Em Mon, Dec 12, 2016 at 11:32:09AM -0300, Arnaldo Carvalho de Melo escreveu: > Em Mon, Dec 12, 2016 at 11:35:42AM +0100, Jiri Olsa escreveu: > > Adding perf_evsel::ignore_missing_cpu_thread bool. > > > > When set true, it allows perf to ignore error of missing > > pid of perf event syscall. > > > > We remove missing thread id from the thread_map, so the > > rest of the processing like ioctl and mmap won't get > > disturbed with -1 fd > > What was the test performed? I tried it here the following while doing > a 'make -j4 allmodconfig' in the background, i.e. tons of threads coming > and going, and I sometimes get: > > [root@jouet ~]# perf record -u acme > Error: > The sys_perf_event_open() syscall returned with 3 (No such process) for > event (cycles:ppp). > /bin/dmesg may provide additional information. > No CONFIG_PERF_EVENTS=y kernel support configured? > [root@jouet ~]#
So I applied the first three patches, will continue after discussing this. - Arnaldo > > The reason for supporting this is to ease up monitoring > > group of pids, that 'disappear' before perf opens their > > event. This currently leads perf to report error and exit > > and makes perf record's -u option unusable under certain > > setup. > > > > With this change we will allow this race and ignore such > > failure with following warning: > > > > WARNING: Ignored open failure for pid 8605 > > > > Signed-off-by: Jiri Olsa <[email protected]> > > Cc: David Ahern <[email protected]> > > Cc: Namhyung Kim <[email protected]> > > Cc: Peter Zijlstra <[email protected]> > > Link: http://lkml.kernel.org/n/[email protected] > > --- > > tools/perf/perf.h | 1 + > > tools/perf/util/evsel.c | 44 ++++++++++++++++++++++++++++++++++++++++++++ > > tools/perf/util/evsel.h | 1 + > > 3 files changed, 46 insertions(+) > > > > diff --git a/tools/perf/perf.h b/tools/perf/perf.h > > index 9a0236a4cf95..1c27d947c2fe 100644 > > --- a/tools/perf/perf.h > > +++ b/tools/perf/perf.h > > @@ -55,6 +55,7 @@ struct record_opts { > > bool all_user; > > bool tail_synthesize; > > bool overwrite; > > + bool ignore_missing_thread; > > unsigned int freq; > > unsigned int mmap_pages; > > unsigned int auxtrace_mmap_pages; > > diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c > > index fd61ebd77c26..ac87f1637ae9 100644 > > --- a/tools/perf/util/evsel.c > > +++ b/tools/perf/util/evsel.c > > @@ -990,6 +990,8 @@ void perf_evsel__config(struct perf_evsel *evsel, > > struct record_opts *opts, > > * it overloads any global configuration. > > */ > > apply_config_terms(evsel, opts); > > + > > + evsel->ignore_missing_thread = opts->ignore_missing_thread; > > } > > > > static int perf_evsel__alloc_fd(struct perf_evsel *evsel, int ncpus, int > > nthreads) > > @@ -1419,6 +1421,33 @@ static int __open_attr__fprintf(FILE *fp, const char > > *name, const char *val, > > return fprintf(fp, " %-32s %s\n", name, val); > > } > > > > +static bool ignore_missing_thread(struct perf_evsel *evsel, > > + struct thread_map *threads, > > + int thread, int err) > > +{ > > + if (!evsel->ignore_missing_thread) > > + return false; > > + > > + /* The system wide setup does not work with threads. */ > > + if (!evsel->system_wide) > > + return false; > > + > > + /* The -ESRCH is perf event syscall errno for pid's not found. */ > > + if (err != -ESRCH) > > + return false; > > + > > + /* If there's only one thread, let it fail. */ > > + if (threads->nr == 1) > > + return false; > > + > > + if (thread_map__remove(threads, thread)) > > + return false; > > + > > + pr_warning("WARNING: Ignored open failure for pid %d\n", > > + thread_map__pid(threads, thread)); > > + return true; > > +} > > + > > static int __perf_evsel__open(struct perf_evsel *evsel, struct cpu_map > > *cpus, > > struct thread_map *threads) > > { > > @@ -1491,6 +1520,21 @@ static int __perf_evsel__open(struct perf_evsel > > *evsel, struct cpu_map *cpus, > > > > if (fd < 0) { > > err = -errno; > > + > > + if (ignore_missing_thread(evsel, threads, > > thread, err)) { > > + /* > > + * We just removed 1 thread, so take a > > step > > + * back on thread index and lower the > > upper > > + * nthreads limit. > > + */ > > + nthreads--; > > + thread--; > > + > > + /* ... and pretend like nothing have > > happened. */ > > + err = 0; > > + continue; > > + } > > + > > pr_debug2("\nsys_perf_event_open failed, error > > %d\n", > > err); > > goto try_fallback; > > diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h > > index 6abb89cd27f9..06ef6f29efa1 100644 > > --- a/tools/perf/util/evsel.h > > +++ b/tools/perf/util/evsel.h > > @@ -120,6 +120,7 @@ struct perf_evsel { > > bool tracking; > > bool per_pkg; > > bool precise_max; > > + bool ignore_missing_thread; > > /* parse modifier helper */ > > int exclude_GH; > > int nr_members; > > -- > > 2.7.4
