> You can optimize a bit by giving the wait queue a priority (and a
> count of how many mebers of the queue are at this priority). When
> a higher priority process arrives, you change the priority (and
> set the count to 1). When all processes at some given priority
> leave, you walk the queue.
the problem is, that in the aforementioned case, there is no connection
leading to raid5d (i dont know the NFS code that well, do we still have
biod's that act as IO agents?). raid5d has an anonymous bh, which will
lead the the page's waitqueue when the IO finishes, but there is no way we
can get to raid5d from the page waitqueue. (it's not even repeatable, eg.
raid1d does read balancing, so a given request can end up anywhere).
-- mingo
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu