On Sunday 08 November 2015 06:01:08 Octavian Purdila wrote:
> >> +static void *sem_alloc(int count)
> >> +{
> >> + struct pthread_sem *sem;
> >> +
> >> + sem = malloc(sizeof(*sem));
> >> + if (!sem)
> >> + return NULL;
> >> +
> >> + pthread_mutex_init(&sem->lock, NULL);
> >> + sem->count = count;
> >> + pthread_cond_init(&sem->cond, NULL);
> >> +
> >> + return sem;
> >> +}
> >
> > What is the reason to have generalized semaphores in the
> > host API rather than a simple mutex?
> >
>
> Currently waking up from idle after an IRQ event requires a semaphore.
> I'll see if we can use a simple mutex for this.
According to the pthread_mutex_unlock() man page, you are not allowed
to unlock a mutex from any thread other than the one that owns the
mutex through pthread_mutex_lock(), so if the IRQ event is sent to
another thread, that would not be safe even if it happens to work
on linux+glibc.
Another option would be to use futexes as the basic primitive, which
might make the implementation for Linux hosts a bit more efficient,
but complicates the implementation for hosts that do not implement
those.
Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-arch" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
