From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753414Ab3KORsi (ORCPT ); Fri, 15 Nov 2013 12:48:38 -0500 Received: from mx1.redhat.com ([209.132.183.28]:8410 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751437Ab3KORsU (ORCPT ); Fri, 15 Nov 2013 12:48:20 -0500 From: Andrea Arcangeli To: Andrew Morton Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, Khalid Aziz , Pravin Shelar , Greg Kroah-Hartman , Ben Hutchings , Christoph Lameter , Johannes Weiner , Mel Gorman , Rik van Riel , Andi Kleen , Minchan Kim , Linus Torvalds Subject: [PATCH 0/3] mm: hugetlbfs: fix hugetlbfs optimization v2 Date: Fri, 15 Nov 2013 18:47:45 +0100 Message-Id: <1384537668-10283-1-git-send-email-aarcange@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi, 1/3 is a bugfix so it should be applied more urgently. 1/3 is not as fast as the current upstream code in the hugetlbfs + directio extreme 8GB/sec benchmark (but 3/3 should fill the gap later). The code is identical to the one I posted in v1 just rebased on upstream and was developed in collaboration with Khalid who already tested it. 2/3 and 3/3 had very little testing yet, and they're incremental optimization. 2/3 is minor and most certainly worth applying later. 3/3 instead complicates things a bit and adds more branches to the THP fast paths, so it should only be applied if the benchmarks of hugetlbfs + directio show that it is very worthwhile (that has not been verified yet). If it's not worthwhile 3/3 should be dropped (and the gap should be filled in some other way if the gap is not caused by the _mapcount mangling as I guessed). Ideally this should bring even more performance than current upstream code, as current upstream code still increased the _mapcount in gup_fast by mistake, while this eliminates the locked op on the tail page cacheline in gup_fast too (which is required for correctness too). As a side note: the _mapcount refcounting on tail pages is only needed for THP as it is a fundamental information required for split_huge_page_refcount to be able to distribute the head refcounts during the split. And it is done on _mapcount instead of the _count, because the _count would screwup badly with the get_page_unless_zero speculative pagecache accesses. Andrea Arcangeli (3): mm: hugetlbfs: fix hugetlbfs optimization mm: hugetlb: use get_page_foll in follow_hugetlb_page mm: tail page refcounting optimization for slab and hugetlbfs include/linux/mm.h | 30 ++++++++- mm/hugetlb.c | 19 +++++- mm/internal.h | 3 +- mm/swap.c | 187 ++++++++++++++++++++++++++++++++++------------------- 4 files changed, 170 insertions(+), 69 deletions(-)