Re: who pages the pager?

[Simon Nieuviarts]

Le Mardi 18 Octobre 2005 17:43, Neal H. Walfield a écrit :
> At Tue, 18 Oct 2005 17:32:46 +0200,
>
> Simon Nieuviarts wrote:
> > And what if the pager does not answer ?
>
> Then you shouldn't have trusted the pager to begin with.  In this
> case, the pager is necessarily part of the task's TCB.
>
> > I think the pager can only give hints
> > to an external paging task that takes the real decisions. The pager can
> > not be considered as reliable.
>
> I don't understand these claims.  What kind of hints is the pager
> going to give?  As far as I can see, there are no hints to give, only
> instructions.  For instance: "if I fault at this address, map this
> frame".  If the pager's pager maps a different frame then the task is
> now broken.

I may have been completely wrong. I was seeing self-paging as a way to let the 
process decide which frame to page out when the system is out of free 
physical frames. The tasks that need memory have to take it from other 
processes, and these processes decide which part of their frames they can 
give.
How is this sharing of memory done with self-paging ? Do you have a link to a 
paper on this topic ?

Simon.