Re: [patch 1/8] fix race in clear_page_dirty_for_io()

From: Miklos Szeredi
Date: Wed Mar 07 2007 - 05:32:37 EST


> (cc's reinstated)
>
> On Wed, 07 Mar 2007 09:09:50 +0100 Miklos Szeredi <miklos@xxxxxxxxxx> wrote:
>
> > There's a race in clear_page_dirty_for_io() that allows a page to have
> > cleared PG_dirty, while being mapped read-write into the page table(s).
>
> I assume you refer to this:
>
> * FIXME! We still have a race here: if somebody
> * adds the page back to the page tables in
> * between the "page_mkclean()" and the "TestClearPageDirty()",
> * we might have it mapped without the dirty bit set.
> */
> if (page_mkclean(page))
> set_page_dirty(page);
> if (TestClearPageDirty(page)) {
> dec_zone_page_state(page, NR_FILE_DIRTY);
> return 1;
> }
>

Yes.

> I guess the comment actually refers to a writefault after the
> set_page_dirty() and before the TestClearPageDirty(). The fault handler
> will run set_page_dirty() and will return to userspace to rerun the write.
> The page then gets set pte-dirty but this thread of control will now make
> the page !PageDirty() and will write it out.

Yes.

> With Nick's proposed lock-the-page-in-pagefaults patches, we have
> lock_page() synchronisation between pagefaults and
> clear_page_dirty_for_io() which I think will fix this.

After a quick look, I don't think it does. It locks the page in
do_no_page(), but not for the whole fault. In particular do_wp_page()
is not affected. But I haven't yet looked closely at that patch, so I
could be wrong.

Miklos
-
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/