Re: [patch] scheduler fixes, 2.5.32-BK

[Ingo Molnar]

On Fri, 30 Aug 2002, Andrew Morton wrote:

> yep, looks like the killing of the semaphore_lock made the race go away.
> 
> But ia64, sparc and x86_64 use semaphore_lock, so they still are
> exposed.

btw., this way completions become special-case semaphores optimized for
the context-switching path, not a correctness issue. At which point it's
also an interesting question whether in fact we need to make completion()  
that much of a different interface?

	Ingo