perf evlist: Avoid iteration for empty evlist.
Ian Rogers
reviewerPATCH 1/1 perf evlist: Avoid iteration for empty evlist.
As seen with 'perf stat --null ..' and reported in:
https://lore.kernel.org/lkml/[email protected]/
--- tools/perf/util/evlist.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
tools/perf/util/evlist.c +4 -1
@@ -354,7 +354,10 @@ struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affin .affinity = affinity, }; -if (itr.affinity) { +if (evlist__empty (evlist)) { +/* Ensure the empty list doesn't iterate. */ + itr.evlist_cpu_map_idx = itr.evlist_cpu_map_nr; + }else if (itr.affinity) { itr.cpu =perf_cpu_map__cpu (evlist->core.all_cpus,0 );affinity__set (itr.affinity, itr.cpu.cpu); itr.cpu_map_idx =perf_cpu_map__idx (itr.evsel->core.cpus, itr.cpu);
Signature · 1 line
2.35.1.723.g4982287a31-goog
Quoted 14 lines · Ian Rogers
As seen with 'perf stat --null ..' and reported in:
https://lore.kernel.org/lkml/[email protected]/Fixes 472832d2c000 ("perf evlist: Refactor evlist__for_each_cpu()")Signed-off-by Ian Rogers <[email protected]>--- tools/perf/util/evlist.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)tools/perf/util/evlist.c +0
-0
tools/perf/util/evlist.c struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affin · line 359
Quoted 7 lines
tools/perf/util/evlist.c +3
-1@@ -354,7 +354,10 @@ struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affin .affinity = affinity, }; -if (itr.affinity) { +if (evlist__empty (evlist)) { +/* Ensure the empty list doesn't iterate. */ + itr.evlist_cpu_map_idx = itr.evlist_cpu_map_nr;
I can't see the crash anymore, but I'm bit confused with the code
if evlist is empty then itr.evsel is bogus.. and the loop code __run_perf_stat is just lucky, right?
I think we need to set itr.evsel to NULL and skip the loop in case evlist is empty
thanks,
jirka
Quoted 7 lines
tools/perf/util/evlist.c +1
-0+ }else if (itr.affinity) { itr.cpu =perf_cpu_map__cpu (evlist->core.all_cpus,0 );affinity__set (itr.affinity, itr.cpu.cpu); itr.cpu_map_idx =perf_cpu_map__idx (itr.evsel->core.cpus, itr.cpu);Signature · 1 line
2.35.1.723.g4982287a31-goog
Ian Rogers
reviewerQuoted 26 lines · Jiri Olsa
Quoted 14 lines · Ian Rogers
As seen with 'perf stat --null ..' and reported in:
https://lore.kernel.org/lkml/[email protected]/Fixes 472832d2c000 ("perf evlist: Refactor evlist__for_each_cpu()")Signed-off-by Ian Rogers <[email protected]>--- tools/perf/util/evlist.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)tools/perf/util/evlist.c +0
-0tools/perf/util/evlist.c struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affin · line 359
Quoted 7 lines
tools/perf/util/evlist.c +3
-1@@ -354,7 +354,10 @@ struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affin .affinity = affinity, }; -if (itr.affinity) { +if (evlist__empty (evlist)) { +/* Ensure the empty list doesn't iterate. */ + itr.evlist_cpu_map_idx = itr.evlist_cpu_map_nr;I can't see the crash anymore, but I'm bit confused with the code
Quoted 2 lines
if evlist is empty then itr.evsel is bogus.. and the loop code
__run_perf_statis just lucky, right?
The itr.evsel is the list head, so bogus.
Quoted 2 lines
I think we need to set itr.evsel to NULL and skip the loop in case evlist is empty
So that's the effect of this change except that the evsel is the list head. I'm not sure it is worth adding the condition to setting the evsel to capture that given that with the end condition having been met it would be invalid to read it. There's a similar problem for the other fields of the iterator.
Thanks,
Ian
Quoted 10 lines
thanks,
jirkaQuoted 7 lines
tools/perf/util/evlist.c +1
-1+ }else if (itr.affinity) { itr.cpu =perf_cpu_map__cpu (evlist->core.all_cpus,0 );affinity__set (itr.affinity, itr.cpu.cpu); itr.cpu_map_idx =perf_cpu_map__idx (itr.evsel->core.cpus, itr.cpu); --2.35.1.723.g4982287a31-goog
Quoted 32 lines · Ian Rogers
Quoted 26 lines · Jiri Olsa
Quoted 14 lines · Ian Rogers
As seen with 'perf stat --null ..' and reported in:
https://lore.kernel.org/lkml/[email protected]/Fixes 472832d2c000 ("perf evlist: Refactor evlist__for_each_cpu()")Signed-off-by Ian Rogers <[email protected]>--- tools/perf/util/evlist.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)tools/perf/util/evlist.c +0
-0tools/perf/util/evlist.c struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affin · line 359
Quoted 7 lines
tools/perf/util/evlist.c +3
-1@@ -354,7 +354,10 @@ struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affin .affinity = affinity, }; -if (itr.affinity) { +if (evlist__empty (evlist)) { +/* Ensure the empty list doesn't iterate. */ + itr.evlist_cpu_map_idx = itr.evlist_cpu_map_nr;I can't see the crash anymore, but I'm bit confused with the code
Quoted 2 lines
if evlist is empty then itr.evsel is bogus.. and the loop code
__run_perf_statis just lucky, right?The itr.evsel is the list head, so bogus.
Quoted 8 lines
Quoted 2 lines
I think we need to set itr.evsel to NULL and skip the loop in case evlist is empty
So that's the effect of this change except that the evsel is the list head. I'm not sure it is worth adding the condition to setting the evsel to capture that given that with the end condition having been met it would be invalid to read it. There's a similar problem for the other fields of the iterator.
ah ok, I got it.. but it seems to be just evsel that's wrong, just as a precaution, could we set it only when it's valid? with something like below
thanks,
jirka
---
tools/perf/util/evlist.c +1 -1
@@ -346,7 +346,7 @@ struct evlist_cpu_iterator evlist__cpu_begin(struct evlist *evlist, struct affin {struct evlist_cpu_iterator itr = { .container = evlist, - .evsel =evlist__first (evlist), + .evsel =evlist__empty (evlist) ?NULL :evlist__first (evlist), .cpu_map_idx =0 , .evlist_cpu_map_idx =0 , .evlist_cpu_map_nr =perf_cpu_map__nr (evlist->core.all_cpus),