Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S936997AbdDSQAi (ORCPT ); Wed, 19 Apr 2017 12:00:38 -0400 Received: from gum.cmpxchg.org ([85.214.110.215]:46206 "EHLO gum.cmpxchg.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1765628AbdDSQAe (ORCPT ); Wed, 19 Apr 2017 12:00:34 -0400 Date: Wed, 19 Apr 2017 12:00:29 -0400 From: Johannes Weiner To: "Huang, Ying" Cc: Andrew Morton , linux-mm@kvack.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH -mm -v9 3/3] mm, THP, swap: Enable THP swap optimization only if has compound map Message-ID: <20170419160029.GB3376@cmpxchg.org> References: <20170419070625.19776-1-ying.huang@intel.com> <20170419070625.19776-4-ying.huang@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170419070625.19776-4-ying.huang@intel.com> User-Agent: Mutt/1.8.0 (2017-02-23) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1625 Lines: 43 On Wed, Apr 19, 2017 at 03:06:25PM +0800, Huang, Ying wrote: > From: Huang Ying > > If there is no compound map for a THP (Transparent Huge Page), it is > possible that the map count of some sub-pages of the THP is 0. So it > is better to split the THP before swapping out. In this way, the > sub-pages not mapped will be freed, and we can avoid the unnecessary > swap out operations for these sub-pages. > > Cc: Johannes Weiner > Signed-off-by: "Huang, Ying" > --- > mm/swap_state.c | 12 +++++++++--- > 1 file changed, 9 insertions(+), 3 deletions(-) > > diff --git a/mm/swap_state.c b/mm/swap_state.c > index 3a3217f68937..b025c9878e5e 100644 > --- a/mm/swap_state.c > +++ b/mm/swap_state.c > @@ -192,9 +192,15 @@ int add_to_swap(struct page *page, struct list_head *list) > VM_BUG_ON_PAGE(!PageLocked(page), page); > VM_BUG_ON_PAGE(!PageUptodate(page), page); > > - /* cannot split, skip it */ > - if (unlikely(PageTransHuge(page)) && !can_split_huge_page(page, NULL)) > - return 0; > + if (unlikely(PageTransHuge(page))) { > + /* cannot split, skip it */ > + if (!can_split_huge_page(page, NULL)) > + return 0; > + /* fallback to split huge page firstly if no PMD map */ > + if (!compound_mapcount(page) && > + split_huge_page_to_list(page, list)) > + return 0; > + } This looks good to me, but could you please elaborate the comment a little bit with what you have in the changelog? Something like: /* * Split pages without a PMD map right away. Chances are * some or all of the tail pages can be freed without IO. */