[PATCH] VFS autmounter support v2

[David Howells <dhowells@redhat.com>]

Hi Linus, Al,

I've revised my patch to make sure a process in one namespace doesn't change
the topology of another namespace (kern_automount() will return an error in
that case, as does (u)mount). As a bonus, check_mnt() has been simplified to
take account of the namespace pointer now in vfsmount.

I've also fixes it so that if an automounted subtree is touched with "mount
--bind" or "mount --move", then the base of that tree has its expiry timeout
cleared, thus making it a permanent fixture (until manually unmounted).

David


diff -uNr linux-2.5.72/fs/namei.c linux-2.5.72-auto/fs/namei.c
--- linux-2.5.72/fs/namei.c	2003-06-17 15:01:51.000000000 +0100
+++ linux-2.5.72-auto/fs/namei.c	2003-06-17 15:06:42.000000000 +0100
@@ -434,23 +434,35 @@
 	return 1;
 }
 
-static int follow_mount(struct vfsmount **mnt, struct dentry **dentry)
+static int follow_mount(struct vfsmount **mnt, struct dentry **dentry, unsigned int flags)
 {
 	int res = 0;
-	while (d_mountpoint(*dentry)) {
-		struct vfsmount *mounted;
-		spin_lock(&dcache_lock);
-		mounted = lookup_mnt(*mnt, *dentry);
-		if (!mounted) {
+	for (;;) {
+
+		if (d_mountpoint(*dentry)) {
+			struct vfsmount *mounted;
+			spin_lock(&dcache_lock);
+			mounted = lookup_mnt(*mnt, *dentry);
+			if (!mounted) {
+				spin_unlock(&dcache_lock);
+				break;
+			}
+			*mnt = mntget(mounted);
 			spin_unlock(&dcache_lock);
+			dput(*dentry);
+			mntput(mounted->mnt_parent);
+			*dentry = dget(mounted->mnt_root);
+			res = 1;
+
+		} else if (d_automount_point(*dentry)) {
+			if (flags & LOOKUP_NOAUTOMOUNT)
+				break;
+			res = kern_automount(*mnt, *dentry);
+			if (res < 0)
+				break;
+		} else {
 			break;
 		}
-		*mnt = mntget(mounted);
-		spin_unlock(&dcache_lock);
-		dput(*dentry);
-		mntput(mounted->mnt_parent);
-		*dentry = dget(mounted->mnt_root);
-		res = 1;
 	}
 	return res;
 }
@@ -510,7 +522,7 @@
 		mntput(*mnt);
 		*mnt = parent;
 	}
-	follow_mount(mnt, dentry);
+	follow_mount(mnt, dentry, 0);
 }
 
 struct path {
@@ -643,7 +655,9 @@
 		if (err)
 			break;
 		/* Check mountpoints.. */
-		follow_mount(&next.mnt, &next.dentry);
+		err = follow_mount(&next.mnt, &next.dentry, 0);
+		if (err < 0)
+			goto out_dput;
 
 		err = -ENOENT;
 		inode = next.dentry->d_inode;
@@ -703,7 +717,10 @@
 		err = do_lookup(nd, &this, &next, 0);
 		if (err)
 			break;
-		follow_mount(&next.mnt, &next.dentry);
+		err = follow_mount(&next.mnt, &next.dentry, nd->flags & LOOKUP_NOAUTOMOUNT);
+		if (err < 0)
+			goto out_dput;
+
 		inode = next.dentry->d_inode;
 		if ((lookup_flags & LOOKUP_FOLLOW)
 		    && inode && inode->i_op && inode->i_op->follow_link) {
diff -uNr linux-2.5.72/fs/namespace.c linux-2.5.72-auto/fs/namespace.c
--- linux-2.5.72/fs/namespace.c	2003-06-17 15:01:51.000000000 +0100
+++ linux-2.5.72-auto/fs/namespace.c	2003-06-18 15:04:30.000000000 +0100
@@ -30,6 +30,8 @@
 static int hash_mask, hash_bits;
 static kmem_cache_t *mnt_cache; 
 
+static void kern_automount_expiry_timeout(void *_data);
+
 static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry)
 {
 	unsigned long tmp = ((unsigned long) mnt / L1_CACHE_BYTES);
@@ -84,13 +86,9 @@
 	return p;
 }
 
-static int check_mnt(struct vfsmount *mnt)
+static inline int check_mnt(struct vfsmount *mnt)
 {
-	spin_lock(&dcache_lock);
-	while (mnt->mnt_parent != mnt)
-		mnt = mnt->mnt_parent;
-	spin_unlock(&dcache_lock);
-	return mnt == current->namespace->root;
+	return mnt->mnt_namespace == current->namespace;
 }
 
 static void detach_mnt(struct vfsmount *mnt, struct nameidata *old_nd)
@@ -142,6 +140,9 @@
 		mnt->mnt_root = dget(root);
 		mnt->mnt_mountpoint = mnt->mnt_root;
 		mnt->mnt_parent = mnt;
+		mnt->mnt_namespace = old->mnt_namespace;
+		mnt->mnt_expiry_timeout = old->mnt_expiry_timeout;
+		mnt->mnt_expires_at = old->mnt_expires_at;
 	}
 	return mnt;
 }
@@ -530,6 +531,7 @@
 	}
 
 	if (mnt) {
+		mnt->mnt_expiry_timeout = 0;
 		err = graft_tree(mnt, nd);
 		if (err) {
 			spin_lock(&dcache_lock);
@@ -622,6 +624,7 @@
 
 	detach_mnt(old_nd.mnt, &parent_nd);
 	attach_mnt(old_nd.mnt, nd);
+	old_nd.mnt->mnt_expiry_timeout = 0;
 out2:
 	spin_unlock(&dcache_lock);
 out1:
@@ -674,6 +677,138 @@
 	return err;
 }
 
+int kern_automount(struct vfsmount *on_mnt, struct dentry *on_dentry)
+{
+	struct nameidata nd;
+	struct vfsmount *mnt;
+	int err;
+
+	if (!on_dentry->d_inode || !S_ISDIR(on_dentry->d_inode->i_mode))
+		return -ENOTDIR;
+
+	if (!check_mnt(on_mnt))
+		return -EACCES;
+
+	mnt = on_dentry->d_op->d_automount(on_dentry);
+	if (IS_ERR(mnt))
+		return PTR_ERR(mnt);
+
+	mnt->mnt_expires_at = get_seconds() + mnt->mnt_expiry_timeout;
+
+	memset(&nd,0,sizeof(nd));
+	nd.dentry = on_dentry;
+	nd.mnt = on_mnt;
+
+	down_write(&current->namespace->sem);
+	/* Something was mounted here while we slept */
+	while(d_mountpoint(nd.dentry) && follow_down(&nd.mnt, &nd.dentry))
+		;
+	err = -EINVAL;
+	if (!check_mnt(nd.mnt))
+		goto unlock;
+
+	/* Refuse the same filesystem on the same mount point */
+	err = -EBUSY;
+	if (nd.mnt->mnt_sb == mnt->mnt_sb && nd.mnt->mnt_root == nd.dentry)
+		goto unlock;
+
+	err = graft_tree(mnt, &nd);
+unlock:
+	up_write(&current->namespace->sem);
+	mntput(mnt);
+	return err;
+}
+
+static inline void kern_automount_set_expiry_timer(struct namespace *namespace,
+						   unsigned long timeout)
+{
+	spin_lock(&dcache_lock);
+
+	if (atomic_read(&namespace->count) > 0) {
+		get_namespace(namespace);
+
+		if (!schedule_delayed_work(&namespace->mnt_expiry_work,
+					   (timeout + 1) * HZ))
+			put_namespace(namespace);
+	}
+
+	spin_unlock(&dcache_lock);
+}
+
+void kern_automount_begin_expiry(struct vfsmount *mnt)
+{
+	mnt->mnt_expires_at = get_seconds() + mnt->mnt_expiry_timeout;
+
+	kern_automount_set_expiry_timer(mnt->mnt_namespace,
+					mnt->mnt_expiry_timeout);
+}
+
+static void kern_automount_expiry_timeout(void *_data)
+{
+	struct namespace *namespace = _data;
+	struct list_head *_p, *_n, graveyard;
+	struct vfsmount *mnt;
+	time_t now;
+	int timeout = INT_MAX, tmp;
+
+	INIT_LIST_HEAD(&graveyard);
+
+	down_write(&namespace->sem);
+
+	now = get_seconds();
+
+	list_for_each_safe(_p, _n, &namespace->list) {
+		mnt = list_entry(_p, struct vfsmount, mnt_list);
+
+		if (mnt->mnt_expiry_timeout &&
+		    atomic_read(&mnt->mnt_count) == 1) {
+			spin_lock(&dcache_lock);
+
+			if (atomic_read(&mnt->mnt_count) == 1) {
+				tmp = (int) mnt->mnt_expires_at - (int) now;
+				if (tmp <= 0) {
+					list_move_tail(&mnt->mnt_list,
+						       &graveyard);
+					list_del_init(&mnt->mnt_child);
+					list_del_init(&mnt->mnt_hash);
+					mnt->mnt_mountpoint->d_mounted--;
+				} else if (tmp < timeout) {
+					timeout = tmp;
+				}
+			}
+
+			spin_unlock(&dcache_lock);
+		}
+	}
+
+	up_write(&namespace->sem);
+
+	while (!list_empty(&graveyard)) {
+		mnt = list_entry(graveyard.next, struct vfsmount, mnt_list);
+		list_del_init(&mnt->mnt_list);
+
+		dput(xchg(&mnt->mnt_mountpoint, mnt->mnt_root));
+		mntput(xchg(&mnt->mnt_parent, mnt));
+
+		if (atomic_read(&mnt->mnt_sb->s_active) == 1) {
+			/* last instance - try to be smart */
+			lock_kernel();
+			DQUOT_OFF(mnt->mnt_sb);
+			acct_auto_close(mnt->mnt_sb);
+			unlock_kernel();
+		}
+
+		mntput(mnt);
+	}
+
+	if (timeout != INT_MAX) {
+		kern_automount_set_expiry_timer(namespace, timeout);
+	}
+	else {
+		put_namespace(namespace);
+	}
+}
+
 static int copy_mount_options (const void __user *data, unsigned long *where)
 {
 	int i;
@@ -800,6 +935,9 @@
 	init_rwsem(&new_ns->sem);
 	new_ns->root = NULL;
 	INIT_LIST_HEAD(&new_ns->list);
+	INIT_WORK(&new_ns->mnt_expiry_work,
+		  kern_automount_expiry_timeout,
+		  new_ns);
 
 	down_write(&tsk->namespace->sem);
 	/* First pass: copy the tree topology */
@@ -816,6 +954,8 @@
 		p = namespace->root;
 		q = new_ns->root;
 		while (p) {
+			q->mnt_namespace = new_ns;
+
 			if (p == fs->rootmnt) {
 				rootmnt = p;
 				fs->rootmnt = mntget(q);
@@ -844,6 +984,8 @@
 	if (altrootmnt)
 		mntput(altrootmnt);
 
+	kern_automount_set_expiry_timer(namespace, 1);
+
 	put_namespace(namespace);
 	return 0;
 
@@ -1079,9 +1221,13 @@
 		panic("Can't allocate initial namespace");
 	atomic_set(&namespace->count, 1);
 	INIT_LIST_HEAD(&namespace->list);
+	INIT_WORK(&namespace->mnt_expiry_work,
+		  kern_automount_expiry_timeout,
+		  namespace);
 	init_rwsem(&namespace->sem);
 	list_add(&mnt->mnt_list, &namespace->list);
 	namespace->root = mnt;
+	mnt->mnt_namespace = namespace;
 
 	init_task.namespace = namespace;
 	read_lock(&tasklist_lock);
diff -uNr linux-2.5.72/fs/stat.c linux-2.5.72-auto/fs/stat.c
--- linux-2.5.72/fs/stat.c	2003-06-17 15:01:51.000000000 +0100
+++ linux-2.5.72-auto/fs/stat.c	2003-06-17 15:06:42.000000000 +0100
@@ -61,7 +61,7 @@
 	struct nameidata nd;
 	int error;
 
-	error = user_path_walk(name, &nd);
+	error = user_path_walk_stat(name, &nd);
 	if (!error) {
 		error = vfs_getattr(nd.mnt, nd.dentry, stat);
 		path_release(&nd);
@@ -74,7 +74,7 @@
 	struct nameidata nd;
 	int error;
 
-	error = user_path_walk_link(name, &nd);
+	error = user_path_walk_link_stat(name, &nd);
 	if (!error) {
 		error = vfs_getattr(nd.mnt, nd.dentry, stat);
 		path_release(&nd);
diff -uNr linux-2.5.72/fs/super.c linux-2.5.72-auto/fs/super.c
--- linux-2.5.72/fs/super.c	2003-06-17 15:01:51.000000000 +0100
+++ linux-2.5.72-auto/fs/super.c	2003-06-17 15:06:42.000000000 +0100
@@ -21,6 +21,7 @@
  */
 
 #include <linux/config.h>
+#include <linux/module.h>
 #include <linux/slab.h>
 #include <linux/smp_lock.h>
 #include <linux/acct.h>
@@ -683,6 +684,7 @@
 	mnt->mnt_root = dget(sb->s_root);
 	mnt->mnt_mountpoint = sb->s_root;
 	mnt->mnt_parent = mnt;
+	mnt->mnt_namespace = current->namespace;
 	up_write(&sb->s_umount);
 	put_filesystem(type);
 	return mnt;
@@ -697,6 +699,8 @@
 	return (struct vfsmount *)sb;
 }
 
+EXPORT_SYMBOL_GPL(do_kern_mount);
+
 struct vfsmount *kern_mount(struct file_system_type *type)
 {
 	return do_kern_mount(type->name, 0, type->name, NULL);
diff -uNr linux-2.5.72/include/linux/dcache.h linux-2.5.72-auto/include/linux/dcache.h
--- linux-2.5.72/include/linux/dcache.h	2003-06-17 15:01:36.000000000 +0100
+++ linux-2.5.72-auto/include/linux/dcache.h	2003-06-17 15:06:42.000000000 +0100
@@ -112,6 +112,7 @@
 	int (*d_delete)(struct dentry *);
 	void (*d_release)(struct dentry *);
 	void (*d_iput)(struct dentry *, struct inode *);
+	struct vfsmount *(*d_automount)(struct dentry *);
 };
 
 /* the dentry parameter passed to d_hash and d_compare is the parent
@@ -307,6 +308,11 @@
 	return dentry->d_mounted;
 }
 
+static inline int d_automount_point(struct dentry *dentry)
+{
+	return dentry->d_op && dentry->d_op->d_automount;
+}
+
 extern struct vfsmount *lookup_mnt(struct vfsmount *, struct dentry *);
 #endif /* __KERNEL__ */
 
diff -uNr linux-2.5.72/include/linux/mount.h linux-2.5.72-auto/include/linux/mount.h
--- linux-2.5.72/include/linux/mount.h	2003-06-17 15:01:35.000000000 +0100
+++ linux-2.5.72-auto/include/linux/mount.h	2003-06-17 15:06:42.000000000 +0100
@@ -31,6 +31,9 @@
 	int mnt_flags;
 	char *mnt_devname;		/* Name of device e.g. /dev/dsk/hda1 */
 	struct list_head mnt_list;
+	struct namespace *mnt_namespace; /* containing namespace */
+	time_t mnt_expires_at;		/* time at which automount expires */
+	unsigned mnt_expiry_timeout;	/* expiry timeout (in seconds) or 0 */
 };
 
 static inline struct vfsmount *mntget(struct vfsmount *mnt)
@@ -40,11 +43,15 @@
 	return mnt;
 }
 
+extern void kern_automount_begin_expiry(struct vfsmount *mnt);
 extern void __mntput(struct vfsmount *mnt);
 
 static inline void mntput(struct vfsmount *mnt)
 {
 	if (mnt) {
+		if (atomic_read(&mnt->mnt_count) == 2 &&
+		    mnt->mnt_expiry_timeout)
+			kern_automount_begin_expiry(mnt);
 		if (atomic_dec_and_test(&mnt->mnt_count))
 			__mntput(mnt);
 	}
diff -uNr linux-2.5.72/include/linux/namei.h linux-2.5.72-auto/include/linux/namei.h
--- linux-2.5.72/include/linux/namei.h	2003-06-17 15:01:36.000000000 +0100
+++ linux-2.5.72-auto/include/linux/namei.h	2003-06-17 15:06:42.000000000 +0100
@@ -31,6 +31,7 @@
 #define LOOKUP_CONTINUE		 4
 #define LOOKUP_PARENT		16
 #define LOOKUP_NOALT		32
+#define LOOKUP_NOAUTOMOUNT	64
 
 
 extern int FASTCALL(__user_walk(const char __user *, unsigned, struct nameidata *));
@@ -38,6 +39,10 @@
 	__user_walk(name, LOOKUP_FOLLOW, nd)
 #define user_path_walk_link(name,nd) \
 	__user_walk(name, 0, nd)
+#define user_path_walk_stat(name,nd) \
+	__user_walk(name, LOOKUP_FOLLOW|LOOKUP_NOAUTOMOUNT, nd)
+#define user_path_walk_link_stat(name,nd) \
+	__user_walk(name, LOOKUP_NOAUTOMOUNT, nd)
 extern int FASTCALL(path_lookup(const char *, unsigned, struct nameidata *));
 extern int FASTCALL(path_walk(const char *, struct nameidata *));
 extern int FASTCALL(link_path_walk(const char *, struct nameidata *));
@@ -52,4 +57,6 @@
 extern struct dentry *lock_rename(struct dentry *, struct dentry *);
 extern void unlock_rename(struct dentry *, struct dentry *);
 
+extern int kern_automount(struct vfsmount *mnt, struct dentry *dentry);
+
 #endif /* _LINUX_NAMEI_H */
diff -uNr linux-2.5.72/include/linux/namespace.h linux-2.5.72-auto/include/linux/namespace.h
--- linux-2.5.72/include/linux/namespace.h	2003-06-17 15:01:36.000000000 +0100
+++ linux-2.5.72-auto/include/linux/namespace.h	2003-06-17 15:06:42.000000000 +0100
@@ -10,6 +10,7 @@
 	struct vfsmount *	root;
 	struct list_head	list;
 	struct rw_semaphore	sem;
+	struct work_struct	mnt_expiry_work;
 };
 
 extern void umount_tree(struct vfsmount *);