Return-Path: Received: from mail-pl1-f194.google.com ([209.85.214.194]:38599 "EHLO mail-pl1-f194.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725937AbeKUNQf (ORCPT ); Wed, 21 Nov 2018 08:16:35 -0500 Received: by mail-pl1-f194.google.com with SMTP id e5so3201564plb.5 for ; Tue, 20 Nov 2018 18:44:13 -0800 (PST) From: Eiichi Tsukata To: andi@firstfloor.org, Chris Mason , Josef Bacik , David Sterba , "Theodore Ts'o" , Andreas Dilger , Jaegeuk Kim , Chao Yu , Miklos Szeredi , Bob Peterson , Andreas Gruenbacher , Alexander Viro , linux-btrfs@vger.kernel.org, linux-ext4@vger.kernel.org, linux-f2fs-devel@lists.sourceforge.net, linux-fsdevel@vger.kernel.org, cluster-devel@redhat.com, linux-unionfs@vger.kernel.org Cc: Eiichi Tsukata Subject: [PATCH v1 0/4] fs: fix race between llseek SEEK_END and write Date: Wed, 21 Nov 2018 11:43:56 +0900 Message-Id: <20181121024400.4346-1-devel@etsukata.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Sender: linux-ext4-owner@vger.kernel.org List-ID: Some file systems (including ext4, xfs, ramfs ...) have the following problem as I've described in the commit message of the 1/4 patch. The commit ef3d0fd27e90 ("vfs: do (nearly) lockless generic_file_llseek") removed almost all locks in llseek() including SEEK_END. It based on the idea that write() updates size atomically. But in fact, write() can be divided into two or more parts in generic_perform_write() when pos straddles over the PAGE_SIZE, which results in updating size multiple times in one write(). It means that llseek() can see the size being updated during write(). This race changes behavior of some applications. 'tail' is one of those applications. It reads range [pos, pos_end] where pos_end is obtained via llseek() SEEK_END. Sometimes, a read line could be broken. reproducer: $ while true; do echo 123456 >> out; done $ while true; do tail out | grep -v 123456 ; done example output(take 30 secs): 12345 1 1234 1 12 1234 Note: Some file systems which indivisually implements llseek() and hold inode mutex lock in it are not affected. ex:) btrfs, ocfs2 Patch 1: re-implements inode lock for SEEK_END in generic_file_llseek() Patch 2 to 4: implement inode lock for SEEK_END in each file systems Eiichi Tsukata (4): vfs: fix race between llseek SEEK_END and write ext4: fix race between llseek SEEK_END and write f2fs: fix race between llseek SEEK_END and write overlayfs: fix race between llseek SEEK_END and write fs/btrfs/file.c | 2 +- fs/ext4/file.c | 10 ++++++++++ fs/f2fs/file.c | 6 +----- fs/fuse/file.c | 5 +++-- fs/gfs2/file.c | 3 ++- fs/overlayfs/file.c | 23 ++++++++++++++++++++--- fs/read_write.c | 37 ++++++++++++++++++++++++++++++++++--- include/linux/fs.h | 2 ++ 8 files changed, 73 insertions(+), 15 deletions(-) -- 2.19.1