Re: Questions about copy-on-write

[Marcus Brinkmann]

At Thu, 28 Oct 2004 00:53:28 +0100,
Sam Mason <address@hidden> wrote:
> >In the HURD I think it's convention that a user task will revoke a 
> >mapping from another task by asking physmem to do it.
> 
> OK, is there any reason for that?  Is it anything to do with the
> guaranteed page accounting process in physmem and if the task forgets
> to tell physmem that it's not using a page anymore it'll be counted as
> being used?

As the memory is mapped from physmem, only physmem can unmap it.

As to why all memory is mapped from physmem, the reason is that we (a)
don't want to rely on direct task-to-task mapping in our design
because then we lose control over page table consumption, and (b) we
don't have any good ideas how to use recursive mappings in a Hurdish
way (and there are always trust issues, too).

Thanks,
Marcus