Re: [patch 2/3] mm: speculative get_page

[Nick Piggin]

Christoph Lameter wrote:
> Looks like the NoNewRefs flag is mostly == 
> spin_is_locked(mapping->tree_lock)? Would it not be better to check the 
> tree_lock?

Well there are other uses for the tree_lock (eg. tag operations)
which do not need the "no new references" guarantee.

> > --- linux-2.6.orig/mm/migrate.c
> > +++ linux-2.6/mm/migrate.c
> >
> > +	SetPageNoNewRefs(page);
> > 	write_lock_irq(&mapping->tree_lock);
>
>
> A dream come true! If this is really working as it sounds then we can 
> move the SetPageNoNewRefs up and avoid the final check under 
> mapping->tree_lock. Then keep SetPageNoNewRefs until the page has been 
> copied. It would basically play the same role as locking the page.

Yes we could do that but at this stage I wouldn't like to seperate
SetPageNoNewRefs from tree_lock, as it is replacing a traditional
guarantee that tree_lock no longer provides.

--
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com 
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/