Quoting Bill Studenmund <wrstuden@netbsd.org>:

> On Tue, 19 Aug 2003, Lennart Augustsson wrote:
> 
> > I'm curious, how does the original paper on SA suggest handling this?
> > Such a fundamental thing as handling page faults cannot have been
> ignored
> > in the paper, can it?
> 
> Probably. It's been ignored up until now. :-) There's only a small
> window
> of vulnerability.

Not quite, the Anderson paper says at the end of 3.1 "The only added 
complication for the kernel is that an upcall to notify the program of a 
page fault may in turn page fault on the same location; the kernel must 
check for this, and when it occurs, delay the subsequent upcall until the 
page fault completes".  Pretty much what we're trying to do.

> > How does FreeBSD handle?  How does Sun handle it?
> 
> They don't have SA phreads. While Sun has LWPs, they dont have the sam
> split scheduler that SA implementations have.

Well, they have "Kernel Scheduler Entities (KSE), [which] is a kernel-
supported threading system similar in design to Scheduler Activations".

-- 
Christian Limpach <chris@pin.lu>