Hello everybody,
let's consider such a situation: you have a priority 6 thread that's running, preventing a priority 3 thread from receiving CPU time; however, a priority 10 thread is waiting for some resource that the priority 3 thread has locked. Since priority 6 thread is "eating" all CPU time, priority 3 thread won't ever be allowed to release its resources, so effectively blocking priority 10 thread.
This problem is known as "priority inversion", and is solved in Windows kernel using a mechanism called "CPU starvation", not all satisfying.
How is it managed in Linux kernel?
Kisses,

Nat

First off, in linux, a priority 3 thread would execute before a priority 6 thread. (Lowest number == highest priority).

But that aside, once the priority 6 thread has completed it's time slice, it is moved to the wait queue for that processor and the priority 3 thread would run, release its lock and then move to the wait queue. Now the priority 10 thread would be able to do its work.

This is all a function of the O(1) scheduler in the 2.6 kernel series.

Beg your pardon for mistake.
More infoes, even an URL?

This is a pretty good article describing the behavior of the 2.6 scheduler: http://www-128.ibm.com/developerwork...y/l-scheduler/