perf stat: Failure with "Operation not supported"
Anton Blanchard [Thu, 1 Dec 2011 22:38:33 +0000 (09:38 +1100)]
perf stat is failing on PowerPC:

  Error: open_counter returned with 95 (Operation not supported). /bin/dmesg may provide additional information.

  Fatal: Not all events could be opened.

commit 370faf1dd046 (perf stat: Fail softly on unsupported events)
added a check for failure returning ENOENT, but the POWER backend
returns EOPNOTSUPP. It looks like alpha, blackfin and mips do the
same.

With the patch applied, things work as expected:

 Performance counter stats for '/bin/true':

          0.362176 task-clock                #    0.623 CPUs utilized
                 0 context-switches          #    0.000 M/sec
                 0 CPU-migrations            #    0.000 M/sec
                28 page-faults               #    0.077 M/sec
         1,677,020 cycles                    #    4.630 GHz
   <not supported> stalled-cycles-frontend
   <not supported> stalled-cycles-backend
           431,220 instructions              #    0.26  insns per cycle
           101,889 branches                  #  281.325 M/sec
             4,145 branch-misses             #    4.07% of all branches

       0.000581361 seconds time elapsed

Cc: <stable@kernel.org> # 3.0+
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20111202093833.5fef7226@kryten
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

tools/perf/builtin-stat.c

index 7d98676..955930e 100644 (file)
@@ -463,7 +463,8 @@ static int run_perf_stat(int argc __used, const char **argv)
 
        list_for_each_entry(counter, &evsel_list->entries, node) {
                if (create_perf_stat_counter(counter, first) < 0) {
-                       if (errno == EINVAL || errno == ENOSYS || errno == ENOENT) {
+                       if (errno == EINVAL || errno == ENOSYS ||
+                           errno == ENOENT || errno == EOPNOTSUPP) {
                                if (verbose)
                                        ui__warning("%s event is not supported by the kernel.\n",
                                                    event_name(counter));