[PATCH 0/9] [RFC v2] safely drop directory dentry on failed revalidate

@@ -1224,6 +1224,163 @@ void shrink_dcache_parent(struct dentry * parent)
 }
 EXPORT_SYMBOL(shrink_dcache_parent);
 
+static int __check_submounts_and_drop(struct dentry *parent,
+					     struct list_head *dispose)
+{
+	struct dentry *this_parent;
+	struct list_head *next;
+	unsigned seq;
+	int found = 0;
+	int locked = 0;
+
+	seq = read_seqbegin(&rename_lock);
+again:
+	this_parent = parent;
+	spin_lock(&this_parent->d_lock);
+repeat:
+	next = this_parent->d_subdirs.next;
+resume:
+	while (next != &this_parent->d_subdirs) {
+		struct list_head *tmp = next;
+		struct dentry *dentry;
+
+		dentry = list_entry(tmp, struct dentry, d_u.d_child);
+		next = tmp->next;
+
+		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
+		if (d_mountpoint(dentry)) {
+			spin_unlock(&dentry->d_lock);
+			found = -EBUSY;
+			goto out;
+		}
+
+		/*
+		 * move only zero ref count dentries to the dispose list.
+		 *
+		 * Those which are presently on the shrink list, being processed
+		 * by shrink_dentry_list(), shouldn't be moved.  Otherwise the
+		 * loop in shrink_dcache_parent() might not make any progress
+		 * and loop forever.
+		 */
+		if (dentry->d_count) {
+			dentry_lru_del(dentry);
+		} else if (!(dentry->d_flags & DCACHE_SHRINK_LIST)) {
+			dentry_lru_move_list(dentry, dispose);
+			dentry->d_flags |= DCACHE_SHRINK_LIST;
+			found++;
+		}
+		/*
+		 * We can return to the caller if we have found some (this
+		 * ensures forward progress). We'll be coming back to find
+		 * the rest.
+		 */
+		if (found && need_resched()) {
+			spin_unlock(&dentry->d_lock);
+			goto out;
+		}
+
+		/*
+		 * Descend a level if the d_subdirs list is non-empty.
+		 */
+		if (!list_empty(&dentry->d_subdirs)) {
+			spin_unlock(&this_parent->d_lock);
+			spin_release(&dentry->d_lock.dep_map, 1, _RET_IP_);
+			this_parent = dentry;
+			spin_acquire(&this_parent->d_lock.dep_map, 0, 1, _RET_IP_);
+			goto repeat;
+		}
+
+		spin_unlock(&dentry->d_lock);
+	}
+	/*
+	 * All done at this level ... ascend and resume the search.
+	 */
+	if (this_parent != parent) {
+		struct dentry *child = this_parent;
+		this_parent = try_to_ascend(this_parent, locked, seq);
+		if (!this_parent)
+			goto rename_retry;
+		next = child->d_u.d_child.next;
+		goto resume;
+	}
+	if (!locked && read_seqretry(&rename_lock, seq)) {
+		spin_unlock(&this_parent->d_lock);
+		goto rename_retry;
+	}
+	if (d_mountpoint(this_parent))
+		found = -EBUSY;
+	if (!found)
+		__d_drop(this_parent);
+out:
+	spin_unlock(&this_parent->d_lock);
+
+	if (locked)
+		write_sequnlock(&rename_lock);
+	return found;
+
+rename_retry:
+	if (found)
+		return found;
+	if (locked)
+		goto again;
+	locked = 1;
+	write_seqlock(&rename_lock);
+	goto again;
+}
+
+/**
+ * check_submounts_and_drop - prune dcache, check for submounts and drop
+ *
+ * All done as a single atomic operation relative to has_unlinked_ancestor().
+ * Returns 0 if successfully unhashed @parent.  If there were submounts then
+ * return -EBUSY.
+ *
+ * @dentry: dentry to prune and drop
+ */
+int check_submounts_and_drop(struct dentry *dentry)
+{
+	int ret = 0;
+
+	/* Negative dentries can be dropped without further checks */
+	if (!dentry->d_inode) {
+		d_drop(dentry);
+		goto out;
+	}
+
+	spin_lock(&dentry->d_lock);
+	if (d_unhashed(dentry))
+		goto out_unlock;
+	if (list_empty(&dentry->d_subdirs)) {
+		if (d_mountpoint(dentry)) {
+			ret = -EBUSY;
+			goto out_unlock;
+		}
+		__d_drop(dentry);
+		goto out_unlock;
+	}
+	spin_unlock(&dentry->d_lock);
+
+	for (;;) {
+		LIST_HEAD(dispose);
+		ret = __check_submounts_and_drop(dentry, &dispose);
+		if (!list_empty(&dispose))
+			shrink_dentry_list(&dispose);
+
+		if (ret <= 0)
+			break;
+
+		cond_resched();
+	}
+
+out:
+	return ret;
+
+out_unlock:
+	spin_unlock(&dentry->d_lock);
+	goto out;
+}
+EXPORT_SYMBOL(check_submounts_and_drop);
+
 /**
  * __d_alloc	-	allocate a dcache entry
  * @sb: filesystem it will belong to

@@ -251,6 +251,7 @@ extern void d_prune_aliases(struct inode *);
 
 /* test whether we have any submounts in a subdir tree */
 extern int have_submounts(struct dentry *);
+extern int check_submounts_and_drop(struct dentry *);
 
 /*
  * This adds the entry to the hash queues.

@@ -1103,6 +1103,41 @@ rename_retry:
 }
 EXPORT_SYMBOL(have_submounts);
 
+static bool __has_unlinked_ancestor(struct dentry *dentry)
+{
+	struct dentry *this;
+
+	for (this = dentry; !IS_ROOT(this); this = this->d_parent) {
+		int is_unhashed;
+
+		/* Need exclusion wrt. check_submounts_and_drop() */
+		spin_lock(&this->d_lock);
+		is_unhashed = d_unhashed(this);
+		spin_unlock(&this->d_lock);
+
+		if (is_unhashed)
+			return true;
+	}
+	return false;
+}
+
+/*
+ * Called by mount code to check if the mountpoint is reachable (e.g. NFS can
+ * unhash a directory dentry and then the complete subtree can become
+ * unreachable).
+ */
+bool has_unlinked_ancestor(struct dentry *dentry)
+{
+	bool found;
+
+	/* Need exclusion wrt. check_submounts_and_drop() */
+	write_seqlock(&rename_lock);
+	found = __has_unlinked_ancestor(dentry);
+	write_sequnlock(&rename_lock);
+
+	return found;
+}
+
 /*
  * Search the dentry child list of the specified parent,
  * and move any unused dentries to the end of the unused

@@ -126,6 +126,7 @@ extern int invalidate_inodes(struct super_block *, bool);
  * dcache.c
  */
 extern struct dentry *__d_alloc(struct super_block *, const struct qstr *);
+extern bool has_unlinked_ancestor(struct dentry *dentry);
 
 /*
  * read_write.c

@@ -634,6 +634,15 @@ static struct mountpoint *new_mountpoint(struct dentry *dentry)
 	}
 	dentry->d_flags |= DCACHE_MOUNTED;
 	spin_unlock(&dentry->d_lock);
+
+	if (has_unlinked_ancestor(dentry)) {
+		spin_lock(&dentry->d_lock);
+		dentry->d_flags &= ~DCACHE_MOUNTED;
+		spin_unlock(&dentry->d_lock);
+		kfree(mp);
+		return ERR_PTR(-ENOENT);
+	}
+
 	mp->m_dentry = dentry;
 	mp->m_count = 1;
 	list_add(&mp->m_hash, chain);

@@ -685,16 +685,12 @@ not_found:
 	spin_unlock(&dentry->d_lock);
 
 out_bad:
-	if (dentry->d_inode) {
-		/* don't unhash if we have submounts */
-		if (have_submounts(dentry))
-			goto out_skip;
-	}
+	/* don't unhash if we have submounts */
+	if (check_submounts_and_drop(dentry) != 0)
+		goto out_skip;
 
 	_debug("dropping dentry %s/%s",
 	       parent->d_name.name, dentry->d_name.name);
-	shrink_dcache_parent(dentry);
-	d_drop(dentry);
 	dput(parent);
 	key_put(key);
 

@@ -93,12 +93,9 @@ invalid_gunlock:
 	if (!had_lock)
 		gfs2_glock_dq_uninit(&d_gh);
 invalid:
-	if (inode && S_ISDIR(inode->i_mode)) {
-		if (have_submounts(dentry))
-			goto valid;
-		shrink_dcache_parent(dentry);
-	}
-	d_drop(dentry);
+	if (check_submounts_and_drop(dentry) != 0)
+		goto valid;
+
 	dput(parent);
 	return 0;
 

@@ -1135,14 +1135,13 @@ out_zap_parent:
 	if (inode && S_ISDIR(inode->i_mode)) {
 		/* Purge readdir caches. */
 		nfs_zap_caches(inode);
-		/* If we have submounts, don't unhash ! */
-		if (have_submounts(dentry))
-			goto out_valid;
 		if (dentry->d_flags & DCACHE_DISCONNECTED)
 			goto out_valid;
-		shrink_dcache_parent(dentry);
 	}
-	d_drop(dentry);
+	/* If we have submounts, don't unhash ! */
+	if (check_submounts_and_drop(dentry) != 0)
+		goto out_valid;
+
 	dput(parent);
 	dfprintk(LOOKUPCACHE, "NFS: %s(%s/%s) is invalid\n",
 			__func__, dentry->d_parent->d_name.name,

@@ -297,7 +297,6 @@ static int sysfs_dentry_delete(const struct dentry *dentry)
 static int sysfs_dentry_revalidate(struct dentry *dentry, unsigned int flags)
 {
 	struct sysfs_dirent *sd;
-	int is_dir;
 	int type;
 
 	if (flags & LOOKUP_RCU)

@@ -341,18 +340,15 @@ out_bad:
 	 * is performed at its new name the dentry will be readded
 	 * to the dcache hashes.
 	 */
-	is_dir = (sysfs_type(sd) == SYSFS_DIR);
 	mutex_unlock(&sysfs_mutex);
-	if (is_dir) {
-		/* If we have submounts we must allow the vfs caches
-		 * to lie about the state of the filesystem to prevent
-		 * leaks and other nasty things.
-		 */
-		if (have_submounts(dentry))
-			goto out_valid;
-		shrink_dcache_parent(dentry);
-	}
-	d_drop(dentry);
+
+	/* If we have submounts we must allow the vfs caches
+	 * to lie about the state of the filesystem to prevent
+	 * leaks and other nasty things.
+	 */
+	if (check_submounts_and_drop(dentry) != 0)
+		goto out_valid;
+
 	return 0;
 }
 

@@ -182,10 +182,11 @@ static int fuse_dentry_revalidate(struct dentry *entry, unsigned int flags)
 	struct inode *inode;
 	struct dentry *parent;
 	struct fuse_conn *fc;
+	int ret;
 
 	inode = ACCESS_ONCE(entry->d_inode);
 	if (inode && is_bad_inode(inode))
-		return 0;
+		goto invalid;
 	else if (fuse_dentry_time(entry) < get_jiffies_64()) {
 		int err;
 		struct fuse_entry_out outarg;

@@ -195,20 +196,23 @@ static int fuse_dentry_revalidate(struct dentry *entry, unsigned int flags)
 
 		/* For negative dentries, always do a fresh lookup */
 		if (!inode)
-			return 0;
+			goto invalid;
 
+		ret = -ECHILD;
 		if (flags & LOOKUP_RCU)
-			return -ECHILD;
+			goto out;
 
 		fc = get_fuse_conn(inode);
 		req = fuse_get_req_nopages(fc);
+		ret = PTR_ERR(req);
 		if (IS_ERR(req))
-			return 0;
+			goto out;
 
 		forget = fuse_alloc_forget();
 		if (!forget) {
 			fuse_put_request(fc, req);
-			return 0;
+			ret = -ENOMEM;
+			goto out;
 		}
 
 		attr_version = fuse_get_attr_version(fc);

@@ -227,7 +231,7 @@ static int fuse_dentry_revalidate(struct dentry *entry, unsigned int flags)
 			struct fuse_inode *fi = get_fuse_inode(inode);
 			if (outarg.nodeid != get_node_id(inode)) {
 				fuse_queue_forget(fc, forget, outarg.nodeid, 1);
-				return 0;
+				goto invalid;
 			}
 			spin_lock(&fc->lock);
 			fi->nlookup++;

@@ -235,7 +239,7 @@ static int fuse_dentry_revalidate(struct dentry *entry, unsigned int flags)
 		}
 		kfree(forget);
 		if (err || (outarg.attr.mode ^ inode->i_mode) & S_IFMT)
-			return 0;
+			goto invalid;
 
 		fuse_change_attributes(inode, &outarg.attr,
 				       entry_attr_timeout(&outarg),

@@ -249,7 +253,13 @@ static int fuse_dentry_revalidate(struct dentry *entry, unsigned int flags)
 			dput(parent);
 		}
 	}
-	return 1;
+	ret = 1;
+out:
+	return ret;
+
+invalid:
+	ret = 0;
+	goto out;
 }
 
 static int invalid_nodeid(u64 nodeid)

@@ -259,6 +259,8 @@ out:
 
 invalid:
 	ret = 0;
+	if (check_submounts_and_drop(entry) != 0)
+		ret = 1;
 	goto out;
 }
 

@@ -267,26 +267,6 @@ int fuse_valid_type(int m)
 		S_ISBLK(m) || S_ISFIFO(m) || S_ISSOCK(m);
 }
 
-/*
- * Add a directory inode to a dentry, ensuring that no other dentry
- * refers to this inode.  Called with fc->inst_mutex.
- */
-static struct dentry *fuse_d_add_directory(struct dentry *entry,
-					   struct inode *inode)
-{
-	struct dentry *alias = d_find_alias(inode);
-	if (alias && !(alias->d_flags & DCACHE_DISCONNECTED)) {
-		/* This tries to shrink the subtree below alias */
-		fuse_invalidate_entry(alias);
-		dput(alias);
-		if (!hlist_empty(&inode->i_dentry))
-			return ERR_PTR(-EBUSY);
-	} else {
-		dput(alias);
-	}
-	return d_splice_alias(inode, entry);
-}
-
 int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
 		     struct fuse_entry_out *outarg, struct inode **inode)
 {

@@ -345,6 +325,24 @@ int fuse_lookup_name(struct super_block *sb, u64 nodeid, struct qstr *name,
 	return err;
 }
 
+static struct dentry *fuse_materialise_dentry(struct dentry *dentry,
+					      struct inode *inode)
+{
+	struct dentry *newent;
+
+	if (inode && S_ISDIR(inode->i_mode)) {
+		struct fuse_conn *fc = get_fuse_conn(inode);
+
+		mutex_lock(&fc->inst_mutex);
+		newent = d_materialise_unique(dentry, inode);
+		mutex_unlock(&fc->inst_mutex);
+	} else {
+		newent = d_materialise_unique(dentry, inode);
+	}
+
+	return newent;
+}
+
 static struct dentry *fuse_lookup(struct inode *dir, struct dentry *entry,
 				  unsigned int flags)
 {

@@ -352,7 +350,6 @@ static struct dentry *fuse_lookup(struct inode *dir, struct dentry *entry,
 	struct fuse_entry_out outarg;
 	struct inode *inode;
 	struct dentry *newent;
-	struct fuse_conn *fc = get_fuse_conn(dir);
 	bool outarg_valid = true;
 
 	err = fuse_lookup_name(dir->i_sb, get_node_id(dir), &entry->d_name,

@@ -368,16 +365,10 @@ static struct dentry *fuse_lookup(struct inode *dir, struct dentry *entry,
 	if (inode && get_node_id(inode) == FUSE_ROOT_ID)
 		goto out_iput;
 
-	if (inode && S_ISDIR(inode->i_mode)) {
-		mutex_lock(&fc->inst_mutex);
-		newent = fuse_d_add_directory(entry, inode);
-		mutex_unlock(&fc->inst_mutex);
-		err = PTR_ERR(newent);
-		if (IS_ERR(newent))
-			goto out_iput;
-	} else {
-		newent = d_splice_alias(inode, entry);
-	}
+	newent = fuse_materialise_dentry(entry, inode);
+	err = PTR_ERR(newent);
+	if (IS_ERR(newent))
+		goto out_err;
 
 	entry = newent ? newent : entry;
 	if (outarg_valid)

@@ -1275,18 +1266,10 @@ static int fuse_direntplus_link(struct file *file,
 	if (!inode)
 		goto out;
 
-	if (S_ISDIR(inode->i_mode)) {
-		mutex_lock(&fc->inst_mutex);
-		alias = fuse_d_add_directory(dentry, inode);
-		mutex_unlock(&fc->inst_mutex);
-		err = PTR_ERR(alias);
-		if (IS_ERR(alias)) {
-			iput(inode);
-			goto out;
-		}
-	} else {
-		alias = d_splice_alias(inode, dentry);
-	}
+	alias = fuse_materialise_dentry(dentry, inode);
+	err = PTR_ERR(alias);
+	if (IS_ERR(alias))
+		goto out;
 
 	if (alias) {
 		dput(dentry);

@@ -1007,8 +1007,6 @@ static int nilfs_tree_was_touched(struct dentry *root_dentry)
  */
 static int nilfs_try_to_shrink_tree(struct dentry *root_dentry)
 {
-	if (have_submounts(root_dentry))
-		return true;
 	shrink_dcache_parent(root_dentry);
 	return nilfs_tree_was_touched(root_dentry);
 }

@@ -994,23 +994,16 @@ static int nilfs_attach_snapshot(struct super_block *s, __u64 cno,
 	return ret;
 }
 
-static int nilfs_tree_was_touched(struct dentry *root_dentry)
-{
-	return d_count(root_dentry) > 1;
-}
-
 /**
- * nilfs_try_to_shrink_tree() - try to shrink dentries of a checkpoint
+ * nilfs_tree_is_busy() - try to shrink dentries of a checkpoint
  * @root_dentry: root dentry of the tree to be shrunk
  *
  * This function returns true if the tree was in-use.
  */
-static int nilfs_try_to_shrink_tree(struct dentry *root_dentry)
+static bool nilfs_tree_is_busy(struct dentry *root_dentry)
 {
-	if (have_submounts(root_dentry))
-		return true;
 	shrink_dcache_parent(root_dentry);
-	return nilfs_tree_was_touched(root_dentry);
+	return d_count(root_dentry) > 1;
 }
 
 int nilfs_checkpoint_is_mounted(struct super_block *sb, __u64 cno)

@@ -1034,8 +1027,7 @@ int nilfs_checkpoint_is_mounted(struct super_block *sb, __u64 cno)
 		if (inode) {
 			dentry = d_find_alias(inode);
 			if (dentry) {
-				if (nilfs_tree_was_touched(dentry))
-					ret = nilfs_try_to_shrink_tree(dentry);
+				ret = nilfs_tree_is_busy(dentry);
 				dput(dentry);
 			}
 			iput(inode);

@@ -1331,11 +1323,8 @@ nilfs_mount(struct file_system_type *fs_type, int flags,
 
 		s->s_flags |= MS_ACTIVE;
 	} else if (!sd.cno) {
-		int busy = false;
-
-		if (nilfs_tree_was_touched(s->s_root)) {
-			busy = nilfs_try_to_shrink_tree(s->s_root);
-			if (busy && (flags ^ s->s_flags) & MS_RDONLY) {
+		if (nilfs_tree_is_busy(s->s_root)) {
+			if ((flags ^ s->s_flags) & MS_RDONLY) {
 				printk(KERN_ERR "NILFS: the device already "
 				       "has a %s mount.\n",
 				       (s->s_flags & MS_RDONLY) ?

@@ -1343,8 +1332,7 @@ nilfs_mount(struct file_system_type *fs_type, int flags,
 				err = -EBUSY;
 				goto failed_super;
 			}
-		}
-		if (!busy) {
+		} else {
 			/*
 			 * Try remount to setup mount states if the current
 			 * tree is not mounted and only snapshots use this sb.