On Tue, 2008-03-18 at 10:26 +0100, [EMAIL PROTECTED] wrote: > > You need kernel memory for the memory maps, at least one for each user > > thread. > > No I don't. That's precisely where it is *not* equivalent. > > In the model I described, the state structures for the blocked requests > (I prefer not to talk about user threads here -- it's possible to use > that abstraction, but it only seems to be confusing) are ordinary user > space data structures, living on the heap. The kernel has nothing to do > with them.
The heap is demand paged virtual memory, for which the kernel must maintain memory maps, as Neal was saying. Thomas
