On Mon, Jul 29, 2013 at 04:19:10PM -0400, Naoya Horiguchi wrote:
> On Mon, Jul 29, 2013 at 02:32:07PM +0900, Joonsoo Kim wrote:
> > If we fail with a reserved page, just calling put_page() is not sufficient,
> > because put_page() invoke free_huge_page() at last step and it doesn't
> > know whether a page comes from a reserved pool or not. So it doesn't do
> > anything related to reserved count. This makes reserve count lower
> > than how we need, because reserve count already decrease in
> > dequeue_huge_page_vma(). This patch fix this situation.
>
> I think we could use a page flag (for example PG_reserve) on a hugepage
> in order to record that the hugepage comes from the reserved pool.
> Furthermore, the reserve flag would be set when dequeueing a free hugepage,
> and cleared when hugepage_fault returns, whether it fails or not.
> I think it's simpler than put_page variant approach, but doesn't it work
> to solve your problem?
Yes. That's good idea.
I thought this idea before, but didn't implement that way, because
I was worry that this may make patchset more larger and complex. But
implementing that way may be better.
Thanks.
>
> Thanks,
> Naoya Horiguchi
>
> > Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
> >
> > diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> > index bb8a45f..6a9ec69 100644
> > --- a/mm/hugetlb.c
> > +++ b/mm/hugetlb.c
> > @@ -649,6 +649,34 @@ struct hstate *size_to_hstate(unsigned long size)
> > return NULL;
> > }
> >
> > +static void put_huge_page(struct page *page, int use_reserve)
> > +{
> > + struct hstate *h = page_hstate(page);
> > + struct hugepage_subpool *spool =
> > + (struct hugepage_subpool *)page_private(page);
> > +
> > + if (!use_reserve) {
> > + put_page(page);
> > + return;
> > + }
> > +
> > + if (!put_page_testzero(page))
> > + return;
> > +
> > + set_page_private(page, 0);
> > + page->mapping = NULL;
> > + BUG_ON(page_count(page));
> > + BUG_ON(page_mapcount(page));
> > +
> > + spin_lock(&hugetlb_lock);
> > + hugetlb_cgroup_uncharge_page(hstate_index(h),
> > + pages_per_huge_page(h), page);
> > + enqueue_huge_page(h, page);
> > + h->resv_huge_pages++;
> > + spin_unlock(&hugetlb_lock);
> > + hugepage_subpool_put_pages(spool, 1);
> > +}
> > +
> > static void free_huge_page(struct page *page)
> > {
> > /*
> > @@ -2625,7 +2653,7 @@ retry_avoidcopy:
> > spin_unlock(&mm->page_table_lock);
> > mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
> >
> > - page_cache_release(new_page);
> > + put_huge_page(new_page, use_reserve);
> > out_old_page:
> > page_cache_release(old_page);
> > out_lock:
> > @@ -2725,7 +2753,7 @@ retry:
> >
> > err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
> > if (err) {
> > - put_page(page);
> > + put_huge_page(page, use_reserve);
> > if (err == -EEXIST)
> > goto retry;
> > goto out;
> > @@ -2798,7 +2826,7 @@ backout:
> > spin_unlock(&mm->page_table_lock);
> > backout_unlocked:
> > unlock_page(page);
> > - put_page(page);
> > + put_huge_page(page, use_reserve);
> > goto out;
> > }
> >
> > --
> > 1.7.9.5
> >
> > --
> > To unsubscribe, send a message with 'unsubscribe linux-mm' in
> > the body to majordomo@xxxxxxxxx. For more info on Linux MM,
> > see: http://www.linux-mm.org/ .
> > Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
> >
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@xxxxxxxxx. For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>