Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753036Ab3EPUfA (ORCPT ); Thu, 16 May 2013 16:35:00 -0400 Received: from mga14.intel.com ([143.182.124.37]:22238 "EHLO mga14.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752270Ab3EPUed (ORCPT ); Thu, 16 May 2013 16:34:33 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.87,686,1363158000"; d="scan'208";a="303598578" Subject: [RFCv2][PATCH 3/5] break up __remove_mapping() To: linux-mm@kvack.org Cc: linux-kernel@vger.kernel.org, akpm@linux-foundation.org, mgorman@suse.de, tim.c.chen@linux.intel.com, Dave Hansen From: Dave Hansen Date: Thu, 16 May 2013 13:34:31 -0700 References: <20130516203427.E3386936@viggo.jf.intel.com> In-Reply-To: <20130516203427.E3386936@viggo.jf.intel.com> Message-Id: <20130516203431.E7D7BAE7@viggo.jf.intel.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 3789 Lines: 128 From: Dave Hansen Our goal here is to eventually reduce the number of repetitive acquire/release operations on mapping->tree_lock. Logically, this patch has two steps: 1. rename __remove_mapping() to lock_remove_mapping() since "__" usually means "this us the unlocked version. 2. Recreate __remove_mapping() to _be_ the lock_remove_mapping() but without the locks. I think this actually makes the code flow around the locking _much_ more straighforward since the locking just becomes: spin_lock_irq(&mapping->tree_lock); ret = __remove_mapping(mapping, page); spin_unlock_irq(&mapping->tree_lock); One non-obvious part of this patch: the freepage = mapping->a_ops->freepage; used to happen under the mapping->tree_lock, but this patch moves it to outside of the lock. All of the other a_ops->freepage users do it outside the lock, and we only assign it when we create inodes, so that makes it safe. Signed-off-by: Dave Hansen Acked-by: Mel Gorman --- linux.git-davehans/mm/vmscan.c | 43 ++++++++++++++++++++++++----------------- 1 file changed, 26 insertions(+), 17 deletions(-) diff -puN mm/vmscan.c~make-remove-mapping-without-locks mm/vmscan.c --- linux.git/mm/vmscan.c~make-remove-mapping-without-locks 2013-05-16 13:27:25.268163133 -0700 +++ linux.git-davehans/mm/vmscan.c 2013-05-16 13:27:25.273163355 -0700 @@ -450,12 +450,12 @@ static pageout_t pageout(struct page *pa * Same as remove_mapping, but if the page is removed from the mapping, it * gets returned with a refcount of 0. */ -static int __remove_mapping(struct address_space *mapping, struct page *page) +static int __remove_mapping(struct address_space *mapping, + struct page *page) { BUG_ON(!PageLocked(page)); BUG_ON(mapping != page_mapping(page)); - spin_lock_irq(&mapping->tree_lock); /* * The non racy check for a busy page. * @@ -482,35 +482,44 @@ static int __remove_mapping(struct addre * and thus under tree_lock, then this ordering is not required. */ if (!page_freeze_refs(page, 2)) - goto cannot_free; + return 0; /* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */ if (unlikely(PageDirty(page))) { page_unfreeze_refs(page, 2); - goto cannot_free; + return 0; } if (PageSwapCache(page)) { __delete_from_swap_cache(page); - spin_unlock_irq(&mapping->tree_lock); + } else { + __delete_from_page_cache(page); + } + return 1; +} + +static int lock_remove_mapping(struct address_space *mapping, struct page *page) +{ + int ret; + BUG_ON(!PageLocked(page)); + + spin_lock_irq(&mapping->tree_lock); + ret = __remove_mapping(mapping, page); + spin_unlock_irq(&mapping->tree_lock); + + /* unable to free */ + if (!ret) + return 0; + + if (PageSwapCache(page)) { swapcache_free_page_entry(page); } else { void (*freepage)(struct page *); - freepage = mapping->a_ops->freepage; - - __delete_from_page_cache(page); - spin_unlock_irq(&mapping->tree_lock); mem_cgroup_uncharge_cache_page(page); - if (freepage != NULL) freepage(page); } - - return 1; - -cannot_free: - spin_unlock_irq(&mapping->tree_lock); - return 0; + return ret; } /* @@ -521,7 +530,7 @@ cannot_free: */ int remove_mapping(struct address_space *mapping, struct page *page) { - if (__remove_mapping(mapping, page)) { + if (lock_remove_mapping(mapping, page)) { /* * Unfreezing the refcount with 1 rather than 2 effectively * drops the pagecache ref for us without requiring another _ -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/