Re: [PATCH v2] perf record: Synthesize COMM event for a command line workload

[Namhyung Kim <namhyung@kernel.org>]

On Wed, Sep 23, 2015 at 12:09:20AM -0300, Arnaldo Carvalho de Melo wrote:
> Le 23 sept. 2015 12:04 AM, "Namhyung Kim" <namhyung@kernel.org> a écrit :
> >
> > Hi Arnaldo,
> >
> > On Tue, Sep 22, 2015 at 04:48:25PM -0300, Arnaldo Carvalho de Melo wrote:
> > > Humm, I think it is working by accident, as you're not allocating space
> > > for machine->id_hdr_size, please take a look at
> > > perf_event__synthesize_thread_map().
> > >
> > > Right now its not a problem this line in perf_event__prepare_comm():
> > >
> > >         memset(event->comm.comm + size, 0, machine->id_hdr_size);
> > >
> > > Because perf_event is an union and some of its elements, like mmap/mmap2
> > > have that PATH_MAX part, but its just a matter of the id_hdr_size
> > > becoming bigger than that and we'll have a problem...
> >
> > Right. I'll send a fix to include the id_hdr part.
> 
> Thanks!


>From 10123021c4e55f14d3bb8e6cc576694bb76d6699 Mon Sep 17 00:00:00 2001
From: Namhyung Kim <namhyung@kernel.org>
Date: Wed, 23 Sep 2015 11:51:40 +0900
Subject: [PATCH] perf record: Allocate area for sample_id_hdr in a synthesized
 comm event

A previous patch added a synthesized comm event for forked child
process but it missed that the event should contain area for
sample_id_hdr at the end.  It worked by accident since the perf_event
union contains bigger event structs like mmap_events.

This patch fixes it by dynamically allocating event struct including
those area like in perf_event__synthesize_thread_map().

Reported-by: Arnaldo Carvalho de Melo <acme@kernel.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/builtin-record.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 6d8302d4612f..5e01c070dbf2 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -637,7 +637,13 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
 	 * Let the child rip
 	 */
 	if (forks) {
-		union perf_event event;
+		union perf_event *event;
+
+		event = malloc(sizeof(event->comm) + machine->id_hdr_size);
+		if (event == NULL) {
+			err = -ENOMEM;
+			goto out_child;
+		}
 
 		/*
 		 * Some H/W events are generated before COMM event
@@ -645,10 +651,11 @@ static int __cmd_record(struct record *rec, int argc, const char **argv)
 		 * cannot see a correct process name for those events.
 		 * Synthesize COMM event to prevent it.
 		 */
-		perf_event__synthesize_comm(tool, &event,
+		perf_event__synthesize_comm(tool, event,
 					    rec->evlist->workload.pid,
 					    process_synthesized_event,
-					    &session->machines.host);
+					    machine);
+		free(event);
 
 		perf_evlist__start_workload(rec->evlist);
 	}
-- 
2.5.0