On 12/13, Andrea Arcangeli wrote:
>
> On Fri, Dec 13, 2013 at 05:22:40PM +0100, Oleg Nesterov wrote:
> >
> > I'll try to make v2 based on -mm and your suggestions.
>
> Ok great!
Yes, it would be great, but I need your help again ;)
Let me quote the pseudo-code you sent me:
put_compound_tail(page) {
page_head = compound_trans_head(page);
if (!__compound_tail_refcounted(page_head)) {
...
return ...;
}
flags = compound_lock_irqsave(page_head);
...
Sure, put_compound_tail() should be the simplified version of
put_compound_page() which doesn't dec page_head->_count, this is clear.
But afaics, compound_lock_irqsave() above looks unsafe without
get_page_unless_zero(page_head) ? If we race with _split, page_head
can be freed and compound_lock() can race with, say, free_pages_check()
which plays with page->flags ?
So it seems that put_compound_tail() should also do get/put(head) like
put_compound_page() does, and this probably means we should factor out
the common code somehow.
Or I missed something?
OTOH, I can't really understand
if (likely(page != page_head && get_page_unless_zero(page_head)))
in __get_page_tail() and put_compound_page().
First of all, is it really possible that page == compound_trans_head(page)?
We already verified that PG_tail was set. Of course this bit can be cleared
and ->first_page can be a dangling pointer but it can never be changed to
point to this page? (in fact, afaics it can be changed at all as long as we
have a reference, but this doesn't matter).
And compound_lock_irqsave() looks racy even after get_page_unless_zero().
For example, suppose that page_head was already freed and then re-allocated
as (say) alloc_pages(__GFP_COMP, 1). get_page_unless_zero() can succeed right
after prep_new_page() does set_page_refcounted(). Now, can't compound_lock()
race with the non-atomic prep_compound_page()->__SetPageHead() ?
Finally. basepage_index(page) after put_page(page) in get_futex_key() looks
confusing imho. I think this is correct, we already checked PageAnon() so it
can't be a thp page. But probably this needs a comment and __basepage_index()
should do BUG_ON(!PageHuge()).
Oleg.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
