perf map: Synthesize maps only for thread group leader
Threads share map_groups, all map events are merged into it.
Thus we could send mmaps only for thread group leader. Otherwise it
took ages to attach and record something from processes with many vmas
and threads.
Thread group leader could be already dead, but it seems perf cannot
handle this case anyway.
Testing dummy:
#include <stdio.h>
#include <stdlib.h>
#include <sys/mman.h>
#include <pthread.h>
#include <unistd.h>
void *thread(void *arg) {
pause();
}
int main(int argc, char **argv) {
int threads = 10000;
int vmas = 50000;
pthread_t th;
for (int i = 0; i < threads; i++)
pthread_create(&th, NULL, thread, NULL);
for (int i = 0; i < vmas; i++)
mmap(NULL, 4096, (i & 1) ? PROT_READ : PROT_WRITE,
MAP_PRIVATE | MAP_ANONYMOUS | MAP_NORESERVE, -1, 0);
sleep(60);
return 0;
}
Comment by Jiri Olsa:
We actualy synthesize the group leader (if we found one) for the thread
even if it's not present in the thread_map, so the process maps are
always in data.
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/153363294102.396323.6277944760215058174.stgit@buzz
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c
index 0c8ecf0..0cd4215 100644
--- a/tools/perf/util/event.c
+++ b/tools/perf/util/event.c
@@ -541,10 +541,17 @@
tgid, process, machine) < 0)
return -1;
+ /*
+ * send mmap only for thread group leader
+ * see thread__init_map_groups
+ */
+ if (pid == tgid &&
+ perf_event__synthesize_mmap_events(tool, mmap_event, pid, tgid,
+ process, machine, mmap_data,
+ proc_map_timeout))
+ return -1;
- return perf_event__synthesize_mmap_events(tool, mmap_event, pid, tgid,
- process, machine, mmap_data,
- proc_map_timeout);
+ return 0;
}
if (machine__is_default_guest(machine))