Re: [PATCH 1/8] migrate: make core migration code aware of hugepage

[Naoya Horiguchi]

On Fri, Jul 19, 2013 at 12:04:56PM +0800, Hillf Danton wrote:
...
> >> > +
> >> > +void putback_active_hugepage(struct page *page)
> >> > +{
> >> > +       VM_BUG_ON(!PageHead(page));
> >> > +       spin_lock(&hugetlb_lock);
> >> > +       list_move_tail(&page->lru, &(page_hstate(page))->hugepage_activelist);
> >> > +       spin_unlock(&hugetlb_lock);
> >> > +       put_page(page);
> >> > +}
> >> > +
> >> > +void putback_active_hugepages(struct list_head *l)
> >> > +{
> >> > +       struct page *page;
> >> > +       struct page *page2;
> >> > +
> >> > +       list_for_each_entry_safe(page, page2, l, lru)
> >> > +               putback_active_hugepage(page);
> >>
> >> Can we acquire hugetlb_lock only once?
> >
> > I'm not sure which is the best. In general, fine-grained locking is
> > preferred because other lock contenders wait less.
> > Could you tell some specific reason to hold lock outside the loop?
> >
> No anything special, looks we can do list splice after taking lock,
> then we no longer contend it.
> 
> >> > @@ -1025,7 +1029,11 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
> >> >                 list_for_each_entry_safe(page, page2, from, lru) {
> >> >                         cond_resched();
> >> >
> >> > -                       rc = unmap_and_move(get_new_page, private,
> >> > +                       if (PageHuge(page))
> >> > +                               rc = unmap_and_move_huge_page(get_new_page,
> >> > +                                               private, page, pass > 2, mode);
> >> > +                       else
> >> > +                               rc = unmap_and_move(get_new_page, private,
> >> >                                                 page, pass > 2, mode);
> >> >
> >> Is this hunk unclean merge?
> >
> > Sorry, I don't catch the point. This patch is based on v3.11-rc1 and
> > the present HEAD has no changes from that release.
> > Or do you mean that other trees have some conflicts? (my brief checking
> > on -mm/-next didn't find that...)
> >
> Looks this hunk should appear in 2/8 or later, as 1/8 is focusing
> on hugepage->lru?

I intended that 1/8 prepares common code used by all users of hugepage
migration. If I put this hunk on a patch which implements one of the
users, the other patchs implementing other users depends on it, which
looks to me an odd dependency...

Naoya
--
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/