Re: [PATCH v3 11/18] perf dso: Update use of pthread mutex

[Ian Rogers]

On Fri, Aug 26, 2022 at 10:34 AM Adrian Hunter <adrian.hunter@xxxxxxxxx> wrote:
>
> On 26/08/22 19:05, Ian Rogers wrote:
> > On Fri, Aug 26, 2022 at 3:37 AM Adrian Hunter <adrian.hunter@xxxxxxxxx> wrote:
> >>
> >> On 24/08/22 18:38, Ian Rogers wrote:
> >>> Switch to the use of mutex wrappers that provide better error checking.
> >>>
> >>> Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
> >>> ---
> >>>  tools/perf/util/dso.c    | 12 ++++++------
> >>
> >> Some not done yet
> >>
> >> $ grep -i pthread_mut tools/perf/util/dso.c
> >> static pthread_mutex_t dso__data_open_lock = PTHREAD_MUTEX_INITIALIZER;
> >>         pthread_mutex_lock(&dso__data_open_lock);
> >>         pthread_mutex_unlock(&dso__data_open_lock);
> >>         if (pthread_mutex_lock(&dso__data_open_lock) < 0)
> >>                 pthread_mutex_unlock(&dso__data_open_lock);
> >>         pthread_mutex_unlock(&dso__data_open_lock);
> >>         pthread_mutex_lock(&dso__data_open_lock);
> >>         pthread_mutex_unlock(&dso__data_open_lock);
> >>         pthread_mutex_lock(&dso__data_open_lock);
> >>         pthread_mutex_unlock(&dso__data_open_lock);
> >
> > Yes, these are all solely dso__data_open_lock that lacks any clear
> > init/exit code to place the initialization/destruction hooks onto. I
> > don't plan to alter these in this patch set.
>
> Perhaps that could be explained in the change log.
>
> But why not just add init / exit code.  Could be called out
> of main(), or maybe use __attribute__((constructor)) /
> __attribute__((destructor))

Because the lock is global and not part of the dso.

Thanks,
Ian