Daniel Phillips <phillips@bonn-fries.net> writes:
> On February 19, 2002 01:03 am, Hugh Dickins wrote:
> > On Tue, 19 Feb 2002, Daniel Phillips wrote:
> > > On February 18, 2002 08:04 pm, Hugh Dickins wrote:
> > > > On Mon, 18 Feb 2002, Daniel Phillips wrote:
> > > > > On February 18, 2002 09:09 am, Hugh Dickins wrote:
> > > > > > Since copy_page_range would not copy shared page tables, I'm wrong to
> > > > > > point there. But __pte_alloc does copy shared page tables (to unshare
>
> > > > > > them), and needs them to be stable while it does so: so locking
> against
>
> > > > > > swap_out really is required. It also needs locking against read
> faults,
>
> > > > > > and they against each other: but there I imagine it's just a matter of
>
> > > > > > dropping the write arg to __pte_alloc, going back to pte_alloc again.
> > >
> > > I'm not sure what you mean here, you're not suggesting we should unshare the
>
> > > page table on read fault are you?
> >
> > I am. But I can understand that you'd prefer not to do it that way.
> > Hugh
>
> No, that's not nearly studly enough ;-)
>
> Since we have gone to all the trouble of sharing the page table, we should
> swap in/out for all sharers at the same time. That is, keep it shared, saving
> memory and cpu.
>
> Now I finally see what you were driving at: before, we could count on the
> mm->page_table_lock for exclusion on read fault, now we can't, at least not
> when ptb->count is great than one[1]. So let's come up with something nice as
> a substitute, any suggestions?
>
> [1] I think that's a big, broad hint.
Something like:
struct mm_share {
spinlock_t page_table_lock;
struct list_head mm_list;
};
struct mm {
struct list_head mm_list;
struct mm_share *mm_share;
.....
};
So we have an overarching structure for all of the shared mm's.
Eric
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Feb 23 2002 - 21:00:18 EST