autofs4: allow RCU-walk to walk through autofs4

This series teaches autofs about RCU-walk so that we don't drop straight
into REF-walk when we hit an autofs directory, and so that we avoid
spinlocks as much as possible when performing an RCU-walk.

This is needed so that the benefits of the recent NFS support for
RCU-walk are fully available when NFS filesystems are automounted.

Patches have been carefully reviewed and tested both with test suites
and in production - thanks a lot to Ian Kent for his support there.

This patch (of 6):

Any attempt to look up a pathname that passes though an autofs4 mount is
currently forced out of RCU-walk into REF-walk.

This can significantly hurt performance of many-thread work loads on
many-core systems, especially if the automounted filesystem supports
RCU-walk but doesn't get to benefit from it.

So if autofs4_d_manage is called with rcu_walk set, only fail with -ECHILD
if it is necessary to wait longer than a spinlock.
Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Ian Kent <raven@themaw.net>
Tested-by: Ian Kent <raven@themaw.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/autofs4/autofs_i.h

@@ -148,7 +148,7 @@ void autofs4_free_ino(struct autofs_info *);
 
 /* Expiration */
 int is_autofs4_dentry(struct dentry *);
-int autofs4_expire_wait(struct dentry *dentry);
+int autofs4_expire_wait(struct dentry *dentry, int rcu_walk);
 int autofs4_expire_run(struct super_block *, struct vfsmount *,
 			struct autofs_sb_info *,
 			struct autofs_packet_expire __user *);

fs/autofs4/dev-ioctl.c

@@ -450,7 +450,7 @@ static int autofs_dev_ioctl_requester(struct file *fp,
 	ino = autofs4_dentry_ino(path.dentry);
 	if (ino) {
 		err = 0;
-		autofs4_expire_wait(path.dentry);
+		autofs4_expire_wait(path.dentry, 0);
 		spin_lock(&sbi->fs_lock);
 		param->requester.uid = from_kuid_munged(current_user_ns(), ino->uid);
 		param->requester.gid = from_kgid_munged(current_user_ns(), ino->gid);

fs/autofs4/expire.c

@@ -461,7 +461,7 @@ struct dentry *autofs4_expire_indirect(struct super_block *sb,
 	return expired;
 }
 
-int autofs4_expire_wait(struct dentry *dentry)
+int autofs4_expire_wait(struct dentry *dentry, int rcu_walk)
 {
 	struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
 	struct autofs_info *ino = autofs4_dentry_ino(dentry);
@@ -471,6 +471,8 @@ int autofs4_expire_wait(struct dentry *dentry)
 	spin_lock(&sbi->fs_lock);
 	if (ino->flags & AUTOFS_INF_EXPIRING) {
 		spin_unlock(&sbi->fs_lock);
+		if (rcu_walk)
+			return -ECHILD;
 
 		DPRINTK("waiting for expire %p name=%.*s",
 			 dentry, dentry->d_name.len, dentry->d_name.name);

fs/autofs4/root.c

@@ -210,7 +210,8 @@ static struct dentry *autofs4_lookup_active(struct dentry *dentry)
 	return NULL;
 }
 
-static struct dentry *autofs4_lookup_expiring(struct dentry *dentry)
+static struct dentry *autofs4_lookup_expiring(struct dentry *dentry,
+					      bool rcu_walk)
 {
 	struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
 	struct dentry *parent = dentry->d_parent;
@@ -229,6 +230,11 @@ static struct dentry *autofs4_lookup_expiring(struct dentry *dentry)
 		struct dentry *expiring;
 		struct qstr *qstr;
 
+		if (rcu_walk) {
+			spin_unlock(&sbi->lookup_lock);
+			return ERR_PTR(-ECHILD);
+		}
+
 		ino = list_entry(p, struct autofs_info, expiring);
 		expiring = ino->dentry;
 
@@ -264,13 +270,15 @@ static struct dentry *autofs4_lookup_expiring(struct dentry *dentry)
 	return NULL;
 }
 
-static int autofs4_mount_wait(struct dentry *dentry)
+static int autofs4_mount_wait(struct dentry *dentry, bool rcu_walk)
 {
 	struct autofs_sb_info *sbi = autofs4_sbi(dentry->d_sb);
 	struct autofs_info *ino = autofs4_dentry_ino(dentry);
 	int status = 0;
 
 	if (ino->flags & AUTOFS_INF_PENDING) {
+		if (rcu_walk)
+			return -ECHILD;
 		DPRINTK("waiting for mount name=%.*s",
 			dentry->d_name.len, dentry->d_name.name);
 		status = autofs4_wait(sbi, dentry, NFY_MOUNT);
@@ -280,20 +288,22 @@ static int autofs4_mount_wait(struct dentry *dentry)
 	return status;
 }
 
-static int do_expire_wait(struct dentry *dentry)
+static int do_expire_wait(struct dentry *dentry, bool rcu_walk)
 {
 	struct dentry *expiring;
 
-	expiring = autofs4_lookup_expiring(dentry);
+	expiring = autofs4_lookup_expiring(dentry, rcu_walk);
+	if (IS_ERR(expiring))
+		return PTR_ERR(expiring);
 	if (!expiring)
-		return autofs4_expire_wait(dentry);
+		return autofs4_expire_wait(dentry, rcu_walk);
 	else {
 		/*
 		 * If we are racing with expire the request might not
 		 * be quite complete, but the directory has been removed
 		 * so it must have been successful, just wait for it.
 		 */
-		autofs4_expire_wait(expiring);
+		autofs4_expire_wait(expiring, 0);
 		autofs4_del_expiring(expiring);
 		dput(expiring);
 	}
@@ -345,7 +355,7 @@ static struct vfsmount *autofs4_d_automount(struct path *path)
 	 * and the directory was removed, so just go ahead and try
 	 * the mount.
 	 */
-	status = do_expire_wait(dentry);
+	status = do_expire_wait(dentry, 0);
 	if (status && status != -EAGAIN)
 		return NULL;
 
@@ -353,7 +363,7 @@ static struct vfsmount *autofs4_d_automount(struct path *path)
 	spin_lock(&sbi->fs_lock);
 	if (ino->flags & AUTOFS_INF_PENDING) {
 		spin_unlock(&sbi->fs_lock);
-		status = autofs4_mount_wait(dentry);
+		status = autofs4_mount_wait(dentry, 0);
 		if (status)
 			return ERR_PTR(status);
 		goto done;
@@ -394,7 +404,7 @@ static struct vfsmount *autofs4_d_automount(struct path *path)
 		}
 		ino->flags |= AUTOFS_INF_PENDING;
 		spin_unlock(&sbi->fs_lock);
-		status = autofs4_mount_wait(dentry);
+		status = autofs4_mount_wait(dentry, 0);
 		spin_lock(&sbi->fs_lock);
 		ino->flags &= ~AUTOFS_INF_PENDING;
 		if (status) {
@@ -430,21 +440,25 @@ static int autofs4_d_manage(struct dentry *dentry, bool rcu_walk)
 		return 0;
 	}
 
-	/* We need to sleep, so we need pathwalk to be in ref-mode */
-	if (rcu_walk)
-		return -ECHILD;
-
 	/* Wait for pending expires */
-	do_expire_wait(dentry);
+	if (do_expire_wait(dentry, rcu_walk) == -ECHILD)
+		return -ECHILD;
 
 	/*
 	 * This dentry may be under construction so wait on mount
 	 * completion.
 	 */
-	status = autofs4_mount_wait(dentry);
+	status = autofs4_mount_wait(dentry, rcu_walk);
 	if (status)
 		return status;
 
+	if (rcu_walk)
+		/* it is always safe to return 0 as the worst that
+		 * will happen is we retry in REF-walk mode.
+		 * Better than always taking a lock.
+		 */
+		return 0;
+
 	spin_lock(&sbi->fs_lock);
 	/*
 	 * If the dentry has been selected for expire while we slept