From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755513AbcKPDMu (ORCPT ); Tue, 15 Nov 2016 22:12:50 -0500 Received: from mga14.intel.com ([192.55.52.115]:6463 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754813AbcKPDMm (ORCPT ); Tue, 15 Nov 2016 22:12:42 -0500 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.31,497,1473145200"; d="scan'208";a="31391326" From: "Huang, Ying" To: Andrew Morton Cc: tim.c.chen@intel.com, dave.hansen@intel.com, andi.kleen@intel.com, aaron.lu@intel.com, linux-mm@kvack.org, linux-kernel@vger.kernel.org, Huang Ying , Andrea Arcangeli , "Kirill A . Shutemov" , Ebru Akagunduz Subject: [PATCH -v5 8/9] mm, THP, swap: Support to split THP in swap cache Date: Wed, 16 Nov 2016 11:10:56 +0800 Message-Id: <20161116031057.12977-9-ying.huang@intel.com> X-Mailer: git-send-email 2.10.2 In-Reply-To: <20161116031057.12977-1-ying.huang@intel.com> References: <20161116031057.12977-1-ying.huang@intel.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org From: Huang Ying This patch enhanced the split_huge_page_to_list() to work properly for the THP (Transparent Huge Page) in the swap cache during swapping out. This is used for delaying splitting the THP during swapping out. Where for a THP to be swapped out, we will allocate a swap cluster, add the THP into the swap cache, then split the THP. The page lock will be held during this process. So in the code path other than swapping out, if the THP need to be split, the PageSwapCache(THP) will be always false. Cc: Andrea Arcangeli Cc: Kirill A. Shutemov Cc: Ebru Akagunduz Signed-off-by: "Huang, Ying" --- mm/huge_memory.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) diff --git a/mm/huge_memory.c b/mm/huge_memory.c index e894154..4e0d8b7 100644 --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -1835,7 +1835,7 @@ static void __split_huge_page_tail(struct page *head, int tail, * atomic_set() here would be safe on all archs (and not only on x86), * it's safer to use atomic_inc()/atomic_add(). */ - if (PageAnon(head)) { + if (PageAnon(head) && !PageSwapCache(head)) { page_ref_inc(page_tail); } else { /* Additional pin to radix tree */ @@ -1846,6 +1846,7 @@ static void __split_huge_page_tail(struct page *head, int tail, page_tail->flags |= (head->flags & ((1L << PG_referenced) | (1L << PG_swapbacked) | + (1L << PG_swapcache) | (1L << PG_mlocked) | (1L << PG_uptodate) | (1L << PG_active) | @@ -1908,7 +1909,11 @@ static void __split_huge_page(struct page *page, struct list_head *list, ClearPageCompound(head); /* See comment in __split_huge_page_tail() */ if (PageAnon(head)) { - page_ref_inc(head); + /* Additional pin to radix tree of swap cache */ + if (PageSwapCache(head)) + page_ref_add(head, 2); + else + page_ref_inc(head); } else { /* Additional pin to radix tree */ page_ref_add(head, 2); @@ -2020,10 +2025,12 @@ int page_trans_huge_mapcount(struct page *page, int *total_mapcount) /* Racy check whether the huge page can be split */ bool can_split_huge_page(struct page *page, int *pextra_pins) { - int extra_pins = 0; + int extra_pins; /* Additional pins from radix tree */ - if (!PageAnon(page)) + if (PageAnon(page)) + extra_pins = PageSwapCache(page) ? HPAGE_PMD_NR : 0; + else extra_pins = HPAGE_PMD_NR; if (pextra_pins) *pextra_pins = extra_pins; @@ -2078,7 +2085,6 @@ int split_huge_page_to_list(struct page *page, struct list_head *list) ret = -EBUSY; goto out; } - extra_pins = 0; mapping = NULL; anon_vma_lock_write(anon_vma); } else { -- 2.10.2