On Wed, 9 Jun 2010 17:19:02 +0800
Dave Young <hidave.darkstar@xxxxxxxxx> wrote:
> > Manually bisected mm patches, the memleak caused by following patch:
> >
> > mm-extend-ksm-refcounts-to-the-anon_vma-root.patch
>
>
> So I guess the refcount break, either drop-without-get or over-drop
I'm guessing I did not run the kernel with enough debug options enabled
when I tested my patches...
Dave & Catalin, thank you for tracking this down.
Dave, does the below patch fix your issue?
Andrew, if the patch below works, you'll probably want to merge it as
mm-extend-ksm-refcounts-to-the-anon_vma-root-fix.patch :)
----------------
With the new anon_vma code we take a refcount on the root anon_vma.
However, the root anon_vma does not have a refcount on itself, so
we should not try to do a drop on itself when it is being unlinked.
Signed-off-by: Rik van Riel <riel@xxxxxxxxxx>
--- linux-2.6-rtavma/mm/rmap.c.orig 2010-06-09 21:10:07.349376896 -0400
+++ linux-2.6-rtavma/mm/rmap.c 2010-06-09 21:10:24.180406299 -0400
@@ -275,7 +275,8 @@ static void anon_vma_unlink(struct anon_
if (empty) {
/* We no longer need the root anon_vma */
- drop_anon_vma(anon_vma->root);
+ if (anon_vma->root != anon_vma)
+ drop_anon_vma(anon_vma->root);
anon_vma_free(anon_vma);
}
}
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>