perf sort: Use default sort if evlist is empty
authorDavid Carrillo-Cisneros <davidcc@google.com>
Fri, 21 Jul 2017 05:11:57 +0000 (22:11 -0700)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Wed, 26 Jul 2017 20:00:07 +0000 (17:00 -0300)
Fixes bug noted by Jiri in https://lkml.org/lkml/2017/6/13/755 and
caused by commit d49dadea7862 ("perf tools: Make 'trace' or
'trace_fields' sort key default for tracepoint events") not taking into
account that evlist is empty in pipe-mode.

Before this commit, pipe mode will only show bogus "100.00%  N/A"
instead of correct output as follows:

  $ perf record -o - sleep 1 | perf report -i -
  # To display the perf.data header info, please use --header/--header-only options.
  #
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.000 MB - ]
  #
  # Total Lost Samples: 0
  #
  # Samples: 8  of event 'cycles:ppH'
  # Event count (approx.): 145658
  #
  # Overhead  Trace output
  # ........  ............
  #
     100.00%  N/A

Correct output, after patch:

  $ perf record -o - sleep 1 | perf report -i -
  # To display the perf.data header info, please use --header/--header-only options.
  #
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.000 MB - ]
  #
  # Total Lost Samples: 0
  #
  # Samples: 8  of event 'cycles:ppH'
  # Event count (approx.): 191331
  #
  # Overhead  Command  Shared Object      Symbol
  # ........  .......  .................  .................................
  #
      81.63%  sleep    libc-2.19.so       [.] _exit
      13.58%  sleep    ld-2.19.so         [.] do_lookup_x
       2.34%  sleep    [kernel.kallsyms]  [k] context_switch
       2.34%  sleep    libc-2.19.so       [.] __GI___libc_nanosleep
       0.11%  perf     [kernel.kallsyms]  [k] __intel_pmu_enable_a

Reported-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lkml.kernel.org/r/20170613185422.GA6092@krava
Signed-off-by: David Carrillo-Cisneros <davidcc@google.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: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Turner <pjt@google.com>
Cc: Simon Que <sque@chromium.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Wang Nan <wangnan0@huawei.com>
Fixes: d49dadea7862 ("perf tools: Make 'trace' or 'trace_fields' sort key default for tracepoint events")
Link: https://lkml.kernel.org/r/20170721051157.47331-1-davidcc@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/util/evlist.h
tools/perf/util/sort.c

index 0843746..bf2c493 100644 (file)
@@ -265,6 +265,11 @@ bool perf_evlist__valid_read_format(struct perf_evlist *evlist);
 void perf_evlist__splice_list_tail(struct perf_evlist *evlist,
                                   struct list_head *list);
 
+static inline bool perf_evlist__empty(struct perf_evlist *evlist)
+{
+       return list_empty(&evlist->entries);
+}
+
 static inline struct perf_evsel *perf_evlist__first(struct perf_evlist *evlist)
 {
        return list_entry(evlist->entries.next, struct perf_evsel, node);
index 8b327c9..12359bd 100644 (file)
@@ -2563,7 +2563,7 @@ static const char *get_default_sort_order(struct perf_evlist *evlist)
 
        BUG_ON(sort__mode >= ARRAY_SIZE(default_sort_orders));
 
-       if (evlist == NULL)
+       if (evlist == NULL || perf_evlist__empty(evlist))
                goto out_no_evlist;
 
        evlist__for_each_entry(evlist, evsel) {