Re: [PATCH v2 07/18] perf record: Update use of pthread mutex

From: Ian Rogers
Date: Wed Aug 24 2022 - 11:05:13 EST


On Wed, Aug 24, 2022 at 3:15 AM Adrian Hunter <adrian.hunter@xxxxxxxxx> wrote:
>
> On 24/08/22 01:09, Ian Rogers wrote:
> > Switch to the use of mutex wrappers that provide better error checking
> > for synth_lock.
>
> It would be better to distinguish patches that make drop-in
> replacements from patches like this that change logic.

The only change here is PTHREAD_MUTEX_INITIALIZER to mutex_init
because PTHREAD_MUTEX_INITIALIZER doesn't have error checking. The two
are morally equivalent and so no logic change is intended - although
one may inadvertently happen by the moving initialization from compile
time to runtime.

> >
> > Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
> > ---
> > tools/perf/builtin-record.c | 13 +++++++++----
> > 1 file changed, 9 insertions(+), 4 deletions(-)
> >
> > diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
> > index 4713f0f3a6cf..02eb85677e99 100644
> > --- a/tools/perf/builtin-record.c
> > +++ b/tools/perf/builtin-record.c
> > @@ -21,6 +21,7 @@
> > #include "util/evsel.h"
> > #include "util/debug.h"
> > #include "util/mmap.h"
> > +#include "util/mutex.h"
> > #include "util/target.h"
> > #include "util/session.h"
> > #include "util/tool.h"
> > @@ -608,17 +609,18 @@ static int process_synthesized_event(struct perf_tool *tool,
> > return record__write(rec, NULL, event, event->header.size);
> > }
> >
> > +static struct mutex synth_lock;
> > +
> > static int process_locked_synthesized_event(struct perf_tool *tool,
> > union perf_event *event,
> > struct perf_sample *sample __maybe_unused,
> > struct machine *machine __maybe_unused)
> > {
> > - static pthread_mutex_t synth_lock = PTHREAD_MUTEX_INITIALIZER;
> > int ret;
> >
> > - pthread_mutex_lock(&synth_lock);
> > + mutex_lock(&synth_lock);
> > ret = process_synthesized_event(tool, event, sample, machine);
> > - pthread_mutex_unlock(&synth_lock);
> > + mutex_unlock(&synth_lock);
> > return ret;
> > }
> >
> > @@ -1917,6 +1919,7 @@ static int record__synthesize(struct record *rec, bool tail)
> > }
> >
> > if (rec->opts.nr_threads_synthesize > 1) {
> > + mutex_init(&synth_lock, /*pshared=*/false);
>
> It would be better to have mutex_init() and mutex_init_shared()
> since /*pshared=*/true is rarely used.

Will change in v3.

Thanks,
Ian

> > perf_set_multithreaded();
> > f = process_locked_synthesized_event;
> > }
> > @@ -1930,8 +1933,10 @@ static int record__synthesize(struct record *rec, bool tail)
> > rec->opts.nr_threads_synthesize);
> > }
> >
> > - if (rec->opts.nr_threads_synthesize > 1)
> > + if (rec->opts.nr_threads_synthesize > 1) {
> > perf_set_singlethreaded();
> > + mutex_destroy(&synth_lock);
> > + }
> >
> > out:
> > return err;
>