perf tools: Fix segfault in cumulative.callchain report

heya,
not completely sure this is what we want to do, but have no streght to dive into annotation code ;-)

---
When cumulative callchain mode is on, we could get samples with with no actual hits. This breaks the assumption of the annotation code, that each sample has annotation counts allocated and leads to segfault.

Fixing this by additional checks for annotation stats.

Cc Arnaldo Carvalho de Melo <[email protected]>

Cc Corey Ashford <[email protected]>

Cc David Ahern <[email protected]>

Cc Frederic Weisbecker <[email protected]>

Cc Ingo Molnar <[email protected]>

Cc Namhyung Kim <[email protected]>

Cc Paul Mackerras <[email protected]>

Cc Peter Zijlstra <[email protected]>

Signed-off-by Jiri Olsa <[email protected]>

---
 tools/perf/ui/browsers/hists.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)

tools/perf/ui/browsers/hists.c +16 -5

index 52c03fb..04a229a 100644
@@ -17,6 +17,7 @@
 #include "../util.h"
 #include "../ui.h"
 #include "map.h"
+#include "annotate.h"
 
 struct hist_browser {
 	struct ui_browser   b;

@@ -1593,13 +1594,18 @@ static int perf_evsel__hists_browse(struct perf_evsel *evsel, int nr_events,
 					 bi->to.sym->name) > 0)
 				annotate_t = nr_options++;
 		} else {
-
 			if (browser->selection != NULL &&
 			    browser->selection->sym != NULL &&
-			    !browser->selection->map->dso->annotate_warned &&
-				asprintf(&options[nr_options], "Annotate %s",
-					 browser->selection->sym->name) > 0)
-				annotate = nr_options++;
+			    !browser->selection->map->dso->annotate_warned) {
+				struct annotation *notes;
+
+				notes = symbol__annotation(browser->selection->sym);
+
+				if (notes->src &&
+				    asprintf(&options[nr_options], "Annotate %s",
+						 browser->selection->sym->name) > 0)
+					annotate = nr_options++;
+			}
 		}
 
 		if (thread != NULL &&

@@ -1656,6 +1662,7 @@ retry_popup_menu:
 
 		if (choice == annotate || choice == annotate_t || choice == annotate_f) {
 			struct hist_entry *he;
+			struct annotation *notes;
 			int err;
 do_annotate:
 			if (!objdump_path && perf_session_env__lookup_objdump(env))

@@ -1679,6 +1686,10 @@ do_annotate:
 				he->ms.map = he->branch_info->to.map;
 			}
 
+			notes = symbol__annotation(he->ms.sym);
+			if (!notes->src)
+				continue;
+
 			/*
 			 * Don't let this be freed, say, by hists__decay_entry.
 			 */

Hi Jiri,

2014-06-15 (일), 10:35 +0200, Jiri Olsa:

Quoted 3 lines

heya,
not completely sure this is what we want to do, but have no streght to dive into annotation code ;-)

Currently, cumulative entries do not accounts annotation info so annotation only works for normal entries. I'm not sure we should support it or not, but anyway it also needs some updates on the annotation code IMHO.

So I'm okay with disabling annotation on pure cumulative entries.

Quoted 7 lines

Hi Jiri,

2014-06-15 (일), 10:35 +0200, Jiri Olsa:

Quoted 3 lines

heya,
not completely sure this is what we want to do, but have no streght to dive into annotation code ;-)

Quoted 6 lines

Currently, cumulative entries do not accounts annotation info so annotation only works for normal entries. I'm not sure we should support it or not, but anyway it also needs some updates on the annotation code IMHO.

So I'm okay with disabling annotation on pure cumulative entries.

is that an ack? ;-)

2014-06-15 (일), 18:34 +0200, Jiri Olsa:

Quoted 11 lines · Namhyung Kim

Quoted 10 lines

2014-06-15 (일), 10:35 +0200, Jiri Olsa:

Quoted 2 lines

not completely sure this is what we want to do, but have no streght to dive into annotation code ;-)

Currently, cumulative entries do not accounts annotation info so annotation only works for normal entries. I'm not sure we should support it or not, but anyway it also needs some updates on the annotation code IMHO.

So I'm okay with disabling annotation on pure cumulative entries.

Quoted 1 line

is that an ack? ;-)

Yep,

Acked-by Namhyung Kim <[email protected]>

From a quick look, this looks sane, i.e. only offering annotation for