[07/13] VFS/namei: abort RCU-walk on symlink if atime needs updating.

Commit Message

NeilBrown March 16, 2015, 4:43 a.m. UTC

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 <[e-mail]>
---
 fs/inode.c         |   26 +++++++++++++++++++-------
 fs/namei.c         |    7 ++++++-
 include/linux/fs.h |    1 +
 3 files changed, 26 insertions(+), 8 deletions(-)



--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [e-mail]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

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)
 {