On Thu, 2018-01-11 at 08:39 +0100, Hannes Reinecke wrote: > I'm actually not that convinced with this change; originally we had been > checking if it's on the wait list, and only _then_ call bt_wait_ptr(). > Now we call bt_wait_ptr() always, meaning we run a chance of increasing > the bitmap_tags wait pointer without actually using it. > Looking at the code I'm not sure this is the correct way of using it ... Hello Hannes, Are you perhaps referring to sbq_index_atomic_inc()? I think it's fine to call bt_wait_ptr() even if the corresponding waitqueue won't be used. My understanding is that the purpose of having multiple waitqueues and of using these in a round-robin fashion is to avoid that if less than or equal to 8 (SBQ_WAIT_QUEUES) threads are waiting for a tag that these won't use the same wait queue. So in the unlikely event of an early return the worst that can happen is that there is more contention on one of the eight wait queues. But since that is an unlikely scenario I don't think we have to worry about this. Bart.
