Re: [PATCH] mm,migration: Avoid race between shift_arg_pages() andrmap_walk() during migration by not migrating temporary stacks

[Rik van Riel]

On 05/11/2010 01:11 PM, Linus Torvalds wrote:
> On Tue, 11 May 2010, Mel Gorman wrote:
> > This patch closes the most important race in relation to exec and
> > migration. With it applied, the swapops bug is no longer triggering for
> > known problem workloads. If you pick it up, it should go with the other
> > mmmigration-* fixes in mm.
>
> Ack. _Much_ better and clearer.
>
> I'm not entirely sure we need that "maybe_stack" (if we need it, that
> would sound like a problem anyway), but I guess it can't hurt either.

Just a heads up - I am looking at creating a patch now that
allows us to _always_ lock the root anon_vma lock, when locking
the anon_vma.

That should take care of the other issue pretty cleanly, while
still allowing us to only walk the VMAs we have to walk in
places like rmap_walk.

--
All rights reversed
--
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/