On 21/01/2015 17:28, Jiri Slaby wrote:
> + if (atomic_fetch_or(&edu->status, EDU_STATUS_COMPUTING) &
> EDU_STATUS_COMPUTING) {
Theoretically the other thread could see EDU_STATUS_COMPUTING here and
not enter the condvar wait. So...
> + break;
> + }
> + qemu_mutex_lock(&edu->thr_mutex);
> + edu->fact = val;
> + qemu_cond_signal(&edu->thr_cond);
> + qemu_mutex_unlock(&edu->thr_mutex);
... just one change:
if (atomic_read(&edu->status) & EDU_STATUS_COMPUTING) {
break;
}
/* EDU_STATUS_COMPUTING cannot go 0->1 concurrently, because
* it is only set in this function and it is under the iothread
* mutex.
*/
qemu_mutex_lock(&edu->thr_mutex);
edu->fact = val;
atomic_or(&edu->status), EDU_STATUS_COMPUTING);
qemu_cond_signal(&edu->thr_cond);
qemu_mutex_unlock(&edu->thr_mutex);
If you are okay with this change, I'll apply the patch to my tree.
Paolo
