From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from psmtp.com (na3sys010amx128.postini.com [74.125.245.128]) by kanga.kvack.org (Postfix) with SMTP id D5ECF6B00D2 for ; Tue, 7 May 2013 17:19:56 -0400 (EDT) Subject: [RFC][PATCH 1/7] defer clearing of page_private() for swap cache pages From: Dave Hansen Date: Tue, 07 May 2013 14:19:55 -0700 References: <20130507211954.9815F9D1@viggo.jf.intel.com> In-Reply-To: <20130507211954.9815F9D1@viggo.jf.intel.com> Message-Id: <20130507211955.7DF88A4F@viggo.jf.intel.com> Sender: owner-linux-mm@kvack.org List-ID: 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 There are only two callers of swapcache_free() which actually pass in a non-NULL 'struct page'. Both of them (__remove_mapping and delete_from_swap_cache()) create a temporary on-stack 'swp_entry_t' and set entry.val to page_private(). They need to do this since __delete_from_swap_cache() does set_page_private(page, 0) and destroys the information. However, I'd like to batch a few of these operations on several pages together in a new version of __remove_mapping(), and I would prefer not to have to allocate temporary storage for each page. The code is pretty ugly, and it's a bit silly to create these on-stack 'swp_entry_t's when it is so easy to just keep the information around in 'struct page'. There should not be any danger in doing this since we are absolutely on the path of freeing these page. There is no turning back, and no other rerferences can be obtained after it comes out of the radix tree. Signed-off-by: Dave Hansen --- linux.git-davehans/mm/swap_state.c | 4 ++-- linux.git-davehans/mm/vmscan.c | 2 ++ 2 files changed, 4 insertions(+), 2 deletions(-) diff -puN mm/swap_state.c~__delete_from_swap_cache-dont-clear-page-private mm/swap_state.c --- linux.git/mm/swap_state.c~__delete_from_swap_cache-dont-clear-page-private 2013-05-07 13:48:13.698044473 -0700 +++ linux.git-davehans/mm/swap_state.c 2013-05-07 13:48:13.703044693 -0700 @@ -146,8 +146,6 @@ void __delete_from_swap_cache(struct pag entry.val = page_private(page); address_space = swap_address_space(entry); radix_tree_delete(&address_space->page_tree, page_private(page)); - set_page_private(page, 0); - ClearPageSwapCache(page); address_space->nrpages--; __dec_zone_page_state(page, NR_FILE_PAGES); INC_CACHE_INFO(del_total); @@ -224,6 +222,8 @@ void delete_from_swap_cache(struct page spin_unlock_irq(&address_space->tree_lock); swapcache_free(entry, page); + set_page_private(page, 0); + ClearPageSwapCache(page); page_cache_release(page); } diff -puN mm/vmscan.c~__delete_from_swap_cache-dont-clear-page-private mm/vmscan.c --- linux.git/mm/vmscan.c~__delete_from_swap_cache-dont-clear-page-private 2013-05-07 13:48:13.700044561 -0700 +++ linux.git-davehans/mm/vmscan.c 2013-05-07 13:48:13.705044783 -0700 @@ -494,6 +494,8 @@ static int __remove_mapping(struct addre __delete_from_swap_cache(page); spin_unlock_irq(&mapping->tree_lock); swapcache_free(swap, page); + set_page_private(page, 0); + ClearPageSwapCache(page); } else { void (*freepage)(struct page *); _ -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org