linux-stable.git/tools/perf, branch v5.16.3

perf metric: Fix metric_leader

2022-01-27T11:03:01+00:00

commit d3e2bb4359f70c8b1d09a6f8e2f57240aab0da3f upstream.

Multiple events may have a metric_leader to aggregate into.

This happens for uncore events where, for example, uncore_imc is
expanded into uncore_imc_0, uncore_imc_1, etc.

Such events all have the same metric_id and should aggregate into the
first event.

The change introducing metric_ids had a bug where the metric_id was
compared to itself, creating an always true condition.

Correct this by comparing the event in the metric_evlist and the
metric_leader.

Fixes: ec5c5b3d2c21b3f3 ("perf metric: Encode and use metric-id as qualifier")
Signed-off-by: Ian Rogers 
Cc: Alexander Shishkin 
Cc: Andi Kleen 
Cc: Jiri Olsa 
Cc: John Garry 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Cc: Stephane Eranian 
Link: http://lore.kernel.org/lkml/20220115062852.1959424-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

perf probe: Fix ppc64 'perf probe add events failed' case

2022-01-27T11:03:01+00:00

commit 4624f199327a704dd1069aca1c3cadb8f2a28c6f upstream.

Because of commit bf794bf52a80c627 ("powerpc/kprobes: Fix kallsyms
lookup across powerpc ABIv1 and ABIv2"), in ppc64 ABIv1, our perf
command eliminates the need to use the prefix "." at the symbol name.

But when the command "perf probe -a schedule" is executed on ppc64
ABIv1, it obtains two symbol address information through /proc/kallsyms,
for example:

  cat /proc/kallsyms | grep -w schedule
  c000000000657020 T .schedule
  c000000000d4fdb8 D schedule

The symbol "D schedule" is not a function symbol, and perf will print:
"p:probe/schedule _text+13958584"Failed to write event: Invalid argument

Therefore, when searching symbols from map and adding probe point for
them, a symbol type check is added. If the type of symbol is not a
function, skip it.

Fixes: bf794bf52a80c627 ("powerpc/kprobes: Fix kallsyms lookup across powerpc ABIv1 and ABIv2")
Signed-off-by: Zechuan Chen 
Acked-by: Masami Hiramatsu 
Cc: Alexander Shishkin 
Cc: Ingo Molnar 
Cc: Jianlin Lv 
Cc: Jin Yao 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Michael Ellerman 
Cc: Namhyung Kim 
Cc: Naveen N. Rao 
Cc: Peter Zijlstra 
Cc: Ravi Bangoria 
Cc: Yang Jihong 
Link: https://lore.kernel.org/r/20211228111338.218602-1-chenzechuan1@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

perf test: Enable system wide for metricgroups test

2022-01-27T11:03:01+00:00

commit 0046686da0ef692a6381260c3aa44291187eafc9 upstream.

Uncore events as group leaders fail in per-thread mode causing exit
errors. Enable system-wide for metricgroup testing. This fixes the HPC
metric group when tested on skylakex.

Fixes: 4a87dea9e60fe100 ("perf test: Workload test of metric and metricgroups")
Signed-off-by: Ian Rogers 
Tested-by: Arnaldo Carvalho de Melo 
Cc: Alexander Shishkin 
Cc: Andi Kleen 
Cc: Ingo Molnar 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Cc: Stephane Eranian 
Link: https://lore.kernel.org/r/20211223183948.3423989-1-irogers@google.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

perf metricgroup: Fix use after free in metric__new()

2022-01-27T11:03:00+00:00

commit e000ea0beffb5497425054b151369fe37a792ece upstream.

We shouldn't free() something that will be used in the next line, fix
it.

Fixes: b85a4d61d3022608 ("perf metric: Allow modifiers on metrics")
Addresses-Coverity-ID: 1494000
Signed-off-by: José Expósito 
Cc: Alexander Shishkin 
Cc: Andi Kleen 
Cc: Ian Rogers 
Cc: Jiri Olsa 
Cc: John Garry 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Link: http://lore.kernel.org/lkml/20211208171113.22089-1-jose.exposito89@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

perf tools: Drop requirement for libstdc++.so for libopencsd check

2022-01-27T11:03:00+00:00

commit ed17b1914978eddb2b01f2d34577f1c82518c650 upstream.

It's possible to link against libopencsd_c_api without having
libstdc++.so available, only libstdc++.so.6.0.28 (or whatever version is
in use) needs to be available. The same holds true for libopencsd.so.
When -lstdc++ (or -lopencsd) is explicitly passed to the linker however
the .so file must be available.

So wrap adding the dependencies into a check for static linking that
actually requires adding them all. The same construct is already used
for some other tests in the same file to reduce dependencies in the
dynamic linking case.

Fixes: 573cf5c9a152 ("perf build: Add missing -lstdc++ when linking with libopencsd")
Reviewed-by: James Clark 
Signed-off-by: Uwe Kleine-König 
Cc: Adrian Bunk 
Cc: Alexander Shishkin 
Cc: Branislav Rankov 
Cc: Diederik de Haas 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Link: https://lore.kernel.org/all/20211203210544.1137935-1-uwe@kleine-koenig.org
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

perf cputopo: Fix CPU topology reading on s/390

2022-01-27T11:03:00+00:00

commit a6e62743621ea29bea461774c0bcc68e5de59068 upstream.

Commit fdf1e29b6118c18f ("perf expr: Add metric literals for topology.")
fails on s390:

 # ./perf test -Fv 7
   ...
 # FAILED tests/expr.c:173 #num_dies >= #num_packages
   ---- end ----
   Simple expression parser: FAILED!
 #

Investigating this issue leads to these functions:
 build_cpu_topology()
   +--> has_die_topology(void)
        {
           struct utsname uts;

           if (uname(&uts) < 0)
                  return false;
           if (strncmp(uts.machine, "x86_64", 6))
                  return false;
           ....
        }

which always returns false on s390. The caller build_cpu_topology()
checks has_die_topology() return value. On false the
the struct cpu_topology::die_cpu_list is not contructed and has zero
entries. This leads to the failing comparison: #num_dies >= #num_packages.
s390 of course has a positive number of packages.

Fix this by adding s390 architecture to support CPU die list.

Output after:
 # ./perf test -Fv 7
  7: Simple expression parser                                        :
  --- start ---
  division by zero
  syntax error
  ---- end ----
  Simple expression parser: Ok
 #

Fixes: fdf1e29b6118c18f ("perf expr: Add metric literals for topology.")
Reviewed-by: Ian Rogers 
Signed-off-by: Thomas Richter 
Cc: Heiko Carstens 
Cc: Ian Rogers 
Cc: Sumanth Korikkar 
Cc: Sven Schnelle 
Cc: Vasily Gorbik 
Link: https://lore.kernel.org/r/20211124090343.9436-1-tmricht@linux.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

perf script: Fix hex dump character output

2022-01-27T11:02:59+00:00

commit 62942e9fda9fd1def10ffcbd5e1c025b3c9eec17 upstream.

Using grep -C with perf script -D can give erroneous results as grep loses
lines due to non-printable characters, for example, below the 0020, 0060
and 0070 lines are missing:

 $ perf script -D | grep -C10 AUX | head
 .  0010:  08 00 00 00 00 00 00 00 1f 00 00 00 00 00 00 00  ................
 .  0030:  01 00 00 00 00 00 00 00 00 04 00 00 00 00 00 00  ................
 .  0040:  00 08 00 00 00 00 00 00 02 00 00 00 00 00 00 00  ................
 .  0050:  00 00 00 00 00 00 00 00 01 00 00 00 00 00 00 00  ................
 .  0080:  02 00 00 00 00 00 00 00 1b 00 00 00 00 00 00 00  ................
 .  0090:  00 00 00 00 00 00 00 00                          ........

 0 0 0x450 [0x98]: PERF_RECORD_AUXTRACE_INFO type: 1
   PMU Type            8
   Time Shift          31

perf's isprint() is a custom implementation from the kernel, but the
kernel's _ctype appears to include characters from Latin-1 Supplement which
is not compatible with, for example, UTF-8. Fix by checking also isascii().

After:

 $ tools/perf/perf script -D | grep -C10 AUX | head
 .  0010:  08 00 00 00 00 00 00 00 1f 00 00 00 00 00 00 00  ................
 .  0020:  03 84 32 2f 00 00 00 00 63 7c 4f d2 fa ff ff ff  ..2/....c|O.....
 .  0030:  01 00 00 00 00 00 00 00 00 04 00 00 00 00 00 00  ................
 .  0040:  00 08 00 00 00 00 00 00 02 00 00 00 00 00 00 00  ................
 .  0050:  00 00 00 00 00 00 00 00 01 00 00 00 00 00 00 00  ................
 .  0060:  00 02 00 00 00 00 00 00 00 c0 03 00 00 00 00 00  ................
 .  0070:  e2 00 00 00 00 00 00 00 02 00 00 00 00 00 00 00  ................
 .  0080:  02 00 00 00 00 00 00 00 1b 00 00 00 00 00 00 00  ................
 .  0090:  00 00 00 00 00 00 00 00                          ........

Fixes: 3052ba56bcb58904 ("tools perf: Move from sane_ctype.h obtained from git to the Linux's original")
Signed-off-by: Adrian Hunter 
Cc: Jiri Olsa 
Link: http://lore.kernel.org/lkml/20220112085057.277205-1-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

perf evsel: Override attr->sample_period for non-libpfm4 events

2022-01-27T11:02:54+00:00

commit 3606c0e1a1050d397ad759a62607e419fd8b0ccb upstream.

A previous patch preventing "attr->sample_period" values from being
overridden in pfm events changed a related behaviour in arm-spe.

Before said patch:

  perf record -c 10000 -e arm_spe_0// -- sleep 1

Would yield an SPE event with period=10000. After the patch, the period
in "-c 10000" was being ignored because the arm-spe code initializes
sample_period to a non-zero value.

This patch restores the previous behaviour for non-libpfm4 events.

Fixes: ae5dcc8abe31 (“perf record: Prevent override of attr->sample_period for libpfm4 events”)
Reported-by: Chase Conklin 
Signed-off-by: German Gomez 
Cc: Alexander Shishkin 
Cc: Ian Rogers 
Cc: Jiri Olsa 
Cc: John Fastabend 
Cc: KP Singh 
Cc: Mark Rutland 
Cc: Martin KaFai Lau 
Cc: Namhyung Kim 
Cc: Song Liu 
Cc: Stephane Eranian 
Cc: Yonghong Song 
Cc: bpf@vger.kernel.org
Cc: netdev@vger.kernel.org
Link: http://lore.kernel.org/lkml/20220118144054.2541-1-german.gomez@arm.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

perf annotate: Avoid TUI crash when navigating in the annotation of recursive functions

2022-01-20T08:12:21+00:00

commit d5962fb7d69073bf68fb647531cfd4f0adf84be3 upstream.

In 'perf report', entering a recursive function from inside of itself
(either directly of indirectly through some other function) results in
calling symbol__annotate2 multiple() times, and freeing the whole
disassembly when exiting from the innermost instance.

The first issue causes the function's disassembly to be duplicated, and
the latter a heap use-after-free (and crash) when trying to access the
disassembly again.

I reproduced the bug on perf 5.11.22 (Ubuntu 20.04.3 LTS) and 5.16.rc8
with the following testcase (compile with gcc recursive.c -o recursive).
To reproduce:

- perf record ./recursive
- perf report
- enter fibonacci and annotate it
- move the cursor on one of the "callq fibonacci" instructions and press enter
  - at this point there will be two copies of the function in the disassembly
- go back by pressing q, and perf will crash

  #include 

  int fibonacci(int n)
  {
      if(n <= 2) return 1;
      return fibonacci(n-1) + fibonacci(n-2);
  }

  int main()
  {
      printf("%d\n", fibonacci(40));
  }

This patch addresses the issue by annotating a function and freeing the
associated memory on exit only if no annotation is already present, so
that a recursive function is only annotated on entry.

Signed-off-by: Dario Petrillo 
Tested-by: Arnaldo Carvalho de Melo 
Cc: Alexander Shishkin 
Cc: Jiri Olsa 
Cc: Mark Rutland 
Cc: Namhyung Kim 
Cc: Peter Zijlstra 
Cc: stable@kernel.org
Link: http://lore.kernel.org/lkml/20220109234441.325106-1-dario.pk1@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo 
Signed-off-by: Greg Kroah-Hartman

Revert "libtraceevent: Increase libtraceevent logging when verbose"

2022-01-07T19:08:50+00:00

This reverts commit 08efcb4a638d260ef7fcbae64ecf7ceceb3f1841.

This breaks the build as it will prefer using libbpf-devel header files,
even when not using LIBBPF_DYNAMIC=1, breaking the build.

This was detected on OpenSuSE Tumbleweed with libtraceevent-devel 1.3.0,
as described by Jiri Slaby:

=======================================================================
It breaks build with LIBTRACEEVENT_DYNAMIC and version 1.3.0:
> util/debug.c: In function ‘perf_debug_option’:
> util/debug.c:243:17: error: implicit declaration of function
‘tep_set_loglevel’ [-Werror=implicit-function-declaration]
>   243 |                 tep_set_loglevel(TEP_LOG_INFO);
>       |                 ^~~~~~~~~~~~~~~~
> util/debug.c:243:34: error: ‘TEP_LOG_INFO’ undeclared (first use in this
function); did you mean ‘TEP_PRINT_INFO’?
>   243 |                 tep_set_loglevel(TEP_LOG_INFO);
>       |                                  ^~~~~~~~~~~~
>       |                                  TEP_PRINT_INFO
> util/debug.c:243:34: note: each undeclared identifier is reported only once
for each function it appears in
> util/debug.c:245:34: error: ‘TEP_LOG_DEBUG’ undeclared (first use in this
function)
>   245 |                 tep_set_loglevel(TEP_LOG_DEBUG);
>       |                                  ^~~~~~~~~~~~~
> util/debug.c:247:34: error: ‘TEP_LOG_ALL’ undeclared (first use in this
function)
>   247 |                 tep_set_loglevel(TEP_LOG_ALL);
>       |                                  ^~~~~~~~~~~

It is because the gcc's command line looks like:
gcc
...
-I/home/abuild/rpmbuild/BUILD/tools/lib/
...
-DLIBTRACEEVENT_VERSION=65790
...
=======================================================================

The proper way to fix this is more involved and so not suitable for this
late in the 5.16-rc stage.

Reported-by: Jiri Slaby 
Link: https://lore.kernel.org/lkml/bc2b0786-8965-1bcd-2316-9d9bb37b9c31@kernel.org
Cc: Andrii Nakryiko 
Cc: Ian Rogers 
Cc: Jiri Olsa 
Cc: Namhyung Kim 
Cc: Song Liu 
Cc: Steven Rostedt 
Link: https://lore.kernel.org/lkml/YddGjjmlMZzxUZbN@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo