() is not there:

  perfbuilder@38a153a1bba8:/$ cat /tmp/build/perf/feature/test-all.make.output
  In file included from test-all.c:178:0:
  test-reallocarray.c: In function 'main_test_reallocarray':
  test-reallocarray.c:7:11: error: implicit declaration of function 'reallocarray' [-Werror=implicit-function-declaration]
    return !!reallocarray(NULL, 1, 1);
             ^
  cc1: all warnings being treated as errors
  perfbuilder@38a153a1bba8:/$

That is the only test that is failing on Ubuntu 16.03.5 LTS, so all
tests are forced:

  perfbuilder@38a153a1bba8:/tmp/build/perf/feature$ ls -lSr *.make.output
  <SNIP successful tests>
  -rw-r--r--. 1 perfbuilder perfbuilder   0 Feb 14 15:00 test-dwarf.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder   0 Feb 14 14:16 test-cplus-demangle.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder   0 Feb 14 15:00 test-bpf.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder   0 Feb 14 15:00 test-backtrace.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder 104 Feb 14 15:00 test-bionic.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder 107 Feb 14 15:00 test-libunwind-x86.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder 115 Feb 14 15:00 test-libunwind-aarch64.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder 122 Feb 14 15:00 test-libbabeltrace.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder 254 Feb 14 15:00 test-reallocarray.make.output
  -rw-r--r--. 1 perfbuilder perfbuilder 312 Feb 14 15:00 test-all.make.output
  perfbuilder@38a153a1bba8:/tmp/build/perf/feature$

And that reallocarray() one shows:

  perfbuilder@38a153a1bba8:/tmp/build/perf/feature$ cat test-reallocarray.make.output
  test-reallocarray.c: In function 'main':
  test-reallocarray.c:7:11: error: implicit declaration of function 'reallocarray' [-Werror=implicit-function-declaration]
    return !!reallocarray(NULL, 1, 1);
             ^
  cc1: all warnings being treated as errors
  perfbuilder@38a153a1bba8:/tmp/build/perf/feature$

Which now generates the expected result:

  perfbuilder@38a153a1bba8:~$ grep reallocarray /tmp/build/perf/FEATURE-DUMP
  feature-reallocarray=0
  perfbuilder@38a153a1bba8:~$

The fallback mechanism kicks in and libbpf and perf are again buildable
in systems without reallocarray():

  $ cat tools/include/tools/libc_compat.h
  // SPDX-License-Identifier: (LGPL-2.0+ OR BSD-2-Clause)
  /* Copyright (C) 2018 Netronome Systems, Inc. */

  #ifndef __TOOLS_LIBC_COMPAT_H
  #define __TOOLS_LIBC_COMPAT_H

  #include <stdlib.h>
  #include <linux/overflow.h>

  #ifdef COMPAT_NEED_REALLOCARRAY
  static inline void *reallocarray(void *ptr, size_t nmemb, size_t size)
  {
  size_t bytes;

  if (unlikely(check_mul_overflow(nmemb, size, &bytes)))
  return NULL;
  return realloc(ptr, bytes);
  }
  #endif
  #endif
  $

Reported-by: Jin Yao <yao.jin@linux.intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexei Starovoitov <ast@fb.com>
Cc: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jakub Kicinski <jakub.kicinski@netronome.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Fixes: 531b014e7a2f ("tools: bpf: make use of reallocarray")
Link: https://lkml.kernel.org/n/tip-aonqku8axii8rxki5g11w40b@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf unwind: Do not put libunwind-{x86,aarch64} in FEATURE_TESTS_BASIC

As it is not normally available on x86_64 not being tested on test-all.c
but being in FEATURE_TESTS_BASIC ends up implying that those features
are present, which leads to trying to link with those libraries and a
build failure now that test-all.c is finally again building
successfully:

  /usr/bin/ld: cannot find -lunwind-x86
  /usr/bin/ld: cannot find -lunwind-aarch64
  collect2: error: ld returned 1 exit status
  make[3]: *** [Makefile:199: /tmp/build/perf/plugin_jbd2.so] Error 1
  make[3]: *** Waiting for unfinished jobs....
  /usr/bin/ld: cannot find -lunwind-x86
  /usr/bin/ld: cannot find -lunwind-aarch64

So remove those features from there and explicitely test them.

And then move this patch to just before the last one that allows this to
be exposed, so that we keep the tree bisectable.

With all this in place we get, at this point:

  $ ldd /tmp/build/perf/feature/test-libunwind.bin
linux-vdso.so.1 (0x00007fffa09c6000)
libunwind-x86_64.so.8 => /lib64/libunwind-x86_64.so.8 (0x00007fbcf4451000)
libunwind.so.8 => /lib64/libunwind.so.8 (0x00007fbcf4435000)
liblzma.so.5 => /lib64/liblzma.so.5 (0x00007fbcf440c000)
libelf.so.1 => /lib64/libelf.so.1 (0x00007fbcf43f2000)
libc.so.6 => /lib64/libc.so.6 (0x00007fbcf422c000)
libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007fbcf4211000)
/lib64/ld-linux-x86-64.so.2 (0x00007fbcf4491000)
libpthread.so.0 => /lib64/libpthread.so.0 (0x00007fbcf41ed000)
libz.so.1 => /lib64/libz.so.1 (0x00007fbcf41d3000)
  $ cat /tmp/build/perf/feature/test-libunwind-x86.make.output
  test-libunwind-x86.c:2:10: fatal error: libunwind-x86.h: No such file or directory
   #include <libunwind-x86.h>
            ^~~~~~~~~~~~~~~~~
  compilation terminated.
  $ cat /tmp/build/perf/feature/test-libunwind-aarch64.make.output
  test-libunwind-aarch64.c:2:10: fatal error: libunwind-aarch64.h: No such file or directory
  #include <libunwind-aarch64.h>
           ^~~~~~~~~~~~~~~~~~~~~
  compilation terminated.
  $
  $ ldd ~/bin/perf | grep unwind
libunwind-x86_64.so.8 => /lib64/libunwind-x86_64.so.8 (0x00007f5ceb24b000)
libunwind.so.8 => /lib64/libunwind.so.8 (0x00007f5ceb22f000)
  $

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jean Pihet <jean.pihet@linaro.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Cc: Will Deacon <will.deacon@arm.com>
Link: https://lkml.kernel.org/n/tip-vs6kwqsvwk7oxhs6z9mq87pp@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf coresight: Do not test for libopencsd by default

Since it is not yet that generally available, avoid testing for the
presence of libcoresight in the fast path test-all.bin feature test.

  # dnf search opencsd
  No matches found.
  # dnf search OpenCSD
  No matches found.
  # cat /etc/fedora-release
  Fedora release 29 (Twenty Nine)
  #

I.e. right now, in my system test-all.bin is failing all the time since
Fedora29 doesn't have libopencsd available:

  $ cat /tmp/build/perf/feature/test-all.make.output
  In file included from test-all.c:174:
  test-libopencsd.c:2:10: fatal error: opencsd/c_api/opencsd_c_api.h: No such file or directory
   #include <opencsd/c_api/opencsd_c_api.h>
            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  compilation terminated.

See:

  6ab2b762befd ("perf build: Disable libbabeltrace check by default")

For the rationale, as soon as libopencsd becomes more generally packaged
and available, we do the same thing we did with babeltrace, enabling it
by default, as done in:

  24787afbcd01 ("perf tools: Enable LIBBABELTRACE by default")

For now, to explicitely ask for opencsd, make sure you have it installed
and use:

   make -C tools/perf CORESIGHT=1

The feature test output will be there as an empty file:

  $ ls -la /tmp/build/perf/feature/test-libopencsd.make.output

Because the binary used for the feature check was successfully built:

  $ ls -la /tmp/build/perf/feature/test-libopencsd.bin
  -rwxrwxr-x. 1 acme acme 18336 Feb 12 14:49 /tmp/build/perf/feature/test-libopencsd.bin
  $ ldd /tmp/build/perf/feature/test-libopencsd.bin
linux-vdso.so.1 (0x00007fffe18cc000)
libopencsd_c_api.so.0 => /lib64/libopencsd_c_api.so.0 (0x00007fb8e67f6000)
libopencsd.so.0 => /lib64/libopencsd.so.0 (0x00007fb8e676f000)
libc.so.6 => /lib64/libc.so.6 (0x00007fb8e65a9000)
libstdc++.so.6 => /lib64/libstdc++.so.6 (0x00007fb8e6411000)
libm.so.6 => /lib64/libm.so.6 (0x00007fb8e628d000)
libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x00007fb8e6272000)
/lib64/ld-linux-x86-64.so.2 (0x00007fb8e6828000)
  $

And the resulting perf binary will be linked with it:

  -rw-rw-r--. 1 acme acme 0 Feb 12 14:49 /tmp/build/perf/feature/test-libopencsd.make.output
  $ ldd ~/bin/perf | grep opencsd
libopencsd_c_api.so.0 => /lib64/libopencsd_c_api.so.0 (0x00007fd43097f000)
libopencsd.so.0 => /lib64/libopencsd.so.0 (0x00007fd4308f8000)
  $

To make sure this gets built before pushing things upstream I have a
ubuntu:19.04-x-arm64 container that has:

  [root@quaco x-arm64]# grep CORESIGHT Dockerfile
  ENV EXTRA_MAKE_ARGS=CORESIGHT=1
  [root@quaco x-arm64]#

So that I always build with libopencsd before pushing things upstream.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kim Phillips <kim.phillips@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Link: https://lkml.kernel.org/n/tip-20vyy39jw9jgrijesi30fgox@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Filter out gnome-terminal* parent

Just like it does with 'sshd', to reduce the feedback loop when doing
system wide tracing on on a gnome GUI.

Need to figure out how to auto-filter the calls to other UI components
tho.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-rjopq5y92itgokppdhe8sc6z@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools build: Add -lrt to FEATURE_CHECK_LDFLAGS-libaio

Since we need it to resolve the AIO symbols, otherwise we fail with:

  $ cat /tmp/build/perf/feature/test-all.make.output
  /usr/bin/ld: /tmp/ccEqrj36.o: undefined reference to symbol 'aio_return64@@GLIBC_2.2.5'
  /usr/bin/ld: //usr/lib64/librt.so.1: error adding symbols: DSO missing from command line
  collect2: error: ld returned 1 exit status
  $

When we added the aio support in 'perf record' only the test-libaio.bin
target got the -lrt, i.e. the feature detection slow path. Fix it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Fixes: 2a07d814747b ("tools build feature: Check if libaio is available")
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty waitid options: Fix up prefix showing logic

When introducing the possibility for selecting if the common prefix to
options such as the waitid ones, i.e. all 'waitid' options start with
'W', so, to make it make it more compact if configured to suppress it,
'perf trace' will do so, other examples include mmap's PROT_ prefix for
its 'prot' argument, etc, which, when showing the syscall argument name
ends up producing duplicated info that clutters the screen, i.e.:

  # perf trace -e mmap --max-events 2 sleep 1
     0.000 ( 0.014 ms): sleep/20886 mmap(len: 112595, prot: PROT_READ, flags: MAP_PRIVATE, fd: 3) = 0x7f3e986d2000
     0.041 ( 0.005 ms): sleep/20886 mmap(len: 8192, prot: PROT_READ|PROT_WRITE, flags: MAP_PRIVATE|MAP_ANONYMOUS) = 0x7f3e986d0000
  #

So it is possible to suppress that and make it more compact by having
this in your ~/.perfconfig:

  # cat ~/.perfconfig
  [trace]
show_prefix = no
  #

  # perf trace -e mmap --max-events 2 sleep 1
     0.000 ( 0.014 ms): sleep/8009 mmap(len: 112595, prot: READ, flags: PRIVATE, fd: 3) = 0x7ff2373de000
     0.040 ( 0.005 ms): sleep/8009 mmap(len: 8192, prot: READ|WRITE, flags: PRIVATE|ANONYMOUS) = 0x7ff2373dc000
  #

To have it look more like strace's output, we instead want to suppress
the arg name and show the prefix, so use:

  # cat ~/.perfconfig
  [trace]
show_prefix = yes
show_arg_names = no
  #
  # perf trace -e mmap --max-events 2 sleep 1
     0.000 ( 0.006 ms): sleep/15513 mmap(NULL, 112595, PROT_READ, MAP_PRIVATE, 3, 0) = 0x7f7a9b6d3000
     0.020 ( 0.002 ms): sleep/15513 mmap(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS) = 0x7f7a9b6d1000
  #

When this logic was introduced a bug came with it when processing the
waitid 'option' arg that ended up expecting 3 strings when just two were
being provided, fix it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: c65c83ffe904 ("perf trace: Allow asking for not suppressing common string prefixes")
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Check if the 'fd' is negative when mapping it to pathname

We were crashing when processing a negative fd:

  Program received signal SIGSEGV, Segmentation fault.
  0x0000000000609bbf in syscall_arg__scnprintf_ioctl_cmd (bf=0x1172eca "", size=2038, arg=0x7fffffff8360) at trace/beauty/ioctl.c:182
  182 if (file->dev_maj == USB_DEVICE_MAJOR)
  Missing separate debuginfos, use: dnf debuginfo-install bzip2-libs-1.0.6-28.fc29.x86_64 elfutils-libelf-0.174-5.fc29.x86_64 elfutils-libs-0.174-5.fc29.x86_64 glib2-2.58.3-1.fc29.x86_64 libbabeltrace-1.5.6-1.fc29.x86_64 libunwind-1.2.1-6.fc29.x86_64 libuuid-2.32.1-1.fc29.x86_64 libxcrypt-4.4.3-2.fc29.x86_64 numactl-libs-2.0.12-1.fc29.x86_64 openssl-libs-1.1.1a-1.fc29.x86_64 pcre-8.42-6.fc29.x86_64 perl-libs-5.28.1-427.fc29.x86_64 popt-1.16-15.fc29.x86_64 python2-libs-2.7.15-11.fc29.x86_64 slang-2.3.2-4.fc29.x86_64 xz-libs-5.2.4-3.fc29.x86_64
  (gdb) bt
  #0  0x0000000000609bbf in syscall_arg__scnprintf_ioctl_cmd (bf=0x1172eca "", size=2038, arg=0x7fffffff8360) at trace/beauty/ioctl.c:182
  #1  0x000000000048e295 in syscall__scnprintf_val (sc=0x123b500, bf=0x1172eca "", size=2038, arg=0x7fffffff8360, val=21519)
      at builtin-trace.c:1594
  #2  0x000000000048e60d in syscall__scnprintf_args (sc=0x123b500, bf=0x1172ec6 "-1, ", size=2042, args=0x7ffff6a7c034 "\377\377\377\377",
      augmented_args=0x7ffff6a7c064, augmented_args_size=4, trace=0x7fffffffa8d0, thread=0x1175cd0) at builtin-trace.c:1661
  #3  0x000000000048f04e in trace__sys_enter (trace=0x7fffffffa8d0, evsel=0xb260b0, event=0x7ffff6a7bfe8, sample=0x7fffffff84f0)
      at builtin-trace.c:1880
  #4  0x00000000004915a4 in trace__handle_event (trace=0x7fffffffa8d0, event=0x7ffff6a7bfe8, sample=0x7fffffff84f0) at builtin-trace.c:2590
  #5  0x0000000000491eed in __trace__deliver_event (trace=0x7fffffffa8d0, event=0x7ffff6a7bfe8) at builtin-trace.c:2818
  #6  0x0000000000492030 in trace__deliver_event (trace=0x7fffffffa8d0, event=0x7ffff6a7bfe8) at builtin-trace.c:2845
  #7  0x0000000000492896 in trace__run (trace=0x7fffffffa8d0, argc=0, argv=0x7fffffffdb58) at builtin-trace.c:3040
  #8  0x000000000049603a in cmd_trace (argc=0, argv=0x7fffffffdb58) at builtin-trace.c:3952
  #9  0x00000000004d5103 in main (argc=1, argv=0x7fffffffdb58) at perf.c:474
  (gdb) p fd
  $1 = -1
  (gdb) p file
  $7 = (struct file *) 0xfffffffffffffff0
  (gdb) p ((struct thread_trace *)arg->thread)->files.table + fd
  $8 = (struct file *) 0xfffffffffffffff0
  (gdb)

Check for that and return NULL instead.

This problem was introduced recently, the other codepaths leading to
thread_trace__files_entry() check for negative fds, like thread__fd_path(),
but we need to do it at thread_trace__files_entry() as more users are now
calling it directly.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: 2d473389f87a ("perf trace beauty: Export function to get the files for a thread")
Link: https://lkml.kernel.org/n/tip-oq7bvaaf07gsd4yqty3107u2@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf beauty ioctl cmd: The 'fd' arg is signed

It is possible to pass a negative number as the fd and that has to be
handled, so stop using 'unsigned int fd' in the ioctl syscall 'cmd'
beautifier.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-b7qwa0l19dswa09h3s41akfu@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools feature: Undef _GNU_SOURCE at the end of feature tests

Since we get all the tests in a single .c file for a first test,
tools/build/feature/test-all.c, if individual tests set that define and
fail to undef it at its end, then it the test-all.c build will fail due
to defining _GNU_SOURCE multiple times, getting us to the slow path,
so undef it at the end in tests that define it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-w6s00jfo1xabgphzczadl59b@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf utils: Silence "Couldn't synthesize bpf events" warning for EPERM

Synthesizing BPF events is only supported for root. Silent warning msg
when non-root user runs perf-record.

Reported-by: David Carrillo-Cisneros <davidca@fb.com>
Signed-off-by: Song Liu <songliubraving@fb.com>
Tested-by: David Carrillo-Cisneros <davidca@fb.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: kernel-team@fb.com
Link: http://lkml.kernel.org/r/20190204193140.719740-1-songliubraving@fb.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events power9: General metrics

Descriptions of metrics for POWER9 processors can be found in the
"POWER9 Performance Monitor Unit User’s Guide", which is currently
available on the "IBM Portal for OpenPOWER"
(https://www-355.ibm.com/systems/power/openpower/welcome.xhtml) at
https://www-355.ibm.com/systems/power/openpower/posting.xhtml?postingId=4948CDE1963C9BCA852582F800718190

This patch is for metric groups:
- general

and other metrics not in a metric group.

Signed-off-by: Paul Clarke <pc@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Cc: linuxppc-dev@ozlabs.org
Link: http://lkml.kernel.org/r/20190209181429.23950-5-pc@us.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events power9: Branch_prediction, instruction_stats, latency, lsu_rejects, memory, prefetch & translation metrics

Descriptions of metrics for POWER9 processors can be found in the
"POWER9 Performance Monitor Unit User’s Guide", which is currently
available on the "IBM Portal for OpenPOWER"
(https://www-355.ibm.com/systems/power/openpower/welcome.xhtml) at
https://www-355.ibm.com/systems/power/openpower/posting.xhtml?postingId=4948CDE1963C9BCA852582F800718190

This patch is for metric groups:
- branch_prediction
- instruction_stats_percent_per_ref
- latency
- lsu_rejects
- memory
- prefetch
- translation

Plus, some whitespace changes.

Signed-off-by: Paul Clarke <pc@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Cc: linuxppc-dev@ozlabs.org
Link: http://lkml.kernel.org/r/20190209181429.23950-4-pc@us.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events power9: Dl1_reloads, instruction_misses, l[23]_stats & pteg_reloads metrics

Descriptions of metrics for POWER9 processors can be found in the
"POWER9 Performance Monitor Unit User’s Guide", which is currently
available on the "IBM Portal for OpenPOWER"
(https://www-355.ibm.com/systems/power/openpower/welcome.xhtml) at
https://www-355.ibm.com/systems/power/openpower/posting.xhtml?postingId=4948CDE1963C9BCA852582F800718190

This patch is for metric groups:
- dl1_reloads_percent_per_inst
- dl1_reloads_percent_per_ref
- instruction_misses_percent_per_inst
- l2_stats
- l3_stats
- pteg_reloads_percent_per_inst
- pteg_reloads_percent_per_ref

Signed-off-by: Paul Clarke <pc@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Cc: linuxppc-dev@ozlabs.org
Link: http://lkml.kernel.org/r/20190209181429.23950-3-pc@us.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events power9: Cpi_breakdown & estimated_dcache_miss_cpi metrics

Descriptions of metrics for POWER9 processors can be found in the
"POWER9 Performance Monitor Unit User’s Guide", which is currently
available on the "IBM Portal for OpenPOWER"
(https://www-355.ibm.com/systems/power/openpower/welcome.xhtml) at
https://www-355.ibm.com/systems/power/openpower/posting.xhtml?postingId=4948CDE1963C9BCA852582F800718190

This patch is for metric groups:
- cpi_breakdown
- estimated_dcache_miss_cpi

Signed-off-by: Paul Clarke <pc@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Cc: linuxppc-dev@ozlabs.org
Link: http://lkml.kernel.org/r/20190209181429.23950-2-pc@us.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events power8: Translaton & general metrics

POWER8 metrics are not well publicized.

Some are here:

  https://www.ibm.com/support/knowledgecenter/en/SSFK5S_2.2.0/com.ibm.cluster.pedev.v2r2.pedev100.doc/bl7ug_derivedmetricspower8.htm

This patch is for metric groups:
- translation
- general

and other metrics not in a metric group.

Signed-off-by: Paul Clarke <pc@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/20190207175314.31813-5-pc@us.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events power8: Branch_prediction, latency, bus_stats, instruction_mix & instruction_stats metrics

POWER8 metrics are not well publicized.  Some are here:

  https://www.ibm.com/support/knowledgecenter/en/SSFK5S_2.2.0/com.ibm.cluster.pedev.v2r2.pedev100.doc/bl7ug_derivedmetricspower8.htm

This patch is for metric groups:
- branch_prediction
- latency
- bus_stats
- instruction_mix
- instruction_stats_percent_per_ref

Signed-off-by: Paul Clarke <pc@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/20190207175314.31813-4-pc@us.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events power8: Dl1_reload, instruction_misses, l2_stats, lsu_rejects, memory & pteg_reloads metrics

POWER8 metrics are not well publicized.

Some are here:

  https://www.ibm.com/support/knowledgecenter/en/SSFK5S_2.2.0/com.ibm.cluster.pedev.v2r2.pedev100.doc/bl7ug_derivedmetricspower8.htm

This patch is for metric groups:
- dl1_reloads_percent_per_inst
- dl1_reloads_percent_per_ref
- instruction_misses_percent_per_inst
- l2_stats
- lsu_rejects
- memory
- pteg_reloads_percent_per_inst
- pteg_reloads_percent_per_ref

Signed-off-by: Paul Clarke <pc@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/20190207175314.31813-3-pc@us.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events power8: Cpi_breakdown & estimated_dcache_miss_cpi metrics

POWER8 metrics are not well publicized.

Some are here:

  https://www.ibm.com/support/knowledgecenter/en/SSFK5S_2.2.0/com.ibm.cluster.pedev.v2r2.pedev100.doc/bl7ug_derivedmetricspower8.htm

This patch is for metric groups:
- cpi_breakdown
- estimated_dcache_miss_cpi

Signed-off-by: Paul Clarke <pc@us.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Cc: Carl Love <cel@us.ibm.com>
Cc: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Cc: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/20190207175314.31813-2-pc@us.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf report: Add s390 diagnosic sampling descriptor size

On IBM z13 machine types 2964 and 2965 the descriptor
sizes for sampling and diagnostic sampling entries
might be missing in the trailer entry and are set to zero.

This leads to a perf report failure when processing diagnostic
sampling entries.

This patch adds missing descriptor sizes when the trailer entry
contains zero for these fields.

Output before:
  [root@s38lp82 perf]#  ./perf report --stdio | fgrep Samples
  0xabbf0 [0x8]: failed to process type: 68
  Error:
  failed to process sample
  [root@s38lp82 perf]#

Output after:
  [root@s38lp82 perf]#  ./perf report --stdio | fgrep Samples
  # Total Lost Samples: 0
  # Samples: 3K of event 'SF_CYCLES_BASIC_DIAG'
  # Samples: 162  of event 'CF_DIAG'
  [root@s38lp82 perf]#

Fixes: 2b1444f2e28b ("perf report: Add raw report support for s390 auxiliary trace")
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: Hendrik Brueckner <brueckner@linux.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Link: http://lkml.kernel.org/r/20190211100627.85714-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add proper header file for symbols

After 'commit e22c1c751140 ("perf thread: Don't include symbol.h,
symbol_conf.h is enough")'

Compilation of the perf tools is broken when using the functionality
provided by the openCSD library:

[...]

...                       timerfd: [ on  ]
...                  sched_getcpu: [ on  ]
...                           sdt: [ OFF ]
...                         setns: [ on  ]
...                    libopencsd: [ on  ]

[...]

  CC       util/arm-spe.o
  CC       util/arm-spe-pkt-decoder.o
  CC       util/s390-cpumsf.o
  CC       util/cs-etm.o
  CC       util/parse-branch-options.o
util/cs-etm.c: In function ‘cs_etm__mem_access’:
util/cs-etm.c:297:24: error: storage size of ‘al’ isn’t known
  struct  addr_location al;

And rightly so since file cs-etm.c doesn't include symbol.h, something
that is rectified in this patch.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190208223543.31836-1-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Implement --affinity=node|cpu option

Implement --affinity=node|cpu option for the record mode defaulting
to system affinity mask bouncing.

Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/083f5422-ece9-10dd-8305-bf59c860f10f@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf/x86/intel: Add counter freezing quirk for Goldmont

A microcode patch is also needed for Goldmont while counter freezing
feature is enabled. Otherwise, there will be some issues, e.g. PMI lost.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: bp@alien8.de
Link: https://lkml.kernel.org/r/1549319013-4522-5-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel: Clean up counter freezing quirk

Clean up counter freezing quirk to use the new facility to check for
min microcode revisions.

Rename the counter freezing quirk related functions. Because other
platforms, e.g. Goldmont, also needs to call the quirk.

Only check the boot CPU, assuming models and features are consistent
over all CPUs.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: bp@alien8.de
Link: https://lkml.kernel.org/r/1549319013-4522-4-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel: Clean up SNB PEBS quirk

Clean up SNB PEBS quirk to use the new facility to check for min
microcode revisions.

Only check the boot CPU, assuming models and features are consistent
over all CPUs.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: bp@alien8.de
Link: https://lkml.kernel.org/r/1549319013-4522-3-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/kvm: Avoid unnecessary work in guest filtering

KVM added a workaround for PEBS events leaking into guests with
commit:

  26a4f3c08de4 ("perf/x86: disable PEBS on a guest entry.")

This uses the VT entry/exit list to add an extra disable of the
PEBS_ENABLE MSR.

Intel also added a fix for this issue to microcode updates on
Haswell/Broadwell/Skylake.

It turns out using the MSR entry/exit list makes VM exits
significantly slower. The list is only needed for disabling
PEBS, because the GLOBAL_CTRL change gets optimized by
KVM into changing the VMCS.

Check for the microcode updates that have the microcode
fix for leaking PEBS, and disable the extra entry/exit list
entry for PEBS_ENABLE. In addition we always clear the
GLOBAL_CTRL for the PEBS counter while running in the guest,
which is enough to make them never fire at the wrong
side of the host/guest transition.

The overhead for VM exits with the filtering active with the patch is
reduced from 8% to 4%.

The microcode patch has already been merged into future platforms.
This patch is one-off thing. The quirks is used here.

For other old platforms which doesn't have microcode patch and quirks,
extra disable of the PEBS_ENABLE MSR is still required.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: bp@alien8.de
Link: https://lkml.kernel.org/r/1549319013-4522-2-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge branch 'x86/cpu' into perf/core, to pick up dependent commit

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

x86/cpufeature: Add facility to check for min microcode revisions

For bug workarounds or checks, it is useful to check for specific
microcode revisions.

Add a new generic function to match the CPU with stepping.
Add the other function to check the min microcode revisions for
the matched CPU.

A new table format is introduced to facilitate the quirk to
fill the related information.

This does not change the existing x86_cpu_id because it's an ABI
shared with modules, and also has quite different requirements,
as in no wildcards, but everything has to be matched exactly.

Originally-by: Andi Kleen <ak@linux.intel.com>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: eranian@google.com
Link: https://lkml.kernel.org/r/1549319013-4522-1-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'perf-core-for-mingo-5.1-20190206' of git://git./linux/kernel/git/acme/linux into perf/core

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

Hardware tracing:

  Adrian Hunter:

  - Handle calls optimized into jumps to a different symbol
    in the thread stack routines used to process hardware traces (Adrian Hunter)

Intel PT:

  Adrian Hunter:

  - Fix overlap calculation for padding.

  - Fix CYC timestamp calculation after OVF.

  - Packet splitting can only happen in 32-bit.

  - Add timestamp to auxtrace errors.

ARM CoreSight:

  Leo Yan:

  - Add last instruction information in packet

  - Set sample flags for instruction range, exception and
    return packets and for a trace discontinuity.

  - Add exception number in exception packet

  - Change tuple from traceID-CPU# to traceID-metadata

  - Add traceID in packet

  Mathieu Poirier:

  - Add "sinks" group to PMU directory

  - Use event attributes to send sink information to kernel

  - Remove set_drv_config() API, no longer used.

perf annotate:

  Jiri Olsa:

  - Delay symbol annotation to the resort phase, speeding up 'perf report'
    startup.

perf record:

  Alexey Budankov:

  - Allow binding userspace buffers to NUMA nodes.

Symbols:

  Adrian Hunter:

  - Fix calculating of symbol sizes when splitting kallsyms into
    maps for kcore processing.

Vendor events:

  William Cohen:

  - Intel: Fix Load_Miss_Real_Latency on CLX

Misc:

  Arnaldo Carvalho de Melo:

  - Streamline headers, removing includes when all that is needed are
    just forward declarations, fixup the fallout for cases where headers
    should have been explicitely included but were instead obtained
    indirectly, by sheer luck.

  - Add fallback versions for CPU_{OR,EQUAL}(), so that code using it
    continue to build on older systems where those were not yet introduced
    or in systems using some other libc than the GNU one where those
    helpers aren't present.

Documentation:

  Changbin Du:

  - Add documentation for BPF event selection.

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

Merge branch 'perf/urgent' into perf/core, to pick up fixes

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

Merge tag 'perf-urgent-for-mingo-5.0-20190205' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fixes from Arnaldo Carvalho de Melo:

perf trace:

  Arnaldo Carvalho de Melo:

    Fix handling of probe:vfs_getname when the probed routine is
    inlined in multiple places, fixing the collection of the 'filename'
    parameter in open syscalls.

perf test:

  Gustavo A. R. Silva:

    Fix bitwise operator usage in evsel-tp-sched test, which made tat
    test always detect fields as signed.

  Jiri Olsa:

    Filter out hidden symbols from labels, added in systems where the
    annobin plugin is used, such as RHEL8, which, if left in place make
    the DWARF unwind 'perf test' to fail on PPC.

  Tony Jones:

    Fix 'perf_event_attr' tests when building with python3.

perf mem/c2c:

  Ravi Bangoria:

    Fix perf_mem_events on PowerPC.

tools headers UAPI:

  Arnaldo Carvalho de Melo:

    Sync linux/in.h copy from the kernel sources, silencing a perf build warning.

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

perf auxtrace: Add timestamp to auxtrace errors

The timestamp can use useful to find part of a trace that has an error
without outputting all of the trace e.g. using the itrace 's' option to
skip initial number of events.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190206103947.15750-6-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Packet splitting can happen only on 32-bit

Data is copied when the trace is stopped, so packets are never split
between buffers except when processing if the buffer cannot fit in the
address space which can only happen on 32-bit systems. Change the logic
to reflect that.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190206103947.15750-5-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Fix CYC timestamp calculation after OVF

CYC packet timestamp calculation depends upon CBR which was being
cleared upon overflow (OVF). That can cause errors due to failing to
synchronize with sideband events. Even if a CBR change has been lost,
the old CBR is still a better estimate than zero. So remove the clearing
of CBR.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/20190206103947.15750-4-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Fix overlap calculation for padding

Auxtrace records might have up to 7 bytes of padding appended. Adjust
the overlap accordingly.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/20190206103947.15750-3-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf auxtrace: Define auxtrace record alignment

Define auxtrace record alignment so that it can be referenced elsewhere.

Note this is preparation for patch "perf intel-pt: Fix overlap calculation
for padding"

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/20190206103947.15750-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf thread-stack: Represent jmps to the start of a different symbol

The compiler might optimize a call/ret combination by making it a jmp.
However the thread-stack does not presently cater for that, so that such
control flow is not visible in the call graph. Make it visible by
recording on the stack a branch to the start of a different symbol.
Note, that means when a ret pops the stack, all jmps must be popped off
first.

Example:

  $ cat jmp-to-fn.c
  __attribute__((noinline)) int bar(void)
  {
          return -1;
  }

  __attribute__((noinline)) int foo(void)
  {
          return bar() + 1;
  }

  int main()
  {
          return foo();
  }
  $ gcc -ggdb3 -Wall -Wextra -O2 -o jmp-to-fn jmp-to-fn.c
  $ objdump -d jmp-to-fn
  <SNIP>
  0000000000001040 <main>:
      1040:       31 c0                   xor    %eax,%eax
      1042:       e9 09 01 00 00          jmpq   1150 <foo>
  <SNIP>
  0000000000001140 <bar>:
      1140:       b8 ff ff ff ff          mov    $0xffffffff,%eax
      1145:       c3                      retq
  <SNIP>
  0000000000001150 <foo>:
      1150:       31 c0                   xor    %eax,%eax
      1152:       e8 e9 ff ff ff          callq  1140 <bar>
      1157:       83 c0 01                add    $0x1,%eax
      115a:       c3                      retq
  <SNIP>
  $ perf record -o jmp-to-fn.perf.data -e intel_pt/cyc/u ./jmp-to-fn
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0,017 MB jmp-to-fn.perf.data ]
  $ perf script -i jmp-to-fn.perf.data --itrace=be -s ~/libexec/perf-core/scripts/python/export-to-sqlite.py jmp-to-fn.db branches calls
  2019-01-08 13:24:58.783069 Creating database...
  2019-01-08 13:24:58.794650 Writing records...
  2019-01-08 13:24:59.008050 Adding indexes
  2019-01-08 13:24:59.015802 Done
  $  ~/libexec/perf-core/scripts/python/exported-sql-viewer.py jmp-to-fn.db

Before:

    main
        -> bar

After:

    main
        -> foo
            -> bar

Committer testing:

Install the python2-pyside package, then select these menu options
on the GUI:

   "Reports"
      "Context sensitive callgraphs"

Then go on expanding the symbols, to get, full picture when doing this
on a fedora:29 with gcc version 8.2.1 20181215 (Red Hat 8.2.1-6) (GCC):

jmp-to-fn
  PID:TID
    _start                (ld-2.28.so)
      __libc_start_main
        main
          foo
            bar

To verify that indeed, this fixes the problem.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: http://lkml.kernel.org/r/20190109091835.5570-5-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf thread-stack: Tidy thread_stack__no_call_return() by adding more local variables

Make thread_stack__no_call_return() more readable by adding more local
variables.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: http://lkml.kernel.org/r/20190109091835.5570-4-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf thread-stack: Tidy thread_stack__push_cp() usage

If 'cp' is checked in thread_stack__push_cp() a number of error checks
can be removed, reducing code size and improving readability.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: http://lkml.kernel.org/r/20190109091835.5570-3-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Fix split_kallsyms_for_kcore() for trampoline symbols

Kallsyms symbols do not have a size, so the size becomes the distance to
the next symbol.

Consequently the recently added trampoline symbols end up with large
sizes because the trampolines are some distance from one another and the
main kernel map.

However, symbols that end outside their map can disrupt the symbol tree
because, after mapping, it can appear incorrectly that they overlap
other symbols.

Add logic to truncate symbol size to the end of the corresponding map.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: stable@vger.kernel.org
Fixes: d83212d5dd67 ("kallsyms, x86: Export addresses of PTI entry trampolines")
Link: http://lkml.kernel.org/r/20190109091835.5570-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf vendor events intel: Fix Load_Miss_Real_Latency on CLX

Fix incorrect event names for the Load_Miss_Real_Latency metric for
Cascadelake server in the same manner as commit 91b2b97025 for SKL/SKX.

Signed-off-by: William Cohen <wcohen@redhat.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190129170536.22510-1-wcohen@redhat.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Set sample flags for exception return packet

When return from exception, we need to distinguish if it's system call
return or for other type exceptions for setting sample flags.  Due to
the exception return packet doesn't contain exception number, so we
cannot decide sample flags based on exception number.

On the other hand, the exception return packet is followed by an
instruction range packet; this range packet deliveries the start address
after exception handling, we can check if it is a SVC instruction just
before the start address.  If there has one SVC instruction is found
ahead the return address, this means it's an exception return for system
call; otherwise it is an normal return for other exceptions.

This patch is to set sample flags for exception return packet, firstly
it simply set sample flags as PERF_IP_FLAG_INTERRUPT for all exception
returns since at this point it doesn't know what's exactly the exception
type.  We will defer to decide if it's an exception return for system
call when the next instruction range packet comes, it checks if there
has one SVC instruction prior to the start address and if so we will
change sample flags to PERF_IP_FLAG_SYSCALLRET for system call return.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Robert Walker <robert.walker@arm.com>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: coresight ml <coresight@lists.linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190129122842.32041-9-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Set sample flags for exception packet

The exception taken and returning are typical flow for instruction jump
but it needs to be handled with exception packets. This patch is to set
sample flags for exception packet.

Since the exception packet contains the exception number, according to
the exception number this patch makes decision for belonging to which
exception types.

The decoder have defined different exception number for ETMv3 and ETMv4
separately, hence this patch needs firstly decide the ETM version by
using the metadata magic number, and this patch adds helper function
cs_etm__get_magic() for easily getting magic number.

Based on different ETM version, the exception packet contains the
exception number, according to the exception number this patch makes
decision for the exception belonging to which exception types.

In this patch, it introduces helper function cs_etm__is_svc_instr(); for
ETMv4 CS_ETMV4_EXC_CALL covers SVC, SMC and HVC cases in the single
exception number, thus need to use cs_etm__is_svc_instr() to decide an
exception taken for system call.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Robert Walker <robert.walker@arm.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: coresight ml <coresight@lists.linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190129122842.32041-8-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add traceID in packet

Add traceID in packet, thus we can use traceID to retrieve metadata
pointer from traceID-metadata tuple.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Robert Walker <robert.walker@arm.com>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: coresight ml <coresight@lists.linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190129122842.32041-7-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Change tuple from traceID-CPU# to traceID-metadata

If packet processing wants to know the packet is bound with which ETM
version, it needs to access metadata to decide that based on metadata
magic number; but we cannot simply to use CPU logic ID number as index
to access metadata sequential array, especially when system have
hotplugged off CPUs, the metadata array are only allocated for online
CPUs but not offline CPUs, so the CPU logic number doesn't match with
its index in the array.

This patch is to change tuple from traceID-CPU# to traceID-metadata,
thus it can use the tuple to retrieve metadata pointer according to
traceID.

For safe accessing metadata fields, this patch provides helper function
cs_etm__get_cpu() which is used to return CPU number according to
traceID; cs_etm_decoder__buffer_packet() is the first consumer for this
helper function.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Robert Walker <robert.walker@arm.com>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: coresight ml <coresight@lists.linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190129122842.32041-6-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add exception number in exception packet

When an exception packet comes, it contains the information for
exception number; the exception number indicates the exception types, so
from it we can know if the exception is taken for interrupt, system call
or other traps, etc.

This patch simply adds a field in cs_etm_packet struct, it records
exception number for exception packet that will then be used to properly
identify exception types to the perf synthesize mechanic.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Robert Walker <robert.walker@arm.com>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: coresight ml <coresight@lists.linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190129122842.32041-5-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Set sample flags for trace discontinuity

In the middle of trace stream, it might be interrupted thus the trace
data is not continuous, the trace stream firstly is ended for previous
trace block and restarted for next block.

To display related information for showing trace is restarted, this
patch set sample flags for trace discontinuity:

- If one discontinuity packet is coming, append flag
  PERF_IP_FLAG_TRACE_END to the previous packet to indicate the trace
  has been ended;
- If one instruction packet is following discontinuity packet, this
  instruction packet is the first one packet to restarting trace.  So
  set flag PERF_IP_FLAG_TRACE_START to discontinuity packet, this flag
  will be used to generate sample when connect with the sequential
  instruction packet.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Robert Walker <robert.walker@arm.com>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: coresight ml <coresight@lists.linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190129122842.32041-4-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Set sample flags for instruction range packet

The perf sample data contains flags to indicate the hardware trace data
is belonging to which type branch instruction, thus this can be used to
print out the human readable string.  Arm CoreSight ETM sample data is
missed to set flags and it is always set to zeros, this results in perf
tool skips to print string for instruction types.

This patch is to set branch instruction flags for instruction range
packet.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Robert Walker <robert.walker@arm.com>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: coresight ml <coresight@lists.linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190129122842.32041-3-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add last instruction information in packet

Decoder provides last instruction related information, these information
can be used for trace analysis; specifically we can get to know what
kind of branch instruction has been executed, mainly the information are
contained in three element fields:

  last_i_type: this is significant type for waypoint calculation, it
  indicates the last instruction is one of immediate branch instruction,
  indirect branch instruction, instruction barrier (ISB), or data
  barrier (DSB/DMB).

  last_i_subtype: this is used for instruction sub type, it can be
  branch with link, ARMv8 return instruction, ARMv8 eret instruction
  (return from exception), or ARMv7 instruction which could imply
  return (e.g. MOV PC, LR; POP { ,PC}).

  last_instr_cond: it indicates if the last instruction was conditional.

But these three fields are not saved into cs_etm_packet struct, thus
cs-etm layer don't know related information and cannot generate sample
flags for branch instructions.

This patch add corresponding three new fields in cs_etm_packet struct
and save related value into the packet structure, it is preparation for
supporting sample flags.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Robert Walker <robert.walker@arm.com>
Cc: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: coresight ml <coresight@lists.linaro.org>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190129122842.32041-2-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add documentation for BPF event selection

Add documentation for how to pass a BPF program as a perf event.

Signed-off-by: Changbin Du <changbin.du@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190201134651.12373-1-changbin.du@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf report: Move symbol annotation to the resort phase

Currently we make the annotation for the IPC column during the entry
display, already outside of the progress bar scope, so it appears like
'perf report' is stuck.

Move the annotation retrieval to the resort phase, so that all the data
are ready for display.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190204141808.23031-4-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Add output_resort_cb method

Add perf_evsel__output_resort_cb() so we have an interface with a
callback for each hist entry. It will be used in the following patch.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190204141808.23031-3-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf hists: Add argument to hists__resort_cb_t callback

Add argument to hists__resort_cb_t so that we can pass data from upper
layers to the callback function. It will be used in the following
patches.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20190204141808.23031-2-jolsa@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf clang: Do not use 'return std::move(something)'

It prevents copy elision, generating this warning when building with
fedora:rawhide's clang:

  clang version 7.0.1 (Fedora 7.0.1-2.fc30)
  Target: x86_64-unknown-linux-gnu
  Thread model: posix
  InstalledDir: /usr/bin
  Found candidate GCC installation: /usr/bin/../lib/gcc/x86_64-redhat-linux/9
  Found candidate GCC installation: /usr/lib/gcc/x86_64-redhat-linux/9
  Selected GCC installation: /usr/bin/../lib/gcc/x86_64-redhat-linux/9
  Candidate multilib: .;@m64
  Candidate multilib: 32;@m32
  Selected multilib: .;@m64

  $ make -C tools/perf CC=clang LIBCLANGLLVM=1
  <SNIP>
  util/c++/clang.cpp: In function 'std::unique_ptr<llvm::SmallVectorImpl<char> > perf::getBPFObjectFromModule(llvm::Module*)':
  util/c++/clang.cpp:163:18: error: moving a local object in a return statement prevents copy elision [-Werror=pessimizing-move]
    163 |  return std::move(Buffer);
        |         ~~~~~~~~~^~~~~~~~
  util/c++/clang.cpp:163:18: note: remove 'std::move' call
  cc1plus: all warnings being treated as errors
  <SNIP>

References:

  http://www.cplusplus.com/forum/general/186411/#msg908572
  https://en.cppreference.com/w/cpp/language/return#Notes
  https://en.cppreference.com/w/cpp/language/copy_elision

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-lehqf5x5q96l0o8myhb6blz6@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Apply affinity masks when reading mmap buffers

Build node cpu masks for mmap data buffers. Apply node cpu masks to tool
thread every time it references data buffers cross node or cross cpu.

Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/b25e4ebc-078d-2c7b-216c-f0bed108d073@linux.intel.com
[ Use cpu-set-sched.h to get the CPU_{EQUAL,OR}() fallbacks for older systems ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add fallback versions for CPU_{OR,EQUAL}()

From the glibc sources, so that we can keep the tooling buildable in
older systems while using recent sched.h CPU_ macros.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexey Budankov <alexey.budankov@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lkml.kernel.org/n/tip-hvm9ysmrjip75ebdzhzoh429@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Bind the AIO user space buffers to nodes

Allocate and bind AIO user space buffers to the memory nodes that mmap
kernel buffers are bound to.

Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/5a5adebc-afe0-4806-81cd-180d49ec043f@linux.intel.com
[ Do not use 'index' as a variable name, it is a define in older glibcs ]
Link: http://lkml.kernel.org/r/20190205151526.GC10613@kernel.org
[ Add -lnuma to the python build when -DHAVE_LIBNUMA_SUPPORT is present, fixing 'perf test python' ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Allocate affinity masks

Allocate affinity option and masks for mmap data buffers and record
thread as well as initialize allocated objects.

Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/526fa2b0-07de-6dbd-a7e9-26ba875593c9@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf pmu: Remove set_drv_config API

CoreSight was the only client of the PMU's set_drv_config() API.  Now
that it is no longer needed by CoreSight remove it from the code base.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-s390@vger.kernel.org
Link: http://lkml.kernel.org/r/20190131184714.20388-8-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf coresight: Remove set_drv_config() API

Now that the event's config2 attribute is used to communicate sink
selection to the kernel, remove the old set_drv_config() implementation
since it is no longer needed.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-s390@vger.kernel.org
Link: http://lkml.kernel.org/r/20190131184714.20388-7-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf arm cs-etm: Use event attributes to send sink information to kernel

The communication of sink information for a trace session doesn't work
when more than on CPU is involved in the scenario due to the static
nature of sysfs.  As such communicate the sink information to each event
by using the perf_event::attr:config2 attribute.  The information sent
to the kernel is an hash of the sink's name, which is unique in a
system.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-s390@vger.kernel.org
Link: http://lkml.kernel.org/r/20190131184714.20388-6-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf pmu: Move EVENT_SOURCE_DEVICE_PATH to PMU header file

Move definition of EVENT_SOURCE_DEVICE_PATH to pmu.h so that it can be
used by other files than pmu.c

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Suzuki K Poulouse <suzuki.poulose@arm.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-s390@vger.kernel.org
Link: http://lkml.kernel.org/r/20190131184714.20388-5-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

coresight: Use event attributes for sink selection

This patch uses the information conveyed by perf_event::attr::config2
to select a sink to use for the session.  That way a sink can easily be
selected to be used by more than one source, something that isn't currently
possible with the sysfs implementation.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-s390@vger.kernel.org
Link: http://lkml.kernel.org/r/20190131184714.20388-4-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

coresight: perf: Add "sinks" group to PMU directory

Add a "sinks" directory entry so that users can see all the sinks
available in the system in a single place.  Individual sink are added
as they are registered with the coresight bus.

Committer tests:

Test built on a ubuntu 18.04 container with a cross build environment to
arm64, the new field is there, need to find a machine with this feature
to do further testing in the future.

  root@d15263e5734a:/git/perf# grep CORESIGHT /tmp/build/v5.0-rc2+/.config
  CONFIG_CORESIGHT=y
  CONFIG_CORESIGHT_LINKS_AND_SINKS=y
  CONFIG_CORESIGHT_LINK_AND_SINK_TMC=y
  CONFIG_CORESIGHT_CATU=y
  CONFIG_CORESIGHT_SINK_TPIU=y
  CONFIG_CORESIGHT_SINK_ETBV10=y
  CONFIG_CORESIGHT_SOURCE_ETM4X=y
  CONFIG_CORESIGHT_DYNAMIC_REPLICATOR=y
  CONFIG_CORESIGHT_STM=y
  CONFIG_CORESIGHT_CPU_DEBUG=m
  root@d15263e5734a:/git/perf#
  root@d15263e5734a:/git/perf# file /tmp/build/v5.0-rc2+/drivers/hwtracing/coresight/*.o
  .../coresight/coresight-catu.o:               ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-cpu-debug.mod.o:      ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-cpu-debug.o:          ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-dynamic-replicator.o: ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-etb10.o:              ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-etm-perf.o:           ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-etm4x-sysfs.o:        ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-etm4x.o:              ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-funnel.o:             ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-replicator.o:         ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-stm.o:                ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-tmc-etf.o:            ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-tmc-etr.o:            ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-tmc.o:                ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight-tpiu.o:               ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/coresight.o:                    ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  .../coresight/of_coresight.o:                 ELF 64-bit MSB relocatable, ARM aarch64, version 1 (SYSV), not stripped
  root@d15263e5734a:/git/perf#

  root@d15263e5734a:/git/perf# pahole -C coresight_device /tmp/build/v5.0-rc2+/drivers/hwtracing/coresight/coresight.o
  struct coresight_device {
          struct coresight_connection * conns;             /*     0     8 */
          int                        nr_inport;            /*     8     4 */
          int                        nr_outport;           /*    12     4 */
          enum coresight_dev_type    type;                 /*    16     4 */
          union coresight_dev_subtype subtype;             /*    20     8 */

          /* XXX 4 bytes hole, try to pack */

          const struct coresight_ops  * ops;               /*    32     8 */
          struct device              dev;                  /*    40  1408 */

          /* XXX last struct has 7 bytes of padding */

          /* --- cacheline 22 boundary (1408 bytes) was 40 bytes ago --- */
          atomic_t *                 refcnt;               /*  1448     8 */
          bool                       orphan;               /*  1456     1 */
          bool                       enable;               /*  1457     1 */
          bool                       activated;            /*  1458     1 */

          /* XXX 5 bytes hole, try to pack */

          struct dev_ext_attribute * ea;                   /*  1464     8 */

          /* size: 1472, cachelines: 23, members: 12 */
          /* sum members: 1463, holes: 2, sum holes: 9 */
          /* paddings: 1, sum paddings: 7 */
  };
  root@d15263e5734a:/git/perf#

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-s390@vger.kernel.org
Link: http://lkml.kernel.org/r/20190131184714.20388-3-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf/aux: Make perf_event accessible to setup_aux()

When pmu::setup_aux() is called the coresight PMU needs to know which
sink to use for the session by looking up the information in the
event's attr::config2 field.

As such simply replace the cpu information by the complete perf_event
structure and change all affected customers.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Suzuki Poulouse <suzuki.poulose@arm.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-s390@vger.kernel.org
Link: http://lkml.kernel.org/r/20190131184714.20388-2-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf bpf-loader: Remove unecessary includes from bpf-loader.h

To cut the header dep tree, to get unecessary object rebuilds to be
reduced when a change happens in headers.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: https://lkml.kernel.org/n/tip-ph72xhl9moqa0g1hxcyudwfn@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf powerpc kvm-stat: Add missing evlist.h header

This header was being obtained indirectly, by sheer luck, add it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hemant Kumar <hemant@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-c3h8oyav16iu5ivput8n4wt6@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf kvm stat: Replace kvm-stat.h includes with forward declarations

To reduce the include header dependency tree and speed up perf builds.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-dngwaxuhfnhksawgdpo6e74n@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf pmu: Remove needless evsel.h include, only needs one fwd decl

To reduce the header dependency tree.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-rc389o1z0htwukqv6ni1viun@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests pmu: Add missing headers

It needs the definitions for PATH_MAX and snprintf, was getting it
by luck from headers it included and that are now being sanitized.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-7bbh3kk0h5mywvfqm64nhv28@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf hist: Remove the needless callchain.h include from hist.h

Nothing that is provided by callchain.h is used there, just things that
should've be directly included in hist.h, such as rbtree.h and a
map_symbol forward declaration.

Remove it so that we reduce the headers dependency tree.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-zivvqfx93w5zzur7hr7h0nlh@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add missing include <callchain.h> in various places

Its getting it from hist.h and that will go away, as that header doesn't
need callchain.h at all.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-6ebl3mwwiqocl79yts44qltu@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf thread: Don't include symbol.h, symbol_conf.h is enough

Also add stdio.h to get the FILE definition.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-8vx5396phynuxhdsxxfbdhsk@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: No need to include symbol.h in evsel.h, symbol_conf.h is enough

To reduce the header dependency and avoid unnecessary rebuilds when
things change in symbol.h.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-6duflwliprh2tr47w5x4t260@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add missing include for symbols.h

Several places were using definitions found in symbols.h but not
including it, getting it by sheer luck from some other headers that now
are in the process of removing that include because they don't need it
or because simply having struct forward declarations is enough, fix it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-xbcvvx296d70kpg9wb0qmeq9@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf hist: Remove symbol.h from hist.h, just fwd decls are needed

To reduce the includes dependencies.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-cmvg5ght75mmfg1efeyna9rn@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests: Add missing headers so far obtained indirectly

We're going to remove symbol.h from some places and this breaks
some of the perf tests, fix it by adding the required includes.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-wpa4b6x0btpnh2kjxzl9no4w@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf map: Move structs and prototypes for map groups to a separate header

And since machine.h only needs what is in there, make it stop including
map.h and instead include this newly introduced map_groups.h instead.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-dbob25fv5rp2rjpwlnterf38@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

pref tools: Add missing map.h includes

Lots of places get the map.h file indirectly, and since we're going to
remove it from machine.h, then those need to include it directly, do it
now, before we remove that dep.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-ob8jehdjda8h5jsrv9dqj9tf@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf symbols: Introduce map_symbol.h

To allow headers just wanting this definition to be able to get it
without all the things in symbol.h, to reduce the include dep tree.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-l32z2qyhs6fe8unf4gk2ead2@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf callchain: Uninline callchain_cursor_reset() to remove map.h dependency

That was the only thing that made including map.h in callchain.h a
requiriment, so uninline it and just add a 'struct map' forward
declaration.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-7fjz4hvv1bpzqaeriku44fn4@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf srccode: Move struct definition from map.h to srccode.h

To reduce the header dependencies, since we already have a srccode.h
header, then there is where the 'struct srccode_state' should be, and
map.h, that is more widely used should have just a forward declaraion
of 'struct srccode_state'.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-64lrkjjaa7wlo1zi2gr5u3es@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>