From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jaegeuk Kim Subject: Re: [PATCH 2/2] f2fs: remove unnecessary wait in write_begin Date: Fri, 17 Feb 2017 10:21:14 -0800 Message-ID: <20170217182114.GA2951@jaegeuk.local> References: <20170217091639.29433-1-heyunlei@huawei.com> <20170217091639.29433-2-heyunlei@huawei.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from sog-mx-2.v43.ch3.sourceforge.com ([172.29.43.192] helo=mx.sourceforge.net) by sfs-ml-3.v29.ch3.sourceforge.com with esmtp (Exim 4.76) (envelope-from ) id 1cen9o-0006fw-TF for linux-f2fs-devel@lists.sourceforge.net; Fri, 17 Feb 2017 18:21:24 +0000 Received: from mail.kernel.org ([198.145.29.136]) by sog-mx-2.v43.ch3.sourceforge.com with esmtps (TLSv1:AES256-SHA:256) (Exim 4.76) id 1cen9m-00082T-CN for linux-f2fs-devel@lists.sourceforge.net; Fri, 17 Feb 2017 18:21:24 +0000 Content-Disposition: inline In-Reply-To: <20170217091639.29433-2-heyunlei@huawei.com> List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: linux-f2fs-devel-bounces@lists.sourceforge.net To: Yunlei He Cc: linux-f2fs-devel@lists.sourceforge.net On 02/17, Yunlei He wrote: > It has already waited page stable in grab_cache_page_write_begin, > so we remove this unnecessary wait. I also got a deadlock in my perf test, and I wrote another patch to resolve it by replacing grab_cache_page_write_begin() with pagecache_get_page(). >>From 8baf724c638813ab8d2e9bad99f5b8683e788da3 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Fri, 17 Feb 2017 09:55:55 -0800 Subject: [PATCH] f2fs: do not wait for writeback in write_begin Otherwise we can get livelock like below. [79880.428136] dbench D 0 18405 18404 0x00000000 [79880.428139] Call Trace: [79880.428142] __schedule+0x219/0x6b0 [79880.428144] schedule+0x36/0x80 [79880.428147] schedule_timeout+0x243/0x2e0 [79880.428152] ? update_sd_lb_stats+0x16b/0x5f0 [79880.428155] ? ktime_get+0x3c/0xb0 [79880.428157] io_schedule_timeout+0xa6/0x110 [79880.428161] __lock_page+0xf7/0x130 [79880.428164] ? unlock_page+0x30/0x30 [79880.428167] pagecache_get_page+0x16b/0x250 [79880.428171] grab_cache_page_write_begin+0x20/0x40 [79880.428182] f2fs_write_begin+0xa2/0xdb0 [f2fs] [79880.428192] ? f2fs_mark_inode_dirty_sync+0x16/0x30 [f2fs] [79880.428197] ? kmem_cache_free+0x79/0x200 [79880.428203] ? __mark_inode_dirty+0x17f/0x360 [79880.428206] generic_perform_write+0xbb/0x190 [79880.428213] ? file_update_time+0xa4/0xf0 [79880.428217] __generic_file_write_iter+0x19b/0x1e0 [79880.428226] f2fs_file_write_iter+0x9c/0x180 [f2fs] [79880.428231] __vfs_write+0xc5/0x140 [79880.428235] vfs_write+0xb2/0x1b0 [79880.428238] SyS_write+0x46/0xa0 [79880.428242] entry_SYSCALL_64_fastpath+0x1e/0xad Fixes: cae96a5c8ab6 ("f2fs: check io submission more precisely") Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 8c61fa7fd27d..5f3bc98a387d 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -1759,7 +1759,12 @@ static int f2fs_write_begin(struct file *file, struct address_space *mapping, goto fail; } repeat: - page = grab_cache_page_write_begin(mapping, index, flags); + /* + * Do not use grab_cache_page_write_begin() to avoid deadlock due to + * wait_for_stable_page. Will wait that below with our IO control. + */ + page = pagecache_get_page(mapping, index, + FGP_LOCK | FGP_WRITE | FGP_CREAT, GFP_NOFS); if (!page) { err = -ENOMEM; goto fail; -- 2.11.0 ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, SlashDot.org! http://sdm.link/slashdot