perf trace: Introduce --min-stack filter

author Arnaldo Carvalho de Melo <acme@redhat.com>

Fri, 15 Apr 2016 14:10:31 +0000 (11:10 -0300)

committer Arnaldo Carvalho de Melo <acme@redhat.com>

Fri, 15 Apr 2016 16:14:20 +0000 (13:14 -0300)
author Arnaldo Carvalho de Melo <acme@redhat.com>
Fri, 15 Apr 2016 14:10:31 +0000 (11:10 -0300)
committer Arnaldo Carvalho de Melo <acme@redhat.com>
Fri, 15 Apr 2016 16:14:20 +0000 (13:14 -0300)
diff --git a/tools/perf/Documentation/perf-trace.txt b/tools/perf/Documentation/perf-trace.txt

index 2ee0c4fee18d3de3ad68b4df21646fe0a35f9af1..4e8baa75a32eaa785268b81460ca3da68a03b7de 100644 (file)
--- a/tools/perf/Documentation/perf-trace.txt
+++ b/tools/perf/Documentation/perf-trace.txt
@@ -138,6 +138,10 @@ the thread executes on the designated CPUs. Default is to monitor all CPUs.
  
          Default: 127
  
+--min-stack::
+        Set the stack depth limit when parsing the callchain, anything
+        below the specified depth will be ignored. Disabled by default.
+
  --proc-map-timeout::
         When processing pre-existing threads /proc/XXX/mmap, it may take a long time,
         because the file may be huge. A time out is needed in such cases.
diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c

index 65f6abe75d714a158d3252e1e12632dcfc20c7ba..6a64cb1344c701c39e41ce62499220f639740dd2 100644 (file)
--- a/tools/perf/builtin-trace.c
+++ b/tools/perf/builtin-trace.c
@@ -108,6 +108,7 @@ struct trace {
                                 proc_getname;
         } stats;
         unsigned int            max_stack;
+       unsigned int            min_stack;
         bool                    not_ev_qualifier;
         bool                    live;
         bool                    full_time;
@@ -1849,7 +1850,7 @@ static int trace__sys_enter(struct trace *trace, struct perf_evsel *evsel,
                         goto out_put;
         }
  
-       if (!(trace->duration_filter || trace->summary_only))
+       if (!(trace->duration_filter || trace->summary_only || trace->min_stack))
                 trace__printf_interrupted_entry(trace, sample);
  
         ttrace->entry_time = sample->time;
@@ -1860,7 +1861,7 @@ static int trace__sys_enter(struct trace *trace, struct perf_evsel *evsel,
                                            args, trace, thread);
  
         if (sc->is_exit) {
-               if (!trace->duration_filter && !trace->summary_only) {
+               if (!(trace->duration_filter || trace->summary_only || trace->min_stack)) {
                         trace__fprintf_entry_head(trace, thread, 1, sample->time, trace->output);
                         fprintf(trace->output, "%-70s\n", ttrace->entry_str);
                 }
@@ -1880,26 +1881,26 @@ out_put:
         return err;
  }
  
-static int trace__fprintf_callchain(struct trace *trace, struct perf_evsel *evsel,
-                                   struct perf_sample *sample)
+static int trace__resolve_callchain(struct trace *trace, struct perf_evsel *evsel,
+                                   struct perf_sample *sample,
+                                   struct callchain_cursor *cursor)
  {
         struct addr_location al;
+
+       if (machine__resolve(trace->host, &al, sample) < 0 ||
+           thread__resolve_callchain(al.thread, cursor, evsel, sample, NULL, NULL, trace->max_stack))
+               return -1;
+
+       return 0;
+}
+
+static int trace__fprintf_callchain(struct trace *trace, struct perf_sample *sample)
+{
         /* TODO: user-configurable print_opts */
         const unsigned int print_opts = EVSEL__PRINT_SYM |
                                         EVSEL__PRINT_DSO |
                                         EVSEL__PRINT_UNKNOWN_AS_ADDR;
  
-       if (sample->callchain == NULL)
-               return 0;
-
-       if (machine__resolve(trace->host, &al, sample) < 0 ||
-           thread__resolve_callchain(al.thread, &callchain_cursor, evsel,
-                                     sample, NULL, NULL, trace->max_stack)) {
-               pr_err("Problem processing %s callchain, skipping...\n",
-                       perf_evsel__name(evsel));
-               return 0;
-       }
-
         return sample__fprintf_callchain(sample, 38, print_opts, &callchain_cursor, trace->output);
  }
  
@@ -1910,7 +1911,7 @@ static int trace__sys_exit(struct trace *trace, struct perf_evsel *evsel,
         long ret;
         u64 duration = 0;
         struct thread *thread;
-       int id = perf_evsel__sc_tp_uint(evsel, id, sample), err = -1;
+       int id = perf_evsel__sc_tp_uint(evsel, id, sample), err = -1, callchain_ret = 0;
         struct syscall *sc = trace__syscall_info(trace, evsel, id);
         struct thread_trace *ttrace;
  
@@ -1942,6 +1943,15 @@ static int trace__sys_exit(struct trace *trace, struct perf_evsel *evsel,
         } else if (trace->duration_filter)
                 goto out;
  
+       if (sample->callchain) {
+               callchain_ret = trace__resolve_callchain(trace, evsel, sample, &callchain_cursor);
+               if (callchain_ret == 0) {
+                       if (callchain_cursor.nr < trace->min_stack)
+                               goto out;
+                       callchain_ret = 1;
+               }
+       }
+
         if (trace->summary_only)
                 goto out;
  
@@ -1982,7 +1992,10 @@ signed_print:
  
         fputc('\n', trace->output);
  
-       trace__fprintf_callchain(trace, evsel, sample);
+       if (callchain_ret > 0)
+               trace__fprintf_callchain(trace, sample);
+       else if (callchain_ret < 0)
+               pr_err("Problem processing %s callchain, skipping...\n", perf_evsel__name(evsel));
  out:
         ttrace->entry_pending = false;
         err = 0;
@@ -2131,7 +2144,10 @@ static int trace__event_handler(struct trace *trace, struct perf_evsel *evsel,
  
         fprintf(trace->output, ")\n");
  
-       trace__fprintf_callchain(trace, evsel, sample);
+       if (sample->callchain) {
+               if (trace__resolve_callchain(trace, evsel, sample, &callchain_cursor) == 0)
+                       trace__fprintf_callchain(trace, sample);
+       }
  
         return 0;
  }
@@ -3082,6 +3098,9 @@ int cmd_trace(int argc, const char **argv, const char *prefix __maybe_unused)
                      &record_parse_callchain_opt),
         OPT_BOOLEAN(0, "kernel-syscall-graph", &trace.kernel_syscallchains,
                     "Show the kernel callchains on the syscall exit path"),
+       OPT_UINTEGER(0, "min-stack", &trace.min_stack,
+                    "Set the minimum stack depth when parsing the callchain, "
+                    "anything below the specified depth will be ignored."),
         OPT_UINTEGER(0, "max-stack", &trace.max_stack,
                      "Set the maximum stack depth when parsing the callchain, "
                      "anything beyond the specified depth will be ignored. "
author	Arnaldo Carvalho de Melo <acme@redhat.com>
	Fri, 15 Apr 2016 14:10:31 +0000 (11:10 -0300)
committer	Arnaldo Carvalho de Melo <acme@redhat.com>
	Fri, 15 Apr 2016 16:14:20 +0000 (13:14 -0300)
tools/perf/Documentation/perf-trace.txt		patch \| blob \| history
tools/perf/builtin-trace.c		patch \| blob \| history