From patchwork Tue Jan 22 10:49:00 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Miao Xie X-Patchwork-Id: 2017681 Return-Path: X-Original-To: patchwork-linux-btrfs@patchwork.kernel.org Delivered-To: patchwork-process-083081@patchwork2.kernel.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by patchwork2.kernel.org (Postfix) with ESMTP id 97DBBDF2EB for ; Tue, 22 Jan 2013 10:48:29 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752687Ab3AVKsZ (ORCPT ); Tue, 22 Jan 2013 05:48:25 -0500 Received: from cn.fujitsu.com ([222.73.24.84]:8976 "EHLO song.cn.fujitsu.com" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1750915Ab3AVKsX (ORCPT ); Tue, 22 Jan 2013 05:48:23 -0500 X-IronPort-AV: E=Sophos;i="4.84,514,1355068800"; d="scan'208";a="6628309" Received: from unknown (HELO tang.cn.fujitsu.com) ([10.167.250.3]) by song.cn.fujitsu.com with ESMTP; 22 Jan 2013 18:46:14 +0800 Received: from fnstmail02.fnst.cn.fujitsu.com (tang.cn.fujitsu.com [127.0.0.1]) by tang.cn.fujitsu.com (8.14.3/8.13.1) with ESMTP id r0MAmK8b019313; Tue, 22 Jan 2013 18:48:21 +0800 Received: from [10.167.225.199] ([10.167.225.199]) by fnstmail02.fnst.cn.fujitsu.com (Lotus Domino Release 8.5.3) with ESMTP id 2013012218472706-1090500 ; Tue, 22 Jan 2013 18:47:27 +0800 Message-ID: <50FE6E9C.2040803@cn.fujitsu.com> Date: Tue, 22 Jan 2013 18:49:00 +0800 From: Miao Xie Reply-To: miaox@cn.fujitsu.com User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:17.0) Gecko/20130110 Thunderbird/17.0.2 MIME-Version: 1.0 To: Linux Btrfs CC: Alex Lyakas Subject: [PATCH 1/5] Btrfs: fix repeated delalloc work allocation X-MIMETrack: Itemize by SMTP Server on mailserver/fnst(Release 8.5.3|September 15, 2011) at 2013/01/22 18:47:27, Serialize by Router on mailserver/fnst(Release 8.5.3|September 15, 2011) at 2013/01/22 18:47:27, Serialize complete at 2013/01/22 18:47:27 Sender: linux-btrfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-btrfs@vger.kernel.org btrfs_start_delalloc_inodes() locks the delalloc_inodes list, fetches the first inode, unlocks the list, triggers btrfs_alloc_delalloc_work/ btrfs_queue_worker for this inode, and then it locks the list, checks the head of the list again. But because we don't delete the first inode that it deals with before, it will fetch the same inode. As a result, this function allocates a huge amount of btrfs_delalloc_work structures, and OOM happens. Fix this problem by splice this delalloc list. Reported-by: Alex Lyakas Signed-off-by: Miao Xie --- fs/btrfs/inode.c | 55 ++++++++++++++++++++++++++++++++++++++++------------- 1 files changed, 41 insertions(+), 14 deletions(-) diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index 67ed24a..86f1d25 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -7545,41 +7545,61 @@ void btrfs_wait_and_free_delalloc_work(struct btrfs_delalloc_work *work) */ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput) { - struct list_head *head = &root->fs_info->delalloc_inodes; struct btrfs_inode *binode; struct inode *inode; struct btrfs_delalloc_work *work, *next; struct list_head works; + struct list_head splice; int ret = 0; if (root->fs_info->sb->s_flags & MS_RDONLY) return -EROFS; INIT_LIST_HEAD(&works); - + INIT_LIST_HEAD(&splice); +again: spin_lock(&root->fs_info->delalloc_lock); - while (!list_empty(head)) { - binode = list_entry(head->next, struct btrfs_inode, + list_splice_init(&root->fs_info->delalloc_inodes, &splice); + while (!list_empty(&splice)) { + binode = list_entry(splice.next, struct btrfs_inode, delalloc_inodes); + + list_del_init(&binode->delalloc_inodes); + inode = igrab(&binode->vfs_inode); if (!inode) - list_del_init(&binode->delalloc_inodes); + continue; + + list_add_tail(&binode->delalloc_inodes, + &root->fs_info->delalloc_inodes); spin_unlock(&root->fs_info->delalloc_lock); - if (inode) { - work = btrfs_alloc_delalloc_work(inode, 0, delay_iput); - if (!work) { - ret = -ENOMEM; - goto out; - } - list_add_tail(&work->list, &works); - btrfs_queue_worker(&root->fs_info->flush_workers, - &work->work); + + work = btrfs_alloc_delalloc_work(inode, 0, delay_iput); + if (unlikely(!work)) { + ret = -ENOMEM; + goto out; } + list_add_tail(&work->list, &works); + btrfs_queue_worker(&root->fs_info->flush_workers, + &work->work); + cond_resched(); spin_lock(&root->fs_info->delalloc_lock); } spin_unlock(&root->fs_info->delalloc_lock); + list_for_each_entry_safe(work, next, &works, list) { + list_del_init(&work->list); + btrfs_wait_and_free_delalloc_work(work); + } + + spin_lock(&root->fs_info->delalloc_lock); + if (!list_empty(&root->fs_info->delalloc_inodes)) { + spin_unlock(&root->fs_info->delalloc_lock); + goto again; + } + spin_unlock(&root->fs_info->delalloc_lock); + /* the filemap_flush will queue IO into the worker threads, but * we have to make sure the IO is actually started and that * ordered extents get created before we return @@ -7592,11 +7612,18 @@ int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput) atomic_read(&root->fs_info->async_delalloc_pages) == 0)); } atomic_dec(&root->fs_info->async_submit_draining); + return 0; out: list_for_each_entry_safe(work, next, &works, list) { list_del_init(&work->list); btrfs_wait_and_free_delalloc_work(work); } + + if (!list_empty_careful(&splice)) { + spin_lock(&root->fs_info->delalloc_lock); + list_splice_tail(&splice, &root->fs_info->delalloc_inodes); + spin_unlock(&root->fs_info->delalloc_lock); + } return ret; }