Messages in this thread | | | Date | Thu, 19 Dec 2013 20:08:46 +0100 | From | Oleg Nesterov <> | Subject | [PATCH 0/1] mm: fix the theoretical compound_lock() vs prep_new_page() race |
| |
On 12/16, Andrea Arcangeli wrote: > > Can you reorder set_page_refcount in your v2?
Please see the patch.
> I wonder if arch_alloc_page needs refcount 1, it sets the page as > stable on s390.
Obviously I have no idea what set_page_stable() does, but it works with page_to_phys(), unlikely the content of "struct page" can matter. And only s390 HAVE_ARCH_ALLOC_PAGE, I added Martin and Heiko.
> the other way around is to move prep_compound_page > before set_page_refcounted (though I think if we can, keeping the > refcounted at the very last with a comment is preferable).
Yes, yes, this looks much more natural.
Oleg.
| |