perf report: Support s390 diag event display on x86

Perf report fails to display s390 specific event numbered bd000
on an x86 platform. For example on s390 this works without error:

[root@m35lp76 perf]# uname -m
s390x
[root@m35lp76 perf]# ./perf record -e rbd000 -- find / >/dev/null
[ perf record: Woken up 3 times to write data ]
[ perf record: Captured and wrote 0.549 MB perf.data ]
[root@m35lp76 perf]# ./perf report -D --stdio  > /dev/null
[root@m35lp76 perf]#

Transfering this perf.data file to an x86 platform and executing
the same report command produces:

[root@f29 perf]# uname -m
x86_64
[root@f29 perf]# ./perf report -i ~/perf.data.m35lp76 --stdio
interpreting bpf_prog_info from systems with endianity is not yet supported
interpreting btf from systems with endianity is not yet supported
0x8c890 [0x8]: failed to process type: 68
Error:
failed to process sample

Event bd000 generates auxiliary data which is stored in big endian
format in the perf data file.
This error is caused by missing endianess handling on the x86 platform
when the data is displayed. Fix this by handling s390 auxiliary event
data depending on the local platform endianness.

Output after on x86:

[root@f29 perf]# ./perf report -D -i ~/perf.data.m35lp76 --stdio > /dev/null
interpreting bpf_prog_info from systems with endianity is not yet supported
interpreting btf from systems with endianity is not yet supported
[root@f29 perf]#

Committer notes:

Fix build breakage on older systems, such as CentOS:6 where using
nesting calls to the endian.h macros end up redefining local variables:

  util/s390-cpumsf.c: In function 's390_cpumsf_trailer_show':
  util/s390-cpumsf.c:333: error: declaration of '__v' shadows a previous local
  util/s390-cpumsf.c:333: error: shadowed declaration is here
  util/s390-cpumsf.c:333: error: declaration of '__x' shadows a previous local
  util/s390-cpumsf.c:333: error: shadowed declaration is here
  util/s390-cpumsf.c:334: error: declaration of '__v' shadows a previous local
  util/s390-cpumsf.c:334: error: shadowed declaration is here
  util/s390-cpumsf.c:334: error: declaration of '__x' shadows a previous local
  util/s390-cpumsf.c:334: error: shadowed declaration is here

  [perfbuilder@455a63ef60dc perf]$ gcc -v |& tail -1
  gcc version 4.4.7 20120313 (Red Hat 4.4.7-23) (GCC)
  [perfbuilder@455a63ef60dc perf]$

Since there are several uses of

  be64toh(te->flags)

Introduce a variable to hold that and then use it, avoiding this case
that causes the above problems:

  -       local.bsdes = be16toh((be64toh(te->flags) >> 16 & 0xffff));
  +       local.bsdes = be16toh((flags >> 16 & 0xffff));

Its the same construct used in s390_cpumsf_diag_show() where we have a
'word' variable that is used just once, s390_cpumsf_basic_show() has
lots of uses and also uses a variable to hold the result of be16toh().

Some of those temp variables needed to be converted from 'unsigned long'
to 'unsigned long long' so as to build on 32-bit arches such as
debian:experimental-x-mipsel, the android NDK ones and
fedora:24-x-ARC-uClibc.

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: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Link: http://lkml.kernel.org/r/20190522064325.25596-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf report: Fix OOM error in TUI mode on s390

Debugging a OOM error using the TUI interface revealed this issue
on s390:

[tmricht@m83lp54 perf]$ cat /proc/kallsyms |sort
....
00000001119b7158 B radix_tree_node_cachep
00000001119b8000 B __bss_stop
00000001119b8000 B _end
000003ff80002850 t autofs_mount [autofs4]
000003ff80002868 t autofs_show_options [autofs4]
000003ff80002a98 t autofs_evict_inode [autofs4]
....

There is a huge gap between the last kernel symbol
__bss_stop/_end and the first kernel module symbol
autofs_mount (from autofs4 module).

After reading the kernel symbol table via functions:

 dso__load()
 +--> dso__load_kernel_sym()
      +--> dso__load_kallsyms()
   +--> __dso_load_kallsyms()
        +--> symbols__fixup_end()

the symbol __bss_stop has a start address of 1119b8000 and
an end address of 3ff80002850, as can be seen by this debug statement:

  symbols__fixup_end __bss_stop start:0x1119b8000 end:0x3ff80002850

The size of symbol __bss_stop is 0x3fe6e64a850 bytes!
It is the last kernel symbol and fills up the space until
the first kernel module symbol.

This size kills the TUI interface when executing the following
code:

  process_sample_event()
    hist_entry_iter__add()
      hist_iter__report_callback()
        hist_entry__inc_addr_samples()
          symbol__inc_addr_samples(symbol = __bss_stop)
            symbol__cycles_hist()
               annotated_source__alloc_histograms(...,
                symbol__size(sym),
                                ...)

This function allocates memory to save sample histograms.
The symbol_size() marco is defined as sym->end - sym->start, which
results in above value of 0x3fe6e64a850 bytes and
the call to calloc() in annotated_source__alloc_histograms() fails.

The histgram memory allocation might fail, make this failure
no-fatal and continue processing.

Output before:
[tmricht@m83lp54 perf]$ ./perf --debug stderr=1 report -vvvvv \
      -i ~/slow.data 2>/tmp/2
[tmricht@m83lp54 perf]$ tail -5 /tmp/2
  __symbol__inc_addr_samples(875): ENOMEM! sym->name=__bss_stop,
start=0x1119b8000, addr=0x2aa0005eb08, end=0x3ff80002850,
func: 0
problem adding hist entry, skipping event
0x938b8 [0x8]: failed to process type: 68 [Cannot allocate memory]
[tmricht@m83lp54 perf]$

Output after:
[tmricht@m83lp54 perf]$ ./perf --debug stderr=1 report -vvvvv \
      -i ~/slow.data 2>/tmp/2
[tmricht@m83lp54 perf]$ tail -5 /tmp/2
   symbol__inc_addr_samples map:0x1597830 start:0x110730000 end:0x3ff80002850
   symbol__hists notes->src:0x2aa2a70 nr_hists:1
   symbol__inc_addr_samples sym:unlink_anon_vmas src:0x2aa2a70
   __symbol__inc_addr_samples: addr=0x11094c69e
   0x11094c670 unlink_anon_vmas: period++ [addr: 0x11094c69e, 0x2e, evidx=0]
    => nr_samples: 1, period: 526008
[tmricht@m83lp54 perf]$

There is no error about failed memory allocation and the TUI interface
shows all entries.

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: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/90cb5607-3e12-5167-682d-978eba7dafa8@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf test 6: Fix missing kvm module load for s390

Command

   # perf test -Fv 6

fails with error

   running test 100 'kvm-s390:kvm_s390_create_vm' failed to parse
    event 'kvm-s390:kvm_s390_create_vm', err -1, str 'unknown tracepoint'
    event syntax error: 'kvm-s390:kvm_s390_create_vm'
                         \___ unknown tracepoint

when the kvm module is not loaded or not built in.

Fix this by adding a valid function which tests if the module
is loaded. Loaded modules (or builtin KVM support) have a
directory named
  /sys/kernel/debug/tracing/events/kvm-s390
for this tracepoint.

Check for existence of this directory.

Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/20190604053504.43073-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf time-utils: Add support for multiple explicit time intervals

Currently only a single explicit time range is accepted. Add support for
multiple ranges separated by spaces, which requires the string to be
quoted. Update the time utils test accordingly.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-20-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tests: Add a test for time-utils

Test time ranges work as expected.

Committer testing:

  $ perf test "time utils"
  59: time utils                                            : Ok
  $ perf test -v "time utils"
  59: time utils                                            :
  --- start ---
  test child forked, pid 31711

  parse_nsec_time("0")
  0

  parse_nsec_time("1")
  1000000000

  parse_nsec_time("0.000000001")
  1

  parse_nsec_time("1.000000001")
  1000000001

  parse_nsec_time("123456.123456")
  123456123456000

  parse_nsec_time("1234567.123456789")
  1234567123456789

  parse_nsec_time("18446744073.709551615")
  18446744073709551615

  perf_time__parse_str("1234567.123456789,1234567.123456789")
  start time 1234567123456789, end time 1234567123456789

  perf_time__parse_str("1234567.123456789,1234567.123456790")
  start time 1234567123456789, end time 1234567123456790

  perf_time__parse_str("1234567.123456789,")
  start time 1234567123456789, end time 0

  perf_time__parse_str(",1234567.123456789")
  start time 0, end time 1234567123456789

  perf_time__parse_str("0,1234567.123456789")
  start time 0, end time 1234567123456789

  perf_time__parse_for_ranges("1234567.123456789,1234567.123456790")
  start time 1234567123456789, end time 1234567123456790

  perf_time__parse_for_ranges("10%/1")
  first_sample_time 7654321000000000 last_sample_time 7654321000000100
  start time 0: 7654321000000000, end time 0: 7654321000000009

  perf_time__parse_for_ranges("10%/2")
  first_sample_time 7654321000000000 last_sample_time 7654321000000100
  start time 0: 7654321000000010, end time 0: 7654321000000019

  perf_time__parse_for_ranges("10%/1,10%/2")
  first_sample_time 11223344000000000 last_sample_time 11223344000000100
  start time 0: 11223344000000000, end time 0: 11223344000000009
  start time 1: 11223344000000010, end time 1: 11223344000000019

  perf_time__parse_for_ranges("10%/1,10%/3,10%/10")
  first_sample_time 11223344000000000 last_sample_time 11223344000000100
  start time 0: 11223344000000000, end time 0: 11223344000000009
  start time 1: 11223344000000020, end time 1: 11223344000000029
  start time 2: 11223344000000090, end time 2: 11223344000000100

  test child finished with 0
  ---- end ----
  time utils: Ok
  $

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-19-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf time-utils: Make perf_time__parse_for_ranges() more logical

Explicit time ranges never contain a percent sign whereas percentage
ranges always do, so it is possible to call the correct parser.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-18-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf time-utils: Simplify perf_time__parse_for_ranges() error paths slightly

Simplify perf_time__parse_for_ranges() error paths slightly.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-17-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf time-utils: Fix --time documentation

Correct some punctuation and spelling and correct the format to show
that the time resolution is nanoseconds not microseconds.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-16-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf time-utils: Prevent percentage time range overlap

Prevent percentage time range overlap. This is only a 1 nanosecond
change but makes the results more logical e.g. a sample cannot be in
both the first 10% and the second 20%.

Note, there is a later patch that adds a test for time-utils.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-15-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf time-utils: Factor out set_percent_time()

Factor out set_percent_time() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-14-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf time-utils: Treat time ranges consistently

Currently, options allow only 1 explicit (non-percentage) time range.
In preparation for adding support for multiple explicit time ranges,
treat time ranges consistently.

Instead of treating some time ranges as inclusive and some as excluding
the end time, treat all time ranges as inclusive. This is only a 1
nanosecond change but is necessary to treat multiple explicit time
ranges in a consistent manner.

Note, there is a later patch that adds a test for time-utils.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-13-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Add support for efficient time interval filtering

Set up time ranges for efficient time interval filtering using the new
"fast forward" facility.

Because decoding is done in time order, intel_pt_time_filter() needs to
look only at the next start or end timestamp - refer intel_pt_next_time().

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-12-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Add support for lookahead

Implement the lookahead callback to let the decoder access subsequent
buffers. intel_pt_lookahead() manages the buffer lifetime and calls the
decoder for each buffer until the decoder returns a non-zero value.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-11-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Factor out intel_pt_get_buffer()

Factor out intel_pt_get_buffer() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-10-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Add intel_pt_fast_forward()

Intel PT decoding is done in time order. In order to support efficient time
interval filtering, add a facility to "fast forward" towards a particular
timestamp. That involves finding the right buffer, stepping to that buffer,
and then stepping forward PSBs. Because decoding must begin at a PSB,
"fast forward" stops at the last PSB that has a timestamp before the target
timestamp.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-9-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Add reposition parameter to intel_pt_get_data()

When the decoder gets the next trace buffer, some state is reset if the
buffer is not consecutive to the previous buffer. Add a parameter
'reposition' so that can be done also to support a "fast forward"
facility.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-8-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Factor out intel_pt_reposition()

Factor out intel_pt_reposition() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-7-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Factor out intel_pt_8b_tsc()

Factor out intel_pt_8b_tsc() so it can be reused.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-6-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Add lookahead callback

Add a callback function to enable the decoder to lookahead at subsequent
trace buffers. This will be used to implement a "fast forward" facility
which will be needed to support efficient time interval filtering.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-5-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf report: Set perf time interval in itrace_synth_ops

Instruction trace decoders can optimize output based on what time
intervals will be filtered, so pass that information in
itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-4-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf script: Set perf time interval in itrace_synth_ops

Instruction trace decoders can optimize output based on what time
intervals will be filtered, so pass that information in
itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-3-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf auxtrace: Add perf time interval to itrace_synth_ops

Instruction trace decoders can optimize output based on what time
intervals will be filtered, so pass that information in
itrace_synth_ops.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190604130017.31207-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf config: Update default value for llvm.clang-bpf-cmd-template

The clang bpf cmdline template has defined default value in the file
tools/perf/util/llvm-utils.c, which has been changed for several times.

This patch updates the documentation to reflect the latest default value
for the configuration llvm.clang-bpf-cmd-template.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Drayton <mbd@fb.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Cc: bpf@vger.kernel.org
Cc: netdev@vger.kernel.org
Fixes: d35b168c3dcd ("perf bpf: Give precedence to bpf header dir")
Fixes: cb76371441d0 ("perf llvm: Allow passing options to llc in addition to clang")
Fixes: 1b16fffa389d ("perf llvm-utils: Add bpf include path to clang command line")
Link: http://lkml.kernel.org/r/20190607143508.18141-1-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Remove duplicate GENMASK() define, use linux/bits.h instead

Suzuki noticed that this should be more useful in a generic header, and
after looking I noticed we have it already in our copy of
include/linux/bits.h in tools/include, so just use it, test built on
x86-64 and ubuntu 19.04 with:

  perfbuilder@46646c9e848e:/$ aarch64-linux-gnu-gcc --version |& head -1
  aarch64-linux-gnu-gcc (Ubuntu/Linaro 8.3.0-6ubuntu1) 8.3.0
  perfbuilder@46646c9e848e:/$

Suggested-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Link: https://lkml.kernel.org/r/68c1c548-33cd-31e8-100d-7ffad008c7b2@arm.com
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org,
Link: https://lkml.kernel.org/n/tip-69pd3mqvxdlh2shddsc7yhyv@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Properly set the value of 'old' and 'head' in snapshot mode

This patch adds the necessary intelligence to properly compute the value
of 'old' and 'head' when operating in snapshot mode.  That way we can
get the latest information in the AUX buffer and be compatible with the
generic AUX ring buffer mechanic.

Tester notes:

> Leo, have you had the chance to test/review this one? Suzuki?

Sure.  I applied this patch on the perf/core branch (with latest
commit 3e4fbf36c1e3 'perf augmented_raw_syscalls: Move reading
filename to the loop') and passed testing with below steps:

  # perf record -e cs_etm/@tmc_etr0/ -S -m,64 --per-thread ./sort &
  [1] 19097
  Bubble sorting array of 30000 elements

  # kill -USR2 19097
  # kill -USR2 19097
  # kill -USR2 19097
  [ perf record: Woken up 4 times to write data ]
  [ perf record: Captured and wrote 0.753 MB perf.data ]

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190605161633.12245-1-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf data: Fix perf.data documentation for HEADER_CPU_TOPOLOGY

The 'die' info isn't in the same array as core and socket ids, and we
missed the 'dies' string list, that comes right after the 'core' +
'socket' id variable length array, followed by the VLA for the dies.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Fixes: c9cb12c5ba08 ("perf header: Add die information in CPU topology")
Link: https://lkml.kernel.org/n/tip-nubi6mxp2n8ofvlx7ph6k3h6@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Apply new CPU topology sysfs attributes

The existing "thread_siblings" and "thread_siblings_list" attribute will
be deprecated.

Use the new CPU topology sysfs attributes, "core_cpus" and
"core_cpus_list", which are synonymous with the deprecated attributes.

Check the new name first. If not available, use the deprecated name to
be compatible with old kernel.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1559688644-106558-5-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf header: Rename "sibling cores" to "sibling sockets"

The "sibling cores" actually shows the sibling CPUs of a socket.  The
name "sibling cores" is very misleading.

Rename "sibling cores" to "sibling sockets"

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1559688644-106558-4-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf stat: Support per-die aggregation

It is useful to aggregate counts per die. E.g. Uncore becomes die-scope
on Xeon Cascade Lake-AP.

Introduce a new option "--per-die" to support per-die aggregation.

The global id for each core has been changed to socket + die id + core
id. The global id for each die is socket + die id.

Add die information for per-core aggregation. The output of per-core
aggregation will be changed from "S0-C0" to "S0-D0-C0". Any scripts
which rely on the output format of per-core aggregation probably be
broken.

For 'perf stat record/report', there is no die information when
processing the old perf.data. The per-die result will be the same as
per-socket.

Committer notes:

Renamed 'die' variable to 'die_id' to fix the build in some systems:

    CC       /tmp/build/perf/builtin-script.o
  cc1: warnings being treated as errors
  builtin-stat.c: In function 'perf_env__get_die':
  builtin-stat.c:963: error: declaration of 'die' shadows a global declaration
  util/util.h:19: error: shadowed declaration is here
  mv: cannot stat `/tmp/build/perf/.builtin-stat.o.tmp': No such file or directory

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lkml.kernel.org/n/tip-bsnhx7vgsuu6ei307mw60mbj@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf header: Add die information in CPU topology

With the new CPUID.1F, a new level type of CPU topology, 'die', is
introduced. The 'die' information in CPU topology should be added in
perf header.

To be compatible with old perf.data, the patch checks the section size
before reading the die information. The new info is added at the end of
the cpu_topology section, the old perf tool ignores the extra data.  It
never reads data crossing the section boundary.

The new perf tool with the patch can be used on legacy kernel. Add a new
function has_die_topology() to check if die topology information is
supported by kernel. The function only check X86 and CPU 0. Assuming
other CPUs have same topology.

Use similar method for core and socket to support die id and sibling
dies string.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1559688644-106558-2-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cpumap: Retrieve die id information

There is no function to retrieve die id information of a given CPU.

Add cpu_map__get_die_id() to retrieve die id information.

Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1559688644-106558-1-git-send-email-kan.liang@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add support for CPU-wide trace scenarios

Add support for CPU-wide trace scenarios by correlating range packets
with timestamp packets.  That way range packets received on different
ETMQ/traceID channels can be processed and synthesized in chronological
order.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-18-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add notion of time to decoding code

This patch deals with timestamp packets received from the decoding
library in order to give the front end packet processing loop a handle
on the time instruction conveyed by range packets have been executed at.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-17-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Linking PE contextID with perf thread mechanic

Link contextID packets received from the decoder with the perf tool
thread mechanic so that we know the specifics of the process currently
executing.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-16-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add support for multiple traceID queues

When operating in CPU-wide trace mode with a source/sink topology of N:1
packets with multiple traceID will end up in the same cs_etm_queue.  In
order to properly decode packets they need to be split in different
queues, i.e one queue per traceID.

As such add support for multiple traceID per cs_etm_queue by adding a
new cs_etm_traceid_queue every time a new traceID is discovered in the
trace stream.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-15-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Use traceID aware memory callback API

When working with CPU-wide traces different traceID may be found in the
same stream.  As such we need to use the decoder callback that provides
the traceID in order to know the thread context being decoded.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-14-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Move tid/pid to traceid_queue

The tid/pid fields of structure cs_etm_queue are CPU dependent and as
such need to be part of the cs_etm_traceid_queue in order to support
CPU-wide trace scenarios.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-13-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Move thread to traceid_queue

The thread field of structure cs_etm_queue is CPU dependent and as such
need to be part of the cs_etm_traceid_queue in order to support CPU-wide
trace scenarios.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-12-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Get rid of unused cpu in struct cs_etm_queue

Nowadays the synthesize code is using the packet's cpu information,
making cs_etm_queue::cpu useless.  As such simply remove it.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-11-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Introduce the concept of trace ID queues

In an ideal world there is one CPU per cs_etm_queue and as such, one
trace ID per cs_etm_queue.  In the real world CoreSight topologies allow
multiple CPUs to use the same sink, which translates to multiple trace
IDs per cs_etm_queue.

To deal with this a new cs_etm_traceid_queue structure is introduced to
enclose all the information related to a single trace ID, allowing a
cs_etm_queue to handle traces generated by any number of CPUs.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-10-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Fix indentation in function cs_etm__process_decoder_queue()

Fixing wrong indentation of the while() loop - no change of
functionality.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Fixes: 3fa0e83e2948 ("perf cs-etm: Modularize main packet processing loop")
Link: http://lkml.kernel.org/r/20190524173508.29044-9-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Move packet queue out of decoder structure

The decoder needs to work with more than one traceID queue if we want to
support CPU-wide scenarios with N:1 source/sink topologies.  As such
move the packet buffer and related fields out of the decoder structure
and into the cs_etm_queue structure.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-8-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Refactor error path in cs_etm_decoder__new()

There is no point in having two different error goto statement since the
openCSD API to free a decoder handles NULL pointers.  As such function
cs_etm_decoder__free() can be called to deal with all aspect of freeing
decoder memory.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-7-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add handling of switch-CPU-wide events

Add handling of SWITCH-CPU-WIDE events in order to add the tid/pid of
the incoming process to the perf tools machine infrastructure.  This
information is later retrieved when a contextID packet is found in the
trace stream.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-6-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Add handling of itrace start events

Add handling of ITRACE events in order to add the tid/pid of the
executing process to the perf tools machine infrastructure.  This
information is later retrieved when a contextID packet is found in the
trace stream.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-5-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Configure SWITCH_EVENTS in CPU-wide mode

Ask the perf core to generate an event when processes are swapped in/out
of context.  That way proper action can be taken by the decoding code
when faced with such event.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-4-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Configure timestamp generation in CPU-wide mode

When operating in CPU-wide mode tracers need to generate timestamps in
order to correlate the code being traced on one CPU with what is executed
on other CPUs.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-3-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf cs-etm: Configure contextID tracing in CPU-wide mode

When operating in CPU-wide mode being notified of contextID changes is
required so that the decoding mechanic is aware of the process context
switch.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Reviewed-by: Suzuki Poulouse <suzuki.poulose@arm.com>
Tested-by: Leo Yan <leo.yan@linaro.org>
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>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: coresight@lists.linaro.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/20190524173508.29044-2-mathieu.poirier@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf evsel: Remove superfluous nthreads system_wide setup in alloc_fd()

It's already setup in the only caller of this method in
perf_evsel__open(), right before calling perf_evsel__alloc_fd(), no need
to do it again.

Also it's better to have it out of the function before we move it to
libperf.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
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-1k8lhyjxfk7o8v4g3r7eyjc9@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Add support to collect callchains from kernel or user space only

One can just record callchains in the kernel or user space with this new
options.

We can use it together with "--all-kernel" options.

This two options is used just like print_stack(sys) or print_ustack(usr)
for systemtap.

Shown below is the usage of this new option combined with "--all-kernel"
options:

1. Configure all used events to run in kernel space and just collect
   kernel callchains.

  $ perf record -a -g --all-kernel --kernel-callchains

2. Configure all used events to run in kernel space and just collect
   user callchains.

  $ perf record -a -g --all-kernel --user-callchains

Committer notes:

Improved documentation to state that asking for kernel callchains really
is asking for excluding user callchains, and vice versa.

Further mentioned that using both won't get both, but nothing, as both
will be excluded.

Signed-off-by: yuzhoujian <yuzhoujian@didichuxing.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1559222962-22891-1-git-send-email-ufo19890607@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf config: Bail out when a handler returns failure for a key-value pair

So perf_config() uses:

  int ret = 0;

  perf_config_set__for_each_entry(config_set, section, item) {
          ...
          ret = fn();
          if (ret < 0)
                  break;
  }

  return ret;

Expecting that that break will imediatelly go to function exit to return
that error value (ret).

The problem is that perf_config_set__for_each_entry() expands into two
nested for() loops, one traversing the sections in a config and the
second the items in each of those sections, so we have to change that
'break' to a goto label right before that final 'return ret'.

With that, for instance 'perf trace' now correctly bails out when a
event that is requested to be added via its 'trace.add_events'
~/.perfconfig entry gets rejected by the kernel BPF verifier:

  # perf trace ls
  event syntax error: '/home/acme/git/perf/tools/perf/examples/bpf/augmented_raw_syscalls.o'
                       \___ Kernel verifier blocks program loading

  (add -v to see detail)
  Run 'perf list' for a list of valid events
  Error: wrong config key-value pair trace.add_events=/home/acme/git/perf/tools/perf/examples/bpf/augmented_raw_syscalls.o
  #

While before it would continue and explode later, when trying to find
maps that would have been in place had that augmented_raw_syscalls.o
precompiled BPF proggie been accepted by the, humm, bast... rigorous
kernel BPF verifier 8-)

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Taeung Song <treeze.taeung@gmail.com>
Cc: Yonghong Song <yhs@fb.com>
Fixes: 8a0a9c7e9146 ("perf config: Introduce new init() and exit()")
Link: https://lkml.kernel.org/n/tip-qvqxfk9d0rn1l7lcntwiezrr@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Exit when failing to build eBPF program

On my Juno board with ARM64 CPUs, perf trace command reports the eBPF
program building failure but the command will not exit and continue to
run.  If we define an eBPF event in config file, the event will be
parsed with below flow:

  perf_config()
    `> trace__config()
         `> parse_events_option()
              `> parse_events__scanner()
                   `-> parse_events_parse()
                         `> parse_events_load_bpf()
                              `> llvm__compile_bpf()

Though the low level functions return back error values when detect eBPF
building failure, but parse_events_option() returns 1 for this case and
trace__config() passes 1 to perf_config(); perf_config() doesn't treat
the returned value 1 as failure and it continues to parse other
configurations.  Thus the perf command continues to run even without
enabling eBPF event successfully.

This patch changes error handling in trace__config(), when it detects
failure it will return -1 rather than directly pass error value (1);
finally, perf_config() will directly bail out and perf will exit for
this case.

Committer notes:

Simplified the patch to just check directly the return of
parse_events_option() and it it is non-zero, change err from its initial
zero value to -1.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Mathieu Poirier <mathieu.poirier@linaro.org>
Cc: Mike Leach <mike.leach@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Suzuki Poulouse <suzuki.poulose@arm.com>
Cc: Yonghong Song <yhs@fb.com>
Fixes: ac96287cae08 ("perf trace: Allow specifying a set of events to add in perfconfig")
Link: https://lkml.kernel.org/n/tip-x4i63f5kscykfok0hqim3zma@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Associate more argument names with the filename beautifier

For instance, the rename* family uses "oldname", "newname", so check if
"name" is at the end and treat it as a filename.

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-wjy7j4bk06g7atzwoz1mid24@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf trace: Consume the augmented_raw_syscalls payload

To support the SCA_FILENAME beautifier in more than one syscall arg, as
needed for syscalls such as the rename* family, we need to, after
processing one such arg, bump the augmented pointers so that the next
augmented arg don't reuse data for the previous augmented arguments.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-4e4cmzyjxb3wkonfo1x9a27y@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf jvmti: Address gcc string overflow warning for strncpy()

We are getting false positive gcc warning when we compile with gcc9 (9.1.1):

     CC       jvmti/libjvmti.o
   In file included from /usr/include/string.h:494,
                    from jvmti/libjvmti.c:5:
   In function ‘strncpy’,
       inlined from ‘copy_class_filename.constprop’ at jvmti/libjvmti.c:166:3:
   /usr/include/bits/string_fortified.h:106:10: error: ‘__builtin_strncpy’ specified bound depends on the length of the source argument [-Werror=stringop-overflow=]
     106 |   return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
         |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   jvmti/libjvmti.c: In function ‘copy_class_filename.constprop’:
   jvmti/libjvmti.c:165:26: note: length computed here
     165 |   size_t file_name_len = strlen(file_name);
         |                          ^~~~~~~~~~~~~~~~~
   cc1: all warnings being treated as errors

As per Arnaldo's suggestion use strlcpy(), which does the same thing and keeps
gcc silent.

Suggested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ben Gainey <ben.gainey@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/20190531131321.GB1281@krava
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf augmented_raw_syscalls: Move reading filename to the loop

Almost there, next step is to copy more than one filename payload.

Probably to read syscall arg structs, etc we'll need just a variation of
this that will decide what to use, if probe_read_str() or plain
probe_read for structs, i.e. fixed size.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-uf6u0pld6xe4xuo16f04owlz@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf augmented_raw_syscalls: Change helper to consider just the augmented_filename part

So that we can use it for multiple args, baby steps not to step into the
verifier toes.

In the process make sure we handle -EFAULT from bpf_prog_read_str(), as
this really is needed now that we'll handle more than one augmented
argument, i.e. if there is failure, then we have the argument that fails
have:

  (size = 0, err = -EFAULT, value = [] )

followed by the next, lets say that worked for a second pathname:

  (size = 4, err = 0, value = "/tmp" )

So we can skip the first while telling the user about the problem and
then process the second.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-deyvqi39um6gp6hux6jovos8@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf augmented_raw_syscalls: Move the probe_read_str to a separate function

One more step into copying multiple filenames to support syscalls like
rename*.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Link: https://lkml.kernel.org/n/tip-xdqtjexdyp81oomm1rkzeifl@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf augmented_raw_syscalls: Tell which args are filenames and how many bytes to copy

Since we know what args are strings from reading the syscall
descriptions in tracefs and also already mark such args to be beautified
using the syscall_arg__scnprintf_filename() helper, all we need is to
fill in this info in the 'syscalls' BPF map we were using to state which
syscalls the user is interested in, i.e. the syscall filter.

Right now just set that with PATH_MAX and unroll the syscall arg in the
BPF program, as the verifier isn't liking something clang generates when
unrolling the loop.

This also makes the augmented_raw_syscalls.c program support all arches,
since we removed that set of defines with the hard coded syscall
numbers, all should be automatically set for all arches, with the
syscall id mapping done correcly.

Doing baby steps here, i.e. just the first string arg for a syscall is
printed, syscalls with more than one, say, the various rename* syscalls,
need further work, but lets get first something that the BPF verifier
accepts before increasing the complexity

To test it, something like:

 # perf trace -e string -e /home/acme/git/perf/tools/perf/examples/bpf/augmented_raw_syscalls.c

With:

  # cat ~/.perfconfig
  [llvm]
dump-obj = true
clang-opt = -g
  [trace]
#add_events = /home/acme/git/perf/tools/perf/examples/bpf/augmented_raw_syscalls.c
show_zeros = yes
show_duration = no
no_inherit = yes
show_timestamp = no
show_arg_names = no
args_alignment = 40
show_prefix = yes
  #

That commented add_events line is needed for developing this
augmented_raw_syscalls.c BPF program, as if we add it via the
'add_events' mechanism so as to shorten the 'perf trace' command lines,
then we end up not setting up the -v option which precludes us having
access to the bpf verifier log :-\

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexei Starovoitov <ast@fb.com>
Cc: Andrii Nakryiko <andriin@fb.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Leo Yan <leo.yan@linaro.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Link: https://lkml.kernel.org/n/tip-dn863ya0cbsqycxuy0olvbt1@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts python: exported-sql-viewer.py: Select find text when find bar is activated

The user probably wants to replace the find text, so select the find
text when the find bar is activated.

That is fairly standard behaviour for search text entry.

Entering text will replace the current text, but using edit keys
(arrows, home, end etc) cancels the selection and enables editing.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-23-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts python: exported-sql-viewer.py: Add IPC information to Call Tree

Enhance the call tree to display IPC information if it is available.

Committer testing:

[acme@quaco adrian.hunter]$ python ~acme/libexec/perf-core/scripts/python/exported-sql-viewer.py ~/c/adrian.hunter/simple-retpoline.db

Reports -> Call Tree, then expand a few trees, then select with the
mouse and press control+C (copy):

Call Path                   Object        Call Time Time  Time(%) Insn  Insn   Cyc   Cyc   IPC Branch Branch
▼ simple-retpolin                                   (ns)          Cnt   Cnt(%) Cnt   Cnt(%)     Count Count(%)
  ▼ 23003:23003
    ▼ _start                ld-2.28.so    112195670 218295 100.0 127746 100.0 207320 100.0 0.62 13046 100.0
      ▶ unknown             unknown       112195987   3202   1.5      0   0.0      0   0.0    0     1   0.0
      ▶ _dl_start           ld-2.28.so    112199189 188471  86.3 123394  96.6 180007  86.8 0.69 12529  96.0
      ▼ _dl_init            ld-2.28.so    112387660  13406   6.1   3207   2.5  14868   7.2 0.22   327   2.5
        ▶ call_init.part.0  ld-2.28.so    112387773    117   0.9     70   2.2    639   4.3 0.11     3   0.9
        ▶ call_init.part.0  ld-2.28.so    112387890  13129  97.9   3103  96.8  14100  94.8 0.22   315  96.3
        ▶ call_init.part.0  ld-2.28.so    112401020      0   0.0      0   0.0      0   0.0    0     2   0.6
      ▼ _start              simple-retpol 112401066  12899   5.9   1142   0.9  11561   5.6 0.10   184   1.4
        ▶ unknown           unknown       112401388    846   6.6      0   0.0      0   0.0    0     1   0.5
        ▼ __libc_start_main libc-2.28.so  112402344  11621  90.1   1129  98.9  10350  89.5 0.11   181  98.4
          ▶ __cxa_atexit    libc-2.28.so  112402360   2302  19.8    101   8.9   1817  17.6 0.06    13   7.2
          ▶ __libc_csu_init simple-retpol 112404673    121   1.0     43   3.8    340   3.3 0.13     8   4.4
          ▶ _setjmp         libc-2.28.so  112404794     74   0.6     46   4.1    206   2.0 0.22     4   2.2
          ▼ main            simple-retpol 112404892     44   0.4     23   2.0    126   1.2 0.18    12   6.6
            ▼ foo           simple-retpol 112404892     19  43.2     12  52.2     55  43.7 0.22     5  41.7
                bar         simple-retpol 112404896     12  63.2      3  25.0     34  61.8 0.09     1  20.0
            ▼ foo           simple-retpol 112404911     25  56.8     11  47.8     71  56.3 0.15     5  41.7
              ▶ bar         simple-retpol 112404924     10  40.0      3  27.3     27  38.0 0.11     1  20.0
          ▶ exit            libc-2.28.so  112404936   9029  77.7    878  77.8   7765  75.0 0.11   139  76.8

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-22-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts python: exported-sql-viewer.py: Add IPC information to Call Graph Graph

Enhance the call graph to display IPC information if it is available.

Committer testing:

[acme@quaco adrian.hunter]$ python ~acme/libexec/perf-core/scripts/python/exported-sql-viewer.py ~/c/adrian.hunter/simple-retpoline.db

Reports -> Context Sensitive Callgraph, then expand a few trees, then
select with the mouse and press control+C:

Call Path                     Object          Count Time(ns) Time(%) Insn Insn   Cyc   Cyc    IPC Branch Branch
▼ simple-retpolin                                                    Cnt  Cnt(%) Cnt   Cnt(%)     Cnt    Cnt(%)
  ▼ 23003:23003
    ▼ _start                  ld-2.28.so         1 218295   100.0  127746 100.0 207320 100.0 0.62 13046  100.0
      ▶ unknown               unknown            1   3202     1.5       0   0.0      0   0.0    0     1    0.0
      ▶ _dl_start             ld-2.28.so         1 188471    86.3  123394  96.6 180007  86.8 0.69 12529   96.0
      ▶ _dl_init              ld-2.28.so         1  13406     6.1    3207   2.5  14868   7.2 0.22   327    2.5
      ▼ _start                simple-retpoline   1  12899     5.9    1142   0.9  11561   5.6 0.10   184    1.4
        ▶ unknown             unknown            1    846     6.6       0   0.0      0   0.0    0     1    0.5
        ▼ __libc_start_main   libc-2.28.so       1  11621    90.1    1129  98.9  10350  89.5 0.11   181   98.4
          ▶ __cxa_atexit      libc-2.28.so       1   2302    19.8     101   8.9   1817  17.6 0.06    13    7.2
          ▶ __libc_csu_init   simple-retpoline   1    121     1.0      43   3.8    340   3.3 0.13     8    4.4
          ▼ _setjmp           libc-2.28.so       1     74     0.6      46   4.1    206   2.0 0.22     4    2.2
            ▼ __sigsetjmp     libc-2.28.so       1     74   100.0      46 100.0    206 100.0 0.22     3   75.0
              ▶ __sigjmp_save libc-2.28.so       1      0     0.0       0   0.0      0   0.0    0     1   33.3
          ▼ main              simple-retpoline   1     44     0.4      23   2.0    126   1.2 0.18    12    6.6
            ▼ foo             simple-retpoline   2     44   100.0      23 100.0    126 100.0 0.18    10   83.3
                bar           simple-retpoline   2     22    50.0       6  26.1     61  48.4 0.10     2   20.0
          ▶ exit              libc-2.28.so       1   9029    77.7     878  77.8   7765  75.0 0.11   139   76.8

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-21-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts python: exported-sql-viewer.py: Add CallGraphModelParams

Add a parameter to call graph and call tree, to determine whether IPC
information is available.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-20-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts python: exported-sql-viewer.py: Add IPC information to the Branch reports

Enhance the "All branches" and "Selected branches" reports to display IPC
information if it is available.

Committer testing:

So, testing this I noticed that it all starts with the left arrow in every
line, that should mean there is some tree there, i.e. look at all those ▶
symbols:

Reports -> All Branches:

Time              CPU Command         PID   TID   Branch Type  In Tx  Insn Cnt  Cyc Cnt  IPC  Branch
▶ 187836112195670 7   simple-retpolin 23003 23003 trace begin  No     0         0        0               0 unknown (unknown) -> 7f6f33d4f110
+_start (ld-2.28.so)
▶ 187836112195987 7   simple-retpolin 23003 23003 trace end    No     0         883      0    7f6f33d4f110 _start (ld-2.28.so) -> 0 unknown
+(unknown)
▶ 187836112199189 7   simple-retpolin 23003 23003 trace begin  No     0         0        0               0 unknown (unknown) -> 7f6f33d4f110
+_start (ld-2.28.so)
▶ 187836112199189 7   simple-retpolin 23003 23003 call         No     0         0        0    7f6f33d4f113 _start+0x3 (ld-2.28.so) -> 7f6f33d4ff50
+_dl_start (ld-2.28.so)
▶ 187836112199544 7   simple-retpolin 23003 23003 trace end    No     17        996      0.02 7f6f33d4ff73 _dl_start+0x23 (ld-2.28.so) -> 0
+unknown (unknown)
▶ 187836112200939 7   simple-retpolin 23003 23003 trace begin  No     0         0        0               0 unknown (unknown) -> 7f6f33d4ff73
+_dl_start+0x23 (ld-2.28.so)
▶ 187836112201229 7   simple-retpolin 23003 23003 trace end    No     1         816      0.00 7f6f33d4ff7a _dl_start+0x2a (ld-2.28.so) -> 0
+unknown (unknown)
▶ 187836112203500 7   simple-retpolin 23003 23003 trace begin  No     0         0        0               0 unknown (unknown) -> 7f6f33d4ff7a
+_dl_start+0x2a (ld-2.28.so)

But if you click on it, that ▶ disappears and a new click doesn't make
it reappear, looks buggy, minor oddity, reported to Adrian.

Reports -> Selected Branches, then ask for branches in the ld-2.28.so
DSO:

Time               CPU  Command          PID    TID    Branch Type        In Tx  Insn Cnt  Cyc Cnt  IPC   Branch
▶ 187836112195987  7    simple-retpolin  23003  23003  trace end          No     0         883      0     7f6f33d4f110 _start (ld-2.28.so) -> 0 unknown (unknown)
▶ 187836112199189  7    simple-retpolin  23003  23003  trace begin        No     0         0        0                0 unknown (unknown) -> 7f6f33d4f110 _start (ld-2.28.so)
▶ 187836112199189  7    simple-retpolin  23003  23003  call               No     0         0        0     7f6f33d4f113 _start+0x3 (ld-2.28.so) -> 7f6f33d4ff50 _dl_start (ld-2.28.so)
▶ 187836112199544  7    simple-retpolin  23003  23003  trace end          No     17        996      0.02  7f6f33d4ff73 _dl_start+0x23 (ld-2.28.so) -> 0 unknown (unknown)
▶ 187836112200939  7    simple-retpolin  23003  23003  trace begin        No     0         0        0                0 unknown (unknown) -> 7f6f33d4ff73 _dl_start+0x23 (ld-2.28.so)
▶ 187836112201229  7    simple-retpolin  23003  23003  trace end          No     1         816      0.00  7f6f33d4ff7a _dl_start+0x2a (ld-2.28.so) -> 0 unknown (unknown)
▶ 187836112203500  7    simple-retpolin  23003  23003  trace begin        No     0         0        0                0 unknown (unknown) -> 7f6f33d4ff7a _dl_start+0x2a (ld-2.28.so)
▶ 187836112203528  7    simple-retpolin  23003  23003  unconditional jump No     0         0        0     7f6f33d4ffe7 _dl_start+0x97 (ld-2.28.so) -> 7f6f33d5000b _dl_start+0xbb (ld-2.28.so)
▶ 187836112203528  7    simple-retpolin  23003  23003  conditional jump   No     0         0        0     7f6f33d5000f _dl_start+0xbf (ld-2.28.so) -> 7f6f33d4fffb _dl_start+0xab (ld-2.28.so)
▶ 187836112203528  7    simple-retpolin  23003  23003  conditional jump   No     0         0        0     7f6f33d5000f _dl_start+0xbf (ld-2.28.so) -> 7f6f33d4fffb _dl_start+0xab (ld-2.28.so)
▶ 187836112203539  7    simple-retpolin  23003  23003  conditional jump   No     0         0        0     7f6f33d50025 _dl_start+0xd5 (ld-2.28.so) -> 7f6f33d50210 _dl_start+0x2c0 (ld-2.28.so)
▶ 187836112203539  7    simple-retpolin  23003  23003  conditional jump   No     0         0        0     7f6f33d5021a _dl_start+0x2ca (ld-2.28.so) -> 7f6f33d50360 _dl_start+0x410 (ld-2.28.so)
▶ 187836112203539  7    simple-retpolin  23003  23003  unconditional jump No     0         0        0     7f6f33d50377 _dl_start+0x427 (ld-2.28.so) -> 7f6f33d4ffff _dl_start+0xaf (ld-2.28.so)
▶ 187836112203539  7    simple-retpolin  23003  23003  conditional jump   No     0         0        0     7f6f33d5000f _dl_start+0xbf (ld-2.28.so) -> 7f6f33d4fffb _dl_start+0xab (ld-2.28.so)
▶ 187836112203562  7    simple-retpolin  23003  23003  conditional jump   No     0         0        0     7f6f33d5000f _dl_start+0xbf (ld-2.28.so) -> 7f6f33d4fffb _dl_start+0xab (ld-2.28.so)

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-19-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts python: export-to-postgresql.py: Export IPC information

Export cycle and instruction counts on samples and calls tables.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-18-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf scripts python: export-to-sqlite.py: Export IPC information

Export cycle and instruction counts on samples and calls tables.

Committer testing:

First runs some workload collecting intel_pt with the 'cyc' ter just for
userspace:

  [root@quaco adrian.hunter]# perf record -o simple-retpoline.perf.data -e intel_pt/cyc/u ./simple-retpoline
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.035 MB simple-retpoline.perf.data ]
  [root@quaco adrian.hunter]#

Then use the export-to-sqlite.py script to see if the changes in this
cset don't make it to break and if the changes in the db schema are the
ones expected:

  [root@quaco adrian.hunter]# perf script -i simple-retpoline.perf.data --itrace=be -s ~acme/libexec/perf-core/scripts/python/export-to-sqlite.py simple-retpoline.db branches calls
  2019-05-31 11:50:46.942710 Creating database ...
  2019-05-31 11:50:46.949663 Writing records...
  2019-05-31 11:50:47.224033 Adding indexes
  2019-05-31 11:50:47.231599 Done
  [root@quaco adrian.hunter]#

Now lets use the db:

  [root@quaco adrian.hunter]# sqlite3 simple-retpoline.db
  SQLite version 3.26.0 2018-12-01 12:34:55
  Enter ".help" for usage hints.
  sqlite> .schema samples
  CREATE TABLE samples (id integer NOT NULL PRIMARY KEY,evsel_id bigint,machine_id bigint,thread_id bigint,comm_id bigint,dso_id bigint,symbol_id bigint,sym_offset bigint,ip bigint,time bigint,cpuinteger,to_dso_id bigint,to_symbol_id bigint,to_sym_offset bigint,to_ip bigint,branch_type integer,in_tx boolean,call_path_id bigint,insn_count bigint,cyc_count bigint);
  sqlite>

Cool, the 'insn_count' and 'cyc_count' are there, now lets see if we can
use them in a query:

  sqlite> select insn_count,cyc_count from samples where cyc_count > 1500 and insn_count < 10;
  6|1507
  sqlite> select insn_count,cyc_count from samples where cyc_count > 1500;
  118|2210
  140|1516
  3783|1861
  132|1521
  6|1507
  sqlite>

Seems to work :-)

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-17-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf db-export: Export IPC information

Export cycle and instruction counts on samples and call-returns.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-16-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf db-export: Add brief documentation

Add brief documentation to explain how the database export maintains
backward and forward compatibility.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-15-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf thread-stack: Accumulate IPC information

Cycle and instruction counts are added to the stack. The IPC of a
function and all functions it calls, is also recorded.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-14-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Document IPC usage

Add brief documentation about instructions-per-cycle (IPC) information
derived from Intel PT.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-13-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Accumulate cycle count from TSC/TMA/MTC packets

When CYC packets are not available, it is still possible to count cycles
using TSC/TMA/MTC timestamps.

As the timestamp increments in TSC ticks, convert to CPU cycles using
the current core-to-bus ratio.

Do not accumulate cycles when control flow packet generation is not
enabled, nor when time has been "lost", typically due to mwait, which is
indicated by a TSC/TMA packet that is not part of PSB+.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-12-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Re-factor TIP cases in intel_pt_walk_to_ip

To make it easier to add new code for different TIP cases, separate each
case.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-11-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Record when decoding PSB+ packets

In preparation for using MTC packets to count cycles, record whether
decoding is between a PSB and PSBEND packets.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-10-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf script: Add output of IPC ratio

Add field 'ipc' to display instructions-per-cycle.

Example:

 perf record -e intel_pt/cyc/u ls
 perf script --insn-trace --xed -F+ipc,-dso,-cpu,-tid

 ls  2670177.697113434:  7f0dfdbcd090 _start+0x0      mov %rsp, %rdi   IPC: 0.00 (1/877)
 ls  2670177.697113434:  7f0dfdbcd093 _start+0x3      callq  0x7f0dfdbce030
 ls  2670177.697113434:  7f0dfdbce030 _dl_start+0x0   pushq  %rbp
 ls  2670177.697113434:  7f0dfdbce031 _dl_start+0x1   mov %rsp, %rbp
 ls  2670177.697113434:  7f0dfdbce034 _dl_start+0x4   pushq  %r15
 ls  2670177.697113434:  7f0dfdbce036 _dl_start+0x6   pushq  %r14
 ls  2670177.697113434:  7f0dfdbce038 _dl_start+0x8   pushq  %r13
 ls  2670177.697113434:  7f0dfdbce03a _dl_start+0xa   pushq  %r12
 ls  2670177.697113434:  7f0dfdbce03c _dl_start+0xc   mov %rdi, %r12
 ls  2670177.697113434:  7f0dfdbce03f _dl_start+0xf   pushq  %rbx
 ls  2670177.697113434:  7f0dfdbce040 _dl_start+0x10  sub $0x38, %rsp
 ls  2670177.697113434:  7f0dfdbce044 _dl_start+0x14  rdtsc
 ls  2670177.697113434:  7f0dfdbce046 _dl_start+0x16  mov %eax, %eax
 ls  2670177.697113434:  7f0dfdbce048 _dl_start+0x18  shl $0x20, %rdx
 ls  2670177.697113434:  7f0dfdbce04c _dl_start+0x1c  or %rax, %rdx
 ls  2670177.697114471:  7f0dfdbce04f _dl_start+0x1f  movq  0x27e22(%rip), %rax        IPC: 0.00 (15/1685)
 ls  2670177.697116177:  7f0dfdbce056 _dl_start+0x26  movq  %rdx, 0x27683(%rip)        IPC: 0.00 (1/881)

Note, the IPC values are low due to page faults at the beginning of
execution. The additional cycles are due to the time to enter the
kernel, not the actual kernel page fault handler.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-9-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Add support for samples to contain IPC ratio

Copy the incremental instruction count and cycle count onto 'instructions'
and 'branches' samples.

Because Intel PT does not update the cycle count on every branch or
instruction, the incremental values will often be zero.

When there are values, they will be the number of instructions and
number of cycles since the last update, and thus represent the average
IPC since the last IPC value.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-8-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf tools: Add IPC information to perf_sample

Add counts of instructions and cycles, in order to represent
instructions-per-cycle (IPC).

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-7-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Accumulate cycle count from CYC packets

In preparation for providing instructions-per-cycle (IPC) information,
accumulate cycle count from CYC packets.

Although CYC packets are optional (requires config term 'cyc' to enable
cycle-accurate mode when recording), the simplest way to count cycles is
with CYC packets.

The first complication is that cycles must be counted only when also
counting instructions.

That means when control flow packet generation is enabled i.e. between
TIP.PGE and TIP.PGD packets.

Also, sampling the cycle count follows the same rules as sampling the
timestamp, that is, not before the instruction to which the decoder is
walking is reached.

In addition, the cycle count is not accurate for any but the first
branch of a TNT packet.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-6-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf intel-pt: Factor out intel_pt_update_sample_time

To eliminate some duplication and make the code more understandable,
factor out intel_pt_update_sample_time.

Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/20190520113728.14389-5-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf record: Allow mixing --user-regs with --call-graph=dwarf

When DWARF stacks were requested and at the same time that the user
specifies a register set using the --user-regs option the full register
context was being captured on samples:

  $ perf record -g --call-graph dwarf,1024 --user-regs=IP,SP,BP -- stack_test2.g.O3

  188143843893585 0x6b48 [0x4f8]: PERF_RECORD_SAMPLE(IP, 0x4002): 23828/23828: 0x401236 period: 1363819 addr: 0x7ffedbdd51ac
  ... FP chain: nr:0
  ... user regs: mask 0xff0fff ABI 64-bit
  .... AX    0x53b
  .... BX    0x7ffedbdd3cc0
  .... CX    0xffffffff
  .... DX    0x33d3a
  .... SI    0x7f09b74c38d0
  .... DI    0x0
  .... BP    0x401260
  .... SP    0x7ffedbdd3cc0
  .... IP    0x401236
  .... FLAGS 0x20a
  .... CS    0x33
  .... SS    0x2b
  .... R8    0x7f09b74c3800
  .... R9    0x7f09b74c2da0
  .... R10   0xfffffffffffff3ce
  .... R11   0x246
  .... R12   0x401070
  .... R13   0x7ffedbdd5db0
  .... R14   0x0
  .... R15   0x0
  ... ustack: size 1024, offset 0xe0
   . data_src: 0x5080021
   ... thread: stack_test2.g.O:23828
   ...... dso: /root/abudanko/stacks/stack_test2.g.O3

I.e. the --user-regs=IP,SP,BP was being ignored, being overridden by the
needs of --call-graph=dwarf.

After applying the change in this patch the sample data contains the
user specified register, but making sure that at least the minimal set
of register needed for DWARF unwinding (DWARF_MINIMAL_REGS) is
requested.

The user is warned that DWARF unwinding may not work if extra registers
end up being needed.

  -g call-graph dwarf,K                         full_regs
  --user-regs=user_regs                         user_regs
  -g call-graph dwarf,K --user-regs=user_regs user_regs + DWARF_MINIMAL_REGS

  $ perf record -g --call-graph dwarf,1024 --user-regs=BP -- ls
  WARNING: The use of --call-graph=dwarf may require all the user registers, specifying a subset with --user-regs may render DWARF unwinding unreliable, so the minimal registers set (IP, SP) is explicitly forced.
  arch   COPYING Documentation  include Kbuild  lbuild    MAINTAINERS modules.builtin  Module.symvers  perf.data.old scripts   System.map  virt
  block  CREDITS drivers        init Kconfig  lib    Makefile modules.builtin.modinfo  net  README security  tools       vmlinux
  certs  crypto fs        ipc kernel  LICENSES  mm modules.order  perf.data  samples sound   usr       vmlinux.o
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.030 MB perf.data (10 samples) ]

  188368474305373 0x5e40 [0x470]: PERF_RECORD_SAMPLE(IP, 0x4002): 23839/23839: 0x401236 period: 1260507 addr: 0x7ffd3d85e96c
  ... FP chain: nr:0
  ... user regs: mask 0x1c0 ABI 64-bit
  .... BP    0x401260
  .... SP    0x7ffd3d85cc20
  .... IP    0x401236
  ... ustack: size 1024, offset 0x58
   . data_src: 0x5080021

Committer notes:

Detected build failures on arches where PERF_REGS_ is not available,
such as debian:experimental-x-{mips,mips64,mipsel}, fedora 24 and 30 for
ARC uClibc and glibc, reported to Alexey that provided a patch moving
the DWARF_MINIMAL_REGS from evsel.c to util/perf_regs.h, where it is
guarded by an HAVE_PERF_REGS_SUPPORT ifdef.

Committer testing:

  # perf record --user-regs=bp,ax -a sleep 1
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 1.955 MB perf.data (1773 samples) ]
  # perf script -F+uregs | grep AX: | head -5
     perf 1719 [000] 181.272398:    1 cycles: ffffffffba06a7c4 native_write_msr+0x4 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffef828fb00
     perf 1719 [000] 181.272402:    1 cycles: ffffffffba06a7c4 native_write_msr+0x4 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffef828fb00
     perf 1719 [000] 181.272403:    8 cycles: ffffffffba06a7c4 native_write_msr+0x4 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffef828fb00
     perf 1719 [000] 181.272405:  181 cycles: ffffffffba06a7c6 native_write_msr+0x6 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffef828fb00
     perf 1719 [000] 181.272406: 4405 cycles: ffffffffba06a7c4 native_write_msr+0x4 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffef828fb00
  # perf record --call-graph=dwarf --user-regs=bp,ax -a sleep 1
  WARNING: The use of --call-graph=dwarf may require all the user registers, specifying a subset with --user-regs may render DWARF unwinding unreliable, so the minimal registers set (IP, SP) is explicitly forced.
  [ perf record: Woken up 55 times to write data ]
  [ perf record: Captured and wrote 24.184 MB perf.data (2841 samples) ]
  [root@quaco ~]# perf script --hide-call-graph -F+uregs | grep AX: | head -5
     perf 1729 [000] 211.268006:    1 cycles: ffffffffba06a7c4 native_write_msr+0x4 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffc8679abb0 SP:0x7ffc8679ab78 IP:0x7fa75223a0db
     perf 1729 [000] 211.268014:    1 cycles: ffffffffba06a7c4 native_write_msr+0x4 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffc8679abb0 SP:0x7ffc8679ab78 IP:0x7fa75223a0db
     perf 1729 [000] 211.268017:    5 cycles: ffffffffba06a7c4 native_write_msr+0x4 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffc8679abb0 SP:0x7ffc8679ab78 IP:0x7fa75223a0db
     perf 1729 [000] 211.268020:   48 cycles: ffffffffba06a7c6 native_write_msr+0x6 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffc8679abb0 SP:0x7ffc8679ab78 IP:0x7fa75223a0db
     perf 1729 [000] 211.268024:  490 cycles: ffffffffba00e471 intel_bts_enable_local+0x21 (/lib/modules/5.2.0-rc1+/build/vmlinux) ABI:2 AX:0xffffffffffffffda BP:0x7ffc8679abb0 SP:0x7ffc8679ab78 IP:0x7fa75223a0db
  #

Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.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/e7fd37b1-af22-0d94-a0dc-5895e803bbfe@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf symbols: Remove unused variable 'err'

Variable 'err' is defined but never used in function symsrc__init(),
remove it and directly return -1 at the end of the function.

Signed-off-by: Leo Yan <leo.yan@linaro.org>
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/20190530093801.20510-1-leo.yan@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf data: Document directory format header: HEADER_DIR_FORMAT

We forgot to update the perf.data file format document for the
HEADER_DIR_FORMAT header, do it now from comments in the patch
introducing 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: Chong Jiang <chongjiang@chromium.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Simon Que <sque@chromium.org>
Fixes: 258031c017c3 ("perf header: Add DIR_FORMAT feature to describe directory data")
Link: https://lkml.kernel.org/n/tip-jbrzb7ijb5al33gi8br6f9rr@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf data: Document clockid header: HEADER_CLOCKID

We forgot to update the perf.data file format document for the
HEADER_CLOCKID header, do it now from comments in the patch introducing
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: Chong Jiang <chongjiang@chromium.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Simon Que <sque@chromium.org>
Fixes: cf7905165fee ("perf record: Encode -k clockid frequency into Perf trace")
Link: https://lkml.kernel.org/n/tip-slhnjp06027j3ae17qqetzxj@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf data: Document memory topology header: HEADER_MEM_TOPOLOGY

We forgot to update the perf.data file format document for the
HEADER_MEM_TOPOLOGY header, do it now from comments in the patch
introducing it.

Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Chong Jiang <chongjiang@chromium.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Simon Que <sque@chromium.org>
Fixes: e2091cedd51b ("perf tools: Add MEM_TOPOLOGY feature to perf data file")
Link: https://lkml.kernel.org/n/tip-h5lcm1nbe9ztxwm61gmadd56@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

perf data: Add description of header HEADER_BPF_PROG_INFO and HEADER_BPF_BTF

This patch addes description of HEADER_BPF_PROG_INFO and HEADER_BPF_BTF to
perf.data-file-format.txt.

Requested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Song Liu <songliubraving@fb.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Fixes: 606f972b1361 ("perf bpf: Save bpf_prog_info information as headers to perf.data")
Link: http://lkml.kernel.org/r/20190521064406.2498925-1-songliubraving@fb.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

Merge branch 'x86/topology' into perf/core, to prepare for new patches

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

perf/x86: Use update attribute groups for default attributes

Using the new pmu::update_attrs attribute group for default
attributes - freeze_on_smi, allow_tsx_force_abort.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-10-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel: Use update attributes for skylake format

Using the new pmu::update_attrs attribute group for
skylake specific format attributes.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-9-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86: Use update attribute groups for extra format

Using the new pmu::update_attrs attribute group for
extra "format" directory.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-8-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86: Use update attribute groups for caps

Using the new pmu::update_attrs attribute group for
"caps" directory.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-7-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86: Add is_visible attribute_group callback for base events

We dont need to pre-filter out unsupported base events,
we can just use its group's is_visible function to do this.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-6-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86: Use the new pmu::update_attrs attribute group

Using the new pmu::update_attrs attribute group to
create detected events for x86_pmu.

Moving the topdown/memory/tsx attributes to separate
attribute groups with specific is_visible functions.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-5-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86: Get rid of x86_pmu::event_attrs

Nobody is using that.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-4-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Add attr_groups_update into struct pmu

Adding attr_update attribute group into pmu, to allow
having multiple attribute groups for same group name.

This will allow us to update "events" or "format"
directories with attributes that depend on various
HW conditions.

For example having group_format_extra group that updates
"format" directory only if pmu version is 2 and higher:

  static umode_t
  exra_is_visible(struct kobject *kobj, struct attribute *attr, int i)
  {
         return x86_pmu.version >= 2 ? attr->mode : 0;
  }

  static struct attribute_group group_format_extra = {
         .name       = "format",
         .is_visible = exra_is_visible,
  };

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-3-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

sysfs: Add sysfs_update_groups function

Adding sysfs_update_groups function to update
multiple groups.

  sysfs_update_groups - given a directory kobject, create a bunch of attribute groups
  @kobj:      The kobject to update the group on
  @groups:    The attribute groups to update, NULL terminated

This function update a bunch of attribute groups.  If an error occurs when
updating a group, all previously updated groups will be removed together
with already existing (not updated) attributes.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190512155518.21468-2-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel/uncore: Add new IMC PCI IDs for KabyLake, AmberLake and WhiskeyLake CPUs

AmberLake and WhiskeyLake have same client uncore events as
KabyLake. Thus add the PCI IDs for AmberLake Y processor lines,
for WhiskeyLake U processor lines and for KabyLake, add H
processor line and workstation.

 Platform Device ID
 ================================
 AML Y 2 Core 590Ch
 KBL H 4 Core 5910h
 KBL 4 Core WorkStation 5918h
 WHL U 4 Core 3ED0h
 WHL U 4 Core 3E34h
 WHL U 2 Core 3E35h
 AML Y 4 Core 590Dh

Signed-off-by: Gayatri Kammela <gayatri.kammela@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Charles Prestopine <charles.d.prestopine@intel.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190511000311.20733-2-gayatri.kammela@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/x86/intel/uncore: Add tabs to Uncore IMC PCI IDs

Improve code readability by adding tabs after #define macros

Signed-off-by: Gayatri Kammela <gayatri.kammela@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Charles Prestopine <charles.d.prestopine@intel.com>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190511000311.20733-1-gayatri.kammela@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

perf/core: Allow non-privileged uprobe for user processes

Currently, non-privileged user could only use uprobe with

    kernel.perf_event_paranoid = -1

However, setting perf_event_paranoid to -1 leaks other users' processes to
non-privileged uprobes.

To introduce proper permission control of uprobes, we are building the
following system:

  A daemon with CAP_SYS_ADMIN is in charge to create uprobes via tracefs;
  Users asks the daemon to create uprobes;
  Then user can attach uprobe only to processes owned by the user.

This patch allows non-privileged user to attach uprobe to processes owned
by the user.

The following example shows how to use uprobe with non-privileged user.
This is based on Brendan's blog post [1]

1. Create uprobe with root:

  sudo perf probe -x 'readline%return +0($retval):string'

2. Then non-root user can use the uprobe as:

  perf record -vvv -e probe_bash:readline__return -p <pid> sleep 20
  perf script

[1] http://www.brendangregg.com/blog/2015-06-28/linux-ftrace-uprobe.html

Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <kernel-team@fb.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190507161545.788381-1-songliubraving@fb.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

Merge tag 'v5.2-rc3' into perf/core, to pick up fixes

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

Linux 5.2-rc3

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

Pull x86 fixes from Ingo Molnar:
 "Two fixes: a quirk for KVM guests running on certain AMD CPUs, and a
  KASAN related build fix"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/CPU/AMD: Don't force the CPB cap when running under a hypervisor
  x86/boot: Provide KASAN compatible aliases for string routines