From: Theodore Ts'o Subject: [PATCH 3/4] ext3: Add replace-on-truncate hueristics for data=writeback mode Date: Fri, 3 Apr 2009 03:01:06 -0400 Message-ID: <1238742067-30814-4-git-send-email-tytso@mit.edu> References: <1238742067-30814-1-git-send-email-tytso@mit.edu> <1238742067-30814-2-git-send-email-tytso@mit.edu> <1238742067-30814-3-git-send-email-tytso@mit.edu> Cc: Linux Kernel Developers List , Ext4 Developers List , Theodore Ts'o To: torvalds@linuxfoundation.org Return-path: Received: from thunk.org ([69.25.196.29]:36438 "EHLO thunker.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751327AbZDCHBL (ORCPT ); Fri, 3 Apr 2009 03:01:11 -0400 In-Reply-To: <1238742067-30814-3-git-send-email-tytso@mit.edu> Sender: linux-ext4-owner@vger.kernel.org List-ID: In data=writeback mode, start an asynchronous flush when closing a file which had been previously truncated down to zero. This lowers the probability of data loss in the case of applications that attempt to replace a file using truncate. Signed-off-by: "Theodore Ts'o" --- fs/ext3/file.c | 4 ++++ fs/ext3/inode.c | 3 +++ include/linux/ext3_fs.h | 1 + 3 files changed, 8 insertions(+), 0 deletions(-) diff --git a/fs/ext3/file.c b/fs/ext3/file.c index 3be1e06..4a04cbb 100644 --- a/fs/ext3/file.c +++ b/fs/ext3/file.c @@ -33,6 +33,10 @@ */ static int ext3_release_file (struct inode * inode, struct file * filp) { + if (EXT3_I(inode)->i_state & EXT3_STATE_FLUSH_ON_CLOSE) { + filemap_flush(inode->i_mapping); + EXT3_I(inode)->i_state &= ~EXT3_STATE_FLUSH_ON_CLOSE; + } /* if we are the last writer on the inode, drop the block reservation */ if ((filp->f_mode & FMODE_WRITE) && (atomic_read(&inode->i_writecount) == 1)) diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c index 5fa453b..0f5bca0 100644 --- a/fs/ext3/inode.c +++ b/fs/ext3/inode.c @@ -2346,6 +2346,9 @@ void ext3_truncate(struct inode *inode) if (!ext3_can_truncate(inode)) return; + if (inode->i_size == 0 && ext3_should_writeback_data(inode)) + ei->i_state |= EXT3_STATE_FLUSH_ON_CLOSE; + /* * We have to lock the EOF page here, because lock_page() nests * outside journal_start(). diff --git a/include/linux/ext3_fs.h b/include/linux/ext3_fs.h index dd495b8..d2630c5 100644 --- a/include/linux/ext3_fs.h +++ b/include/linux/ext3_fs.h @@ -208,6 +208,7 @@ static inline __u32 ext3_mask_flags(umode_t mode, __u32 flags) #define EXT3_STATE_JDATA 0x00000001 /* journaled data exists */ #define EXT3_STATE_NEW 0x00000002 /* inode is newly created */ #define EXT3_STATE_XATTR 0x00000004 /* has in-inode xattrs */ +#define EXT3_STATE_FLUSH_ON_CLOSE 0x00000008 /* Used to pass group descriptor data when online resize is done */ struct ext3_new_group_input { -- 1.5.6.3