linux-stable.git/tools/perf/util, branch v6.2.4

perf stat: Avoid merging/aggregating metric counts twice

2023-03-10T08:28:44+00:00

[ Upstream commit 37f322cd58d81a9d46456531281c908de9ef6e42 ]

The added perf_stat_merge_counters combines uncore counters. When
metrics are enabled, the counts are merged into a metric_leader via the
stat-shadow saved_value logic. As the leader now is passed an aggregated
count, it leads to all counters being added together twice and counts
appearing approximately doubled in metrics.

This change disables the saved_value merging of counts for evsels that
are merged. It is recommended that later changes remove the saved_value
entirely as the two layers of aggregation in the code is confusing.

Fixes: 942c5593393d9418 ("perf stat: Add perf_stat_merge_counters()")
Reported-by: Perry Taylor 
Signed-off-by: Ian Rogers 
Acked-by: Namhyung Kim 
Cc: Alexander Shishkin 
Cc: Eduard Zingerman 
Cc: Florian Fischer 
Cc: Ingo Molnar 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Cc: Stephane Eranian 
Cc: Xing Zhengjun 
Link: https://lore.kernel.org/r/20230209064447.83733-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf stat: Hide invalid uncore event output for aggr mode

2023-03-10T08:28:44+00:00

[ Upstream commit dd15480a3d67b9cf04a1f6f5d60f1c0dc018e22f ]

The current display code for perf stat iterates given cpus and build the
aggr map to collect the event data for the aggregation mode.

But uncore events have their own cpu maps and it won't guarantee that
it'd match to the aggr map.  For example, per-package uncore events
would generate a single value for each socket.  When user asks per-core
aggregation mode, the output would contain 0 values for other cores.

Thus it needs to check the uncore PMU's cpumask and if it matches to the
current aggregation id.

Before:
  $ sudo ./perf stat -a --per-core -e power/energy-pkg/ sleep 1

   Performance counter stats for 'system wide':

  S0-D0-C0              1               3.73 Joules power/energy-pkg/
  S0-D0-C1              0       Joules power/energy-pkg/
  S0-D0-C2              0       Joules power/energy-pkg/
  S0-D0-C3              0       Joules power/energy-pkg/

         1.001404046 seconds time elapsed

  Some events weren't counted. Try disabling the NMI watchdog:
  	echo 0 > /proc/sys/kernel/nmi_watchdog
  	perf stat ...
  	echo 1 > /proc/sys/kernel/nmi_watchdog

The core 1, 2 and 3 should not be printed because the event is handled
in a cpu in the core 0 only.  With this change, the output becomes like
below.

After:
  $ sudo ./perf stat -a --per-core -e power/energy-pkg/ sleep 1

   Performance counter stats for 'system wide':

  S0-D0-C0              1               2.09 Joules power/energy-pkg/

Fixes: b897613510890d6e ("perf stat: Update event skip condition for system-wide per-thread mode and merged uncore and hybrid events")
Signed-off-by: Namhyung Kim 
Tested-by: Ian Rogers 
Acked-by: Ian Rogers 
Cc: Adrian Hunter 
Cc: Athira Rajeev 
Cc: Ingo Molnar 
Cc: Jiri Olsa 
Cc: Michael Petlan 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Link: https://lore.kernel.org/r/20230125192431.2929677-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf intel-pt: Do not try to queue auxtrace data on pipe

2023-03-10T08:28:44+00:00

[ Upstream commit aeb802f872a7c42e4381f36041e77d1745908255 ]

When it processes AUXTRACE_INFO, it calls to auxtrace_queue_data() to
collect AUXTRACE data first.  That won't work with pipe since it needs
lseek() to read the scattered aux data.

  $ perf record -o- -e intel_pt// true | perf report -i- --itrace=i100
  # To display the perf.data header info, please use --header/--header-only options.
  #
  0x4118 [0xa0]: failed to process type: 70
  Error:
  failed to process sample

For the pipe mode, it can handle the aux data as it gets.  But there's
no guarantee it can get the aux data in time.  So the following warning
will be shown at the beginning:

  WARNING: Intel PT with pipe mode is not recommended.
           The output cannot relied upon.  In particular,
           time stamps and the order of events may be incorrect.

Fixes: dbd134322e74f19d ("perf intel-pt: Add support for decoding AUX area samples")
Reviewed-by: Adrian Hunter 
Reviewed-by: James Clark 
Signed-off-by: Namhyung Kim 
Cc: Adrian Hunter 
Cc: Ian Rogers 
Cc: Ingo Molnar 
Cc: Jiri Olsa 
Cc: Leo Yan 
Cc: Peter Zijlstra 
Cc: Stephane Eranian 
Link: https://lore.kernel.org/r/20230131023350.1903992-3-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf llvm: Fix inadvertent file creation

2023-03-10T08:28:43+00:00

[ Upstream commit 9f19aab47ced012eddef1e2bc96007efc7713b61 ]

The LLVM template is first echo-ed into command_out and then
command_out executed. The echo surrounds the template with double
quotes, however, the template itself may contain quotes. This is
generally innocuous but in tools/perf/tests/bpf-script-test-prologue.c
we see:
...
SEC("func=null_lseek file->f_mode offset orig")
...
where the first double quote ends the double quote of the echo, then
the > redirects output into a file called f_mode.

To avoid this inadvertent behavior substitute redirects and similar
characters to be ASCII control codes, then substitute the output in
the echo back again.

Fixes: 5eab5a7ee032acaa ("perf llvm: Display eBPF compiling command in debug output")
Signed-off-by: Ian Rogers 
Cc: Alexander Shishkin 
Cc: Andrii Nakryiko 
Cc: bpf@vger.kernel.org
Cc: Ingo Molnar 
Cc: Jiri Olsa 
Cc: llvm@lists.linux.dev
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Nathan Chancellor 
Cc: Nick Desaulniers 
Cc: Peter Zijlstra 
Cc: Tom Rix 
Link: https://lore.kernel.org/r/20230105082609.344538-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Sasha Levin

perf buildid-cache: Fix the file mode with copyfile() while adding file to build-id cache

2023-01-18T13:39:16+00:00

The test "build id cache operations" fails on powerpc as below:

	Adding 5a0fd882b53084224ba47b624c55a469 ./tests/shell/../pe-file.exe: Ok
	build id: 5a0fd882b53084224ba47b624c55a469
	link: /tmp/perf.debug.ZTu/.build-id/5a/0fd882b53084224ba47b624c55a469
	file: /tmp/perf.debug.ZTu/.build-id/5a/../../root/linux/tools/perf/tests/pe-file.exe/5a0fd882b53084224ba47b624c55a469/elf
	failed: file /tmp/perf.debug.ZTu/.build-id/5a/../../root/linux/tools/perf/tests/pe-file.exe/5a0fd882b53084224ba47b624c55a469/elf does not exist
	test child finished with -1
	---- end ----
	build id cache operations: FAILED!

The failing test is when trying to add pe-file.exe to build id cache.

'perf buildid-cache' can be used to add/remove/manage files from the
build-id cache. "-a" option is used to add a file to the build-id cache.

Simple command to do so for a PE exe file:

  # ls -ltr tests/pe-file.exe
  -rw-r--r--. 1 root root 75595 Jan 10 23:35 tests/pe-file.exe

  The file is in home directory.

  # mkdir  /tmp/perf.debug.TeY1
  # perf --buildid-dir /tmp/perf.debug.TeY1 buildid-cache -v -a tests/pe-file.exe

The above will create ".build-id" folder in build id directory, which is
/tmp/perf.debug.TeY1. Also adds file to this folder under build id.
Example:

  # ls -ltr /tmp/perf.debug.TeY1/.build-id/5a/0fd882b53084224ba47b624c55a469/
  total 76
  -rw-r--r--. 1 root root     0 Jan 11 00:38 probes
  -rwxr-xr-x. 1 root root 75595 Jan 11 00:38 elf

We can see in the results that file mode for original file and file in
build id directory is different. ie, build id file has executable
permission whereas original file doesn’t have.

The code path and function (build_id_cache__add  to add a file to the
cache is in "util/build-id.c". In build_id_cache__add() function, it
first attempts to link the original file to destination cache folder.

If linking the file fails (which can happen if the destination and
source is on a different mount points), it will copy the file to
destination.  Here copyfile() routine explicitly uses mode as "755" and
hence file in the destination will have executable permission.

Code snippet:

 if (link(realname, filename) && errno != EEXIST && copyfile(name, filename))

strace logs:

	172285 link("/home//linux/tools/perf/tests/pe-file.exe", "/tmp/perf.debug.TeY1/home//linux/tools/perf/tests/pe-file.exe/5a0fd882b53084224ba47b624c55a469/elf") = -1 EXDEV (Invalid cross-device link)
	172285 newfstatat(AT_FDCWD, "tests/pe-file.exe", {st_mode=S_IFREG|0644, st_size=75595, ...}, 0) = 0
	172285 openat(AT_FDCWD, "/tmp/perf.debug.TeY1/home//linux/tools/perf/tests/pe-file.exe/5a0fd882b53084224ba47b624c55a469/.elf.KbAnsl", O_RDWR|O_CREAT|O_EXCL, 0600) = 3
	172285 fchmod(3, 0755)                  = 0
	172285 openat(AT_FDCWD, "tests/pe-file.exe", O_RDONLY) = 4
	172285 mmap(NULL, 75595, PROT_READ, MAP_PRIVATE, 4, 0) = 0x7fffa5cd0000
	172285 pwrite64(3, "MZ\220\0\3\0\0\0\4\0\0\0\377\377\0\0\270\0\0\0\0\0\0\0@\0\0\0\0\0\0\0"..., 75595, 0) = 75595

Whereas if the link succeeds, it succeeds in the first attempt itself
and the file in the build-id dir will have same permission as original
file.

Example, above uses /tmp. Instead if we use "--buildid-dir /home/build",
linking will work here since mount points are same. Hence the
destination file will not have executable permission.

Since the testcase "tests/shell/buildid.sh" always looks for executable
file, test fails in powerpc environment when test is run from /root.

The patch adds a change in build_id_cache__add() to use copyfile_mode()
which also passes the file’s original mode as argument. This way the
destination file mode also will be same as original file.

Signed-off-by: Athira Jajeev 
Cc: Andi Kleen 
Cc: Disha Goel 
Cc: Ian Rogers 
Cc: James Clark 
Cc: Jiri Olsa 
Cc: Kajol Jain 
Cc: Madhavan Srinivasan 
Cc: Michael Ellerman 
Cc: Nageswara R Sastry 
Cc: Namhyung Kim 
Cc: linuxppc-dev@lists.ozlabs.org
Link: https://lore.kernel.org/r/20230116050131.17221-1-atrajeev@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo

perf expr: Prevent normalize() from reading into undefined memory in the expression lexer

2023-01-18T13:33:00+00:00

The current implementation does not account for a trailing backslash
followed by a null-byte.

If a null-byte is encountered following a backslash, normalize() will
continue reading (and potentially writing) into garbage memory ignoring
the EOS null-byte.

Signed-off-by: Sohom Datta 
Acked-by: Ian Rogers 
Acked-by: Jiri Olsa 
Cc: Alexander Shishkin 
Cc: Ingo Molnar 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Link: https://lore.kernel.org/r/20221204105836.1012885-1-sohomdatta1+git@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo

perf auxtrace: Fix address filter duplicate symbol selection

2023-01-11T17:03:44+00:00

When a match has been made to the nth duplicate symbol, return
success not error.

Example:

  Before:

    $ cat file.c
    cat: file.c: No such file or directory
    $ cat file1.c
    #include 

    static void func(void)
    {
            printf("First func\n");
    }

    void other(void);

    int main()
    {
            func();
            other();
            return 0;
    }
    $ cat file2.c
    #include 

    static void func(void)
    {
            printf("Second func\n");
    }

    void other(void)
    {
            func();
    }

    $ gcc -Wall -Wextra -o test file1.c file2.c
    $ perf record -e intel_pt//u --filter 'filter func @ ./test' -- ./test
    Multiple symbols with name 'func'
    #1      0x1149  l       func
                    which is near           main
    #2      0x1179  l       func
                    which is near           other
    Disambiguate symbol name by inserting #n after the name e.g. func #2
    Or select a global symbol by inserting #0 or #g or #G
    Failed to parse address filter: 'filter func @ ./test'
    Filter format is: filter|start|stop|tracestop  [/ ] [@]
    Where multiple filters are separated by space or comma.
    $ perf record -e intel_pt//u --filter 'filter func #2 @ ./test' -- ./test
    Failed to parse address filter: 'filter func #2 @ ./test'
    Filter format is: filter|start|stop|tracestop  [/ ] [@]
    Where multiple filters are separated by space or comma.

  After:

    $ perf record -e intel_pt//u --filter 'filter func #2 @ ./test' -- ./test
    First func
    Second func
    [ perf record: Woken up 1 times to write data ]
    [ perf record: Captured and wrote 0.016 MB perf.data ]
    $ perf script --itrace=b -Ftime,flags,ip,sym,addr --ns
    1231062.526977619:   tr strt                               0 [unknown] =>     558495708179 func
    1231062.526977619:   tr end  call               558495708188 func =>     558495708050 _init
    1231062.526979286:   tr strt                               0 [unknown] =>     55849570818d func
    1231062.526979286:   tr end  return             55849570818f func =>     55849570819d other

Fixes: 1b36c03e356936d6 ("perf record: Add support for using symbols in address filters")
Reported-by: Dmitrii Dolgov <9erthalion6@gmail.com>
Signed-off-by: Adrian Hunter 
Tested-by: Dmitry Dolgov <9erthalion6@gmail.com>
Cc: Adrian Hunter 
Cc: Ian Rogers 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230110185659.15979-1-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo

perf build: Properly guard libbpf includes

2023-01-10T13:51:39+00:00

Including libbpf header files should be guarded by HAVE_LIBBPF_SUPPORT.
In bpf_counter.h, move the skeleton utilities under HAVE_BPF_SKEL.

Fixes: d6a735ef3277c45f ("perf bpf_counter: Move common functions to bpf_counter.h")
Reported-by: Mike Leach 
Signed-off-by: Ian Rogers 
Tested-by: Arnaldo Carvalho de Melo 
Tested-by: Jiri Olsa 
Tested-by: Mike Leach 
Cc: Alexander Shishkin 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Link: http://lore.kernel.org/lkml/20230105172243.7238-1-mike.leach@linaro.org
Signed-off-by: Arnaldo Carvalho de Melo

perf tools: Fix build on uClibc systems by adding missing sys/types.h include

2023-01-04T19:44:01+00:00

Not all libc implementations define ssize_t as part of stdio.h like
glibc does since the standard only requires this type to be defined by
unistd.h and sys/types.h. For this reason the perf build is currently
broken for toolchains based on uClibc, for instance.

Include sys/types.h explicitly to fix that.

Committer notes:

In addition, in the past this worked in uClibc test systems as there was
another way to get to sys/types.h that got removed in that cset:

  tools/perf/util/trace-event.h
    /usr/include/traceevent/event_parse.h # This got removed from util/trace-event.h in 378ef0f5d9d7f465
      /usr/include/regex.h
        /usr/include/sys/types.h
          typedef __ssize_t ssize_t;

So the size_t that is used in tools/perf/util/trace-event.h was being
obtained indirectly, by chance.

Fixes: 378ef0f5d9d7f465 ("perf build: Use libtraceevent from the system")
Signed-off-by: Jesus Sanchez-Palencia 
Cc: Alexander Shishkin 
Cc: Ian Rogers 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Link: http://lore.kernel.org/lkml/20230104193414.606905-1-jesussanp@google.com
Signed-off-by: Arnaldo Carvalho de Melo

perf stat: Fix handling of --for-each-cgroup with --bpf-counters to match non BPF mode

2023-01-04T14:11:41+00:00

The --for-each-cgroup can have the same cgroup multiple times, but this
confuses BPF counters (since they have the same cgroup id), making only
the last cgroup events to be counted.

Let's check the cgroup name before adding a new entry to the cgroups
list.

Before:

  $ sudo ./perf stat -a --bpf-counters --for-each-cgroup /,/ sleep 1

   Performance counter stats for 'system wide':

        msec cpu-clock                        /
             context-switches                 /
             cpu-migrations                   /
             page-faults                      /
             cycles                           /
             instructions                     /
             branches                         /
             branch-misses                    /
            8,016.04 msec cpu-clock                        /                #    7.998 CPUs utilized
               6,152      context-switches                 /                #  767.461 /sec
                 250      cpu-migrations                   /                #   31.187 /sec
                 442      page-faults                      /                #   55.139 /sec
         613,111,487      cycles                           /                #    0.076 GHz
         280,599,604      instructions                     /                #    0.46  insn per cycle
          57,692,724      branches                         /                #    7.197 M/sec
           3,385,168      branch-misses                    /                #    5.87% of all branches

         1.002220125 seconds time elapsed

After it becomes similar to the non-BPF mode:

  $ sudo ./perf stat -a --bpf-counters --for-each-cgroup /,/  sleep 1

   Performance counter stats for 'system wide':

            8,013.38 msec cpu-clock                        /                #    7.998 CPUs utilized
               6,859      context-switches                 /                #  855.944 /sec
                 334      cpu-migrations                   /                #   41.680 /sec
                 345      page-faults                      /                #   43.053 /sec
         782,326,119      cycles                           /                #    0.098 GHz
         471,645,724      instructions                     /                #    0.60  insn per cycle
          94,963,430      branches                         /                #   11.851 M/sec
           3,685,511      branch-misses                    /                #    3.88% of all branches

         1.001864539 seconds time elapsed

Committer notes:

As a reminder, to test with BPF counters one has to use BUILD_BPF_SKEL=1
in the make command line and have clang/llvm installed when building
perf, otherwise the --bpf-counters option will not be available:

  # perf stat -a --bpf-counters --for-each-cgroup /,/ sleep 1
  Error: unknown option `bpf-counters'

   Usage: perf stat [] []

      -a, --all-cpus        system-wide collection from all CPUs
  
  #

Fixes: bb1c15b60b981d10 ("perf stat: Support regex pattern in --for-each-cgroup")
Signed-off-by: Namhyung Kim 
Tested-by: Arnaldo Carvalho de Melo 
Cc: Adrian Hunter 
Cc: bpf@vger.kernel.org
Cc: Ian Rogers 
Cc: Ingo Molnar 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Cc: Song Liu 
Link: https://lore.kernel.org/r/20230104064402.1551516-5-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo