From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753156AbbCPEo4 (ORCPT ); Mon, 16 Mar 2015 00:44:56 -0400 Received: from cantor2.suse.de ([195.135.220.15]:56223 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753066AbbCPEox (ORCPT ); Mon, 16 Mar 2015 00:44:53 -0400 From: NeilBrown To: Al Viro Date: Mon, 16 Mar 2015 15:43:20 +1100 Subject: [PATCH 07/13] VFS/namei: abort RCU-walk on symlink if atime needs updating. Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org Message-ID: <20150316044320.23648.11587.stgit@notabene.brown> In-Reply-To: <20150316043602.23648.52734.stgit@notabene.brown> References: <20150316043602.23648.52734.stgit@notabene.brown> User-Agent: StGit/0.17.1-dirty MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org touch_atime is not RCU-safe, and so cannot be called on an RCU walk. However in situations where RCU-walk makes a difference, the symlink will likely to accessed much more often than it is useful to update the atime. So split out the test of "Does the atime actually need to be updated" into atime_needs_update(), and only allow RCU-walk on a symlink if that fails. Signed-off-by: NeilBrown --- fs/inode.c | 26 +++++++++++++++++++------- fs/namei.c | 7 ++++++- include/linux/fs.h | 1 + 3 files changed, 26 insertions(+), 8 deletions(-) diff --git a/fs/inode.c b/fs/inode.c index f00b16f45507..a0da920e4650 100644 --- a/fs/inode.c +++ b/fs/inode.c @@ -1584,30 +1584,41 @@ static int update_time(struct inode *inode, struct timespec *time, int flags) * This function automatically handles read only file systems and media, * as well as the "noatime" flag and inode specific "noatime" markers. */ -void touch_atime(const struct path *path) +int atime_needs_update(const struct path *path) { struct vfsmount *mnt = path->mnt; struct inode *inode = path->dentry->d_inode; struct timespec now; if (inode->i_flags & S_NOATIME) - return; + return 0; if (IS_NOATIME(inode)) - return; + return 0; if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode)) - return; + return 0; if (mnt->mnt_flags & MNT_NOATIME) - return; + return 0; if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode)) - return; + return 0; now = current_fs_time(inode->i_sb); if (!relatime_need_update(mnt, inode, now)) - return; + return 0; if (timespec_equal(&inode->i_atime, &now)) + return 0; + return 1; +} + +void touch_atime(const struct path *path) +{ + struct vfsmount *mnt = path->mnt; + struct inode *inode = path->dentry->d_inode; + struct timespec now; + + if (!atime_needs_update(path)) return; if (!sb_start_write_trylock(inode->i_sb)) @@ -1624,6 +1635,7 @@ void touch_atime(const struct path *path) * We may also fail on filesystems that have the ability to make parts * of the fs read only, e.g. subvolumes in Btrfs. */ + now = current_fs_time(inode->i_sb); update_time(inode, &now, S_ATIME); __mnt_drop_write(mnt); skip_update: diff --git a/fs/namei.c b/fs/namei.c index e0f889192f59..1663d21a3eb4 100644 --- a/fs/namei.c +++ b/fs/namei.c @@ -909,7 +909,12 @@ follow_link(struct path *link, struct nameidata *nd, void **p) cond_resched(); current->nameidata->total_link_count++; - touch_atime(link); + if (nd->flags & LOOKUP_RCU) { + error = -ECHILD; + if (atime_needs_update(link)) + goto out_put_nd_path; + } else + touch_atime(link); nd_set_link(NULL); error = security_inode_follow_link(link->dentry, nd->flags); diff --git a/include/linux/fs.h b/include/linux/fs.h index b7d578d552bf..41e6d99031dd 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -1844,6 +1844,7 @@ enum file_time_flags { S_VERSION = 8, }; +extern int atime_needs_update(const struct path *); extern void touch_atime(const struct path *); static inline void file_accessed(struct file *file) {