From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932538AbbD0Lu0 (ORCPT ); Mon, 27 Apr 2015 07:50:26 -0400 Received: from cantor2.suse.de ([195.135.220.15]:50887 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752619AbbD0LuZ (ORCPT ); Mon, 27 Apr 2015 07:50:25 -0400 Message-ID: <553E2281.4050102@suse.cz> Date: Mon, 27 Apr 2015 13:50:25 +0200 From: Vlastimil Babka User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:31.0) Gecko/20100101 Thunderbird/31.6.0 MIME-Version: 1.0 To: "Kirill A. Shutemov" , Andrew Morton CC: linux-mm@kvack.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH] compaction: fix isolate_migratepages_block() for THP=n References: <1430134006-215317-1-git-send-email-kirill.shutemov@linux.intel.com> In-Reply-To: <1430134006-215317-1-git-send-email-kirill.shutemov@linux.intel.com> Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 27.4.2015 13:26, Kirill A. Shutemov wrote: > PageTrans* helpers are always-false if THP is disabled compile-time. > It means the fucntion will fail to detect hugetlb pages in this case. > > Let's use PageCompound() instead. With small tweak to how we calculate > next low_pfn it will make function ready to see tail pages. > > Signed-off-by: Kirill A. Shutemov Acked-by: Vlastimil Babka > --- > mm/compaction.c | 25 +++++++++++++------------ > 1 file changed, 13 insertions(+), 12 deletions(-) > > diff --git a/mm/compaction.c b/mm/compaction.c > index 018f08da99a2..6ef2fdf1d6b6 100644 > --- a/mm/compaction.c > +++ b/mm/compaction.c > @@ -732,18 +732,18 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn, > * splitting and collapsing (collapsing has already happened > * if PageLRU is set) but the lock is not necessarily taken > * here and it is wasteful to take it just to check transhuge. > - * Check TransHuge without lock and skip the whole pageblock if > - * it's either a transhuge or hugetlbfs page, as calling > + * Check PageCompound without lock and skip the whole pageblock > + * if it's either a transhuge or hugetlbfs page, as calling > * compound_order() without preventing THP from splitting the > * page underneath us may return surprising results. > */ > - if (PageTransHuge(page)) { > - if (!locked) > - low_pfn = ALIGN(low_pfn + 1, > - pageblock_nr_pages) - 1; > + if (PageCompound(page)) { > + int nr; > + if (locked) > + nr = 1 << compound_order(page); > else > - low_pfn += (1 << compound_order(page)) - 1; > - > + nr = pageblock_nr_pages; > + low_pfn = ALIGN(low_pfn + 1, nr) - 1; > continue; > } > > @@ -763,11 +763,12 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn, > if (!locked) > break; > > - /* Recheck PageLRU and PageTransHuge under lock */ > + /* Recheck PageLRU and PageCompound under lock */ > if (!PageLRU(page)) > continue; > - if (PageTransHuge(page)) { > - low_pfn += (1 << compound_order(page)) - 1; > + if (PageCompound(page)) { > + int nr = 1 << compound_order(page); > + low_pfn = ALIGN(low_pfn + 1, nr) - 1; > continue; > } > } > @@ -778,7 +779,7 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn, > if (__isolate_lru_page(page, isolate_mode) != 0) > continue; > > - VM_BUG_ON_PAGE(PageTransCompound(page), page); > + VM_BUG_ON_PAGE(PageCompound(page), page); > > /* Successfully isolated */ > del_page_from_lru_list(page, lruvec, page_lru(page)); >