From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752989Ab1DUG7t (ORCPT ); Thu, 21 Apr 2011 02:59:49 -0400 Received: from mail-iy0-f174.google.com ([209.85.210.174]:55795 "EHLO mail-iy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751318Ab1DUG7s convert rfc822-to-8bit (ORCPT ); Thu, 21 Apr 2011 02:59:48 -0400 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type:content-transfer-encoding; b=aE7E/poulcFDfhwvtwJU3FAGhiZgwzGBehNHqitEMaGeW+J7uMlUde7RSVPLvHTG4X mvl8Ay6ZGvDAXWRKmRZmBIrvF5A1DSK24yFa1yM8piF5m+wQSYccXhQf0jnF4kiTJNK6 K9SV6ABkHGYTyx2Q56FsZz7LNBHJYq1pOIAh4= MIME-Version: 1.0 In-Reply-To: <20110415101248.GB22688@suse.de> References: <20110415101248.GB22688@suse.de> Date: Thu, 21 Apr 2011 15:59:47 +0900 Message-ID: Subject: Re: [PATCH] mm: Check if PTE is already allocated during page fault From: Minchan Kim To: Mel Gorman Cc: akpm@linux-foundation.org, Andrea Arcangeli , raz ben yehuda , riel@redhat.com, kosaki.motohiro@jp.fujitsu.com, lkml , linux-mm@kvack.org, stable@kernel.org Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Mel, On Fri, Apr 15, 2011 at 7:12 PM, Mel Gorman wrote: > With transparent hugepage support, handle_mm_fault() has to be careful > that a normal PMD has been established before handling a PTE fault. To > achieve this, it used __pte_alloc() directly instead of pte_alloc_map > as pte_alloc_map is unsafe to run against a huge PMD. pte_offset_map() > is called once it is known the PMD is safe. > > pte_alloc_map() is smart enough to check if a PTE is already present > before calling __pte_alloc but this check was lost. As a consequence, > PTEs may be allocated unnecessarily and the page table lock taken. > Thi useless PTE does get cleaned up but it's a performance hit which > is visible in page_test from aim9. > > This patch simply re-adds the check normally done by pte_alloc_map to > check if the PTE needs to be allocated before taking the page table > lock. The effect is noticable in page_test from aim9. > > AIM9 >                2.6.38-vanilla 2.6.38-checkptenone > creat-clo      446.10 ( 0.00%)   424.47 (-5.10%) > page_test       38.10 ( 0.00%)    42.04 ( 9.37%) > brk_test        52.45 ( 0.00%)    51.57 (-1.71%) > exec_test      382.00 ( 0.00%)   456.90 (16.39%) > fork_test       60.11 ( 0.00%)    67.79 (11.34%) > MMTests Statistics: duration > Total Elapsed Time (seconds)                611.90    612.22 > > (While this affects 2.6.38, it is a performance rather than a > functional bug and normally outside the rules -stable. While the big > performance differences are to a microbench, the difference in fork > and exec performance may be significant enough that -stable wants to > consider the patch) > > Reported-by: Raz Ben Yehuda > Signed-off-by: Mel Gorman > -- >  mm/memory.c |    2 +- >  1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/mm/memory.c b/mm/memory.c > index 5823698..1659574 100644 > --- a/mm/memory.c > +++ b/mm/memory.c > @@ -3322,7 +3322,7 @@ int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma, >         * run pte_offset_map on the pmd, if an huge pmd could >         * materialize from under us from a different thread. >         */ > -       if (unlikely(__pte_alloc(mm, vma, pmd, address))) > +       if (unlikely(pmd_none(*pmd)) && __pte_alloc(mm, vma, pmd, address)) >                return VM_FAULT_OOM; >        /* if an huge pmd materialized from under us just retry later */ >        if (unlikely(pmd_trans_huge(*pmd))) > Reviewed-by: Minchan Kim Sorry for jumping in too late. I have a just nitpick. We have another place, do_huge_pmd_anonymous_page. Although it isn't workload of page_test, is it valuable to expand your patch to cover it? If there is workload there are many thread and share one shared anon vma in ALWAYS THP mode, same problem would happen. -- Kind regards, Minchan Kim