[PATCH v3 107/110] VFS/namei: make the use of touch_atime() in get_link() RCU-safe.

[Al Viro]

From: NeilBrown <neilb@xxxxxxx>

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 have get_link() unlazy if it finds that
it will need to do that update.

Signed-off-by: NeilBrown <neilb@xxxxxxx>
Signed-off-by: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
---
 fs/inode.c         | 30 +++++++++++++++++++++---------
 fs/namei.c         | 12 +++++++++---
 include/linux/fs.h |  1 +
 3 files changed, 31 insertions(+), 12 deletions(-)

diff --git a/fs/inode.c b/fs/inode.c
index 952fb48..e8d6268 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -1585,36 +1585,47 @@ 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)
+bool atime_needs_update(const struct path *path, struct inode *inode)
 {
 	struct vfsmount *mnt = path->mnt;
-	struct inode *inode = d_inode(path->dentry);
 	struct timespec now;
 
 	if (inode->i_flags & S_NOATIME)
-		return;
+		return false;
 	if (IS_NOATIME(inode))
-		return;
+		return false;
 	if ((inode->i_sb->s_flags & MS_NODIRATIME) && S_ISDIR(inode->i_mode))
-		return;
+		return false;
 
 	if (mnt->mnt_flags & MNT_NOATIME)
-		return;
+		return false;
 	if ((mnt->mnt_flags & MNT_NODIRATIME) && S_ISDIR(inode->i_mode))
-		return;
+		return false;
 
 	now = current_fs_time(inode->i_sb);
 
 	if (!relatime_need_update(mnt, inode, now))
-		return;
+		return false;
 
 	if (timespec_equal(&inode->i_atime, &now))
+		return false;
+
+	return true;
+}
+
+void touch_atime(const struct path *path)
+{
+	struct vfsmount *mnt = path->mnt;
+	struct inode *inode = d_inode(path->dentry);
+	struct timespec now;
+
+	if (!atime_needs_update(path, inode))
 		return;
 
 	if (!sb_start_write_trylock(inode->i_sb))
 		return;
 
-	if (__mnt_want_write(mnt))
+	if (__mnt_want_write(mnt) != 0)
 		goto skip_update;
 	/*
 	 * File systems can error out when updating inodes if they need to
@@ -1625,6 +1636,7 @@ 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 4d6065d..9ef8c50 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -958,13 +958,19 @@ const char *get_link(struct nameidata *nd)
 	int error;
 	const char *res;
 
-	if (nd->flags & LOOKUP_RCU) {
+	if (!(nd->flags & LOOKUP_RCU)) {
+		touch_atime(&last->link);
+	} else if (atime_needs_update(&last->link, inode)) {
 		if (unlikely(unlazy_walk(nd, NULL, 0)))
 			return ERR_PTR(-ECHILD);
+		touch_atime(&last->link);
 	}
-	cond_resched();
 
-	touch_atime(&last->link);
+	_cond_resched();
+	if (nd->flags & LOOKUP_RCU) {
+		if (unlikely(unlazy_walk(nd, NULL, 0)))
+			return ERR_PTR(-ECHILD);
+	}
 
 	error = security_inode_follow_link(dentry, inode,
 					   nd->flags & LOOKUP_RCU);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 8f73851..1426c43 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -1880,6 +1880,7 @@ enum file_time_flags {
 	S_VERSION = 8,
 };
 
+extern bool atime_needs_update(const struct path *, struct inode *);
 extern void touch_atime(const struct path *);
 static inline void file_accessed(struct file *file)
 {
-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/