Re: why should physmem not know about which frames are extra?

[Sam Mason]

Marcus Brinkmann wrote:
>Now you know how much I know about garbage collectors.

My knowledge isn't exactly deep!  I've quickly read over Knuth's
treatment of them in TAOCP and I've tried to read a bit about more
modern approaches (one of the joys of using Java for anything
remotely serious is that you do actually have to know about this
sort of thing), but I've never gone as far as actually writing one.

>Well, the lazy approach is what Neal suggests and which is so far the
>accepted solution, even by me, and even given the obvious problems
>arising from it.

Just out of interest, what are the "obvious problems"?  I've had
a quick re-read of the thread and your argument seems to be that
if physmem 'just knew' which pages to evict then it may be able to
miss out on setting up a bit of state and doing a small amount of
IPC (which, as I've been finding out, isn't particularly expensive
under L4) to get the information on demand.

Admittedly there's still a bit of slightly thorny bit of work to be
done in figuring out how to determine if a task has gone to sleep
on us or not.

  Sam