1 files changed, 404 insertions, 167 deletions

diff --git a/fs/notify/mark.c b/fs/notify/mark.c
index 9991f8826734..55a03bb05aa1 100644
--- a/fs/notify/mark.c
+++ b/fs/notify/mark.c
@@ -1,19 +1,6 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
 /*
  *  Copyright (C) 2008 Red Hat, Inc., Eric Paris <eparis@redhat.com>
- *
- *  This program is free software; you can redistribute it and/or modify
- *  it under the terms of the GNU General Public License as published by
- *  the Free Software Foundation; either version 2, or (at your option)
- *  any later version.
- *
- *  This program is distributed in the hope that it will be useful,
- *  but WITHOUT ANY WARRANTY; without even the implied warranty of
- *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- *  GNU General Public License for more details.
- *
- *  You should have received a copy of the GNU General Public License
- *  along with this program; see the file COPYING.  If not, write to
- *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
 /*
@@ -82,6 +69,7 @@
 #include <linux/slab.h>
 #include <linux/spinlock.h>
 #include <linux/srcu.h>
+#include <linux/ratelimit.h>
 
 #include <linux/atomic.h>
 
@@ -105,52 +93,220 @@ static DECLARE_WORK(connector_reaper_work, fsnotify_connector_destroy_workfn);
 
 void fsnotify_get_mark(struct fsnotify_mark *mark)
 {
-	WARN_ON_ONCE(!atomic_read(&mark->refcnt));
-	atomic_inc(&mark->refcnt);
+	WARN_ON_ONCE(!refcount_read(&mark->refcnt));
+	refcount_inc(&mark->refcnt);
+}
+
+static fsnotify_connp_t *fsnotify_object_connp(void *obj,
+				enum fsnotify_obj_type obj_type)
+{
+	switch (obj_type) {
+	case FSNOTIFY_OBJ_TYPE_INODE:
+		return &((struct inode *)obj)->i_fsnotify_marks;
+	case FSNOTIFY_OBJ_TYPE_VFSMOUNT:
+		return &real_mount(obj)->mnt_fsnotify_marks;
+	case FSNOTIFY_OBJ_TYPE_SB:
+		return fsnotify_sb_marks(obj);
+	case FSNOTIFY_OBJ_TYPE_MNTNS:
+		return &((struct mnt_namespace *)obj)->n_fsnotify_marks;
+	default:
+		return NULL;
+	}
+}
+
+static __u32 *fsnotify_conn_mask_p(struct fsnotify_mark_connector *conn)
+{
+	if (conn->type == FSNOTIFY_OBJ_TYPE_INODE)
+		return &fsnotify_conn_inode(conn)->i_fsnotify_mask;
+	else if (conn->type == FSNOTIFY_OBJ_TYPE_VFSMOUNT)
+		return &fsnotify_conn_mount(conn)->mnt_fsnotify_mask;
+	else if (conn->type == FSNOTIFY_OBJ_TYPE_SB)
+		return &fsnotify_conn_sb(conn)->s_fsnotify_mask;
+	else if (conn->type == FSNOTIFY_OBJ_TYPE_MNTNS)
+		return &fsnotify_conn_mntns(conn)->n_fsnotify_mask;
+	return NULL;
+}
+
+__u32 fsnotify_conn_mask(struct fsnotify_mark_connector *conn)
+{
+	if (WARN_ON(!fsnotify_valid_obj_type(conn->type)))
+		return 0;
+
+	return READ_ONCE(*fsnotify_conn_mask_p(conn));
+}
+
+static void fsnotify_get_sb_watched_objects(struct super_block *sb)
+{
+	atomic_long_inc(fsnotify_sb_watched_objects(sb));
+}
+
+static void fsnotify_put_sb_watched_objects(struct super_block *sb)
+{
+	atomic_long_t *watched_objects = fsnotify_sb_watched_objects(sb);
+
+	/* the superblock can go away after this decrement */
+	if (atomic_long_dec_and_test(watched_objects))
+		wake_up_var(watched_objects);
+}
+
+static void fsnotify_get_inode_ref(struct inode *inode)
+{
+	ihold(inode);
+	fsnotify_get_sb_watched_objects(inode->i_sb);
+}
+
+static void fsnotify_put_inode_ref(struct inode *inode)
+{
+	/* read ->i_sb before the inode can go away */
+	struct super_block *sb = inode->i_sb;
+
+	iput(inode);
+	fsnotify_put_sb_watched_objects(sb);
 }
 
 /*
- * Get mark reference when we found the mark via lockless traversal of object
- * list. Mark can be already removed from the list by now and on its way to be
- * destroyed once SRCU period ends.
+ * Grab or drop watched objects reference depending on whether the connector
+ * is attached and has any marks attached.
  */
-static bool fsnotify_get_mark_safe(struct fsnotify_mark *mark)
+static void fsnotify_update_sb_watchers(struct super_block *sb,
+					struct fsnotify_mark_connector *conn)
 {
-	return atomic_inc_not_zero(&mark->refcnt);
+	struct fsnotify_sb_info *sbinfo = fsnotify_sb_info(sb);
+	bool is_watched = conn->flags & FSNOTIFY_CONN_FLAG_IS_WATCHED;
+	struct fsnotify_mark *first_mark = NULL;
+	unsigned int highest_prio = 0;
+
+	if (conn->obj)
+		first_mark = hlist_entry_safe(conn->list.first,
+					      struct fsnotify_mark, obj_list);
+	if (first_mark)
+		highest_prio = first_mark->group->priority;
+	if (WARN_ON(highest_prio >= __FSNOTIFY_PRIO_NUM))
+		highest_prio = 0;
+
+	/*
+	 * If the highest priority of group watching this object is prio,
+	 * then watched object has a reference on counters [0..prio].
+	 * Update priority >= 1 watched objects counters.
+	 */
+	for (unsigned int p = conn->prio + 1; p <= highest_prio; p++)
+		atomic_long_inc(&sbinfo->watched_objects[p]);
+	for (unsigned int p = conn->prio; p > highest_prio; p--)
+		atomic_long_dec(&sbinfo->watched_objects[p]);
+	conn->prio = highest_prio;
+
+	/* Update priority >= 0 (a.k.a total) watched objects counter */
+	BUILD_BUG_ON(FSNOTIFY_PRIO_NORMAL != 0);
+	if (first_mark && !is_watched) {
+		conn->flags |= FSNOTIFY_CONN_FLAG_IS_WATCHED;
+		fsnotify_get_sb_watched_objects(sb);
+	} else if (!first_mark && is_watched) {
+		conn->flags &= ~FSNOTIFY_CONN_FLAG_IS_WATCHED;
+		fsnotify_put_sb_watched_objects(sb);
+	}
 }
 
-static void __fsnotify_recalc_mask(struct fsnotify_mark_connector *conn)
+/*
+ * Grab or drop inode reference for the connector if needed.
+ *
+ * When it's time to drop the reference, we only clear the HAS_IREF flag and
+ * return the inode object. fsnotify_drop_object() will be resonsible for doing
+ * iput() outside of spinlocks. This happens when last mark that wanted iref is
+ * detached.
+ */
+static struct inode *fsnotify_update_iref(struct fsnotify_mark_connector *conn,
+					  bool want_iref)
+{
+	bool has_iref = conn->flags & FSNOTIFY_CONN_FLAG_HAS_IREF;
+	struct inode *inode = NULL;
+
+	if (conn->type != FSNOTIFY_OBJ_TYPE_INODE ||
+	    want_iref == has_iref)
+		return NULL;
+
+	if (want_iref) {
+		/* Pin inode if any mark wants inode refcount held */
+		fsnotify_get_inode_ref(fsnotify_conn_inode(conn));
+		conn->flags |= FSNOTIFY_CONN_FLAG_HAS_IREF;
+	} else {
+		/* Unpin inode after detach of last mark that wanted iref */
+		inode = fsnotify_conn_inode(conn);
+		conn->flags &= ~FSNOTIFY_CONN_FLAG_HAS_IREF;
+	}
+
+	return inode;
+}
+
+static void *__fsnotify_recalc_mask(struct fsnotify_mark_connector *conn)
 {
 	u32 new_mask = 0;
+	bool want_iref = false;
 	struct fsnotify_mark *mark;
 
 	assert_spin_locked(&conn->lock);
+	/* We can get detached connector here when inode is getting unlinked. */
+	if (!fsnotify_valid_obj_type(conn->type))
+		return NULL;
 	hlist_for_each_entry(mark, &conn->list, obj_list) {
-		if (mark->flags & FSNOTIFY_MARK_FLAG_ATTACHED)
-			new_mask |= mark->mask;
+		if (!(mark->flags & FSNOTIFY_MARK_FLAG_ATTACHED))
+			continue;
+		new_mask |= fsnotify_calc_mask(mark);
+		if (conn->type == FSNOTIFY_OBJ_TYPE_INODE &&
+		    !(mark->flags & FSNOTIFY_MARK_FLAG_NO_IREF))
+			want_iref = true;
 	}
-	if (conn->flags & FSNOTIFY_OBJ_TYPE_INODE)
-		conn->inode->i_fsnotify_mask = new_mask;
-	else if (conn->flags & FSNOTIFY_OBJ_TYPE_VFSMOUNT)
-		real_mount(conn->mnt)->mnt_fsnotify_mask = new_mask;
+	/*
+	 * We use WRITE_ONCE() to prevent silly compiler optimizations from
+	 * confusing readers not holding conn->lock with partial updates.
+	 */
+	WRITE_ONCE(*fsnotify_conn_mask_p(conn), new_mask);
+
+	return fsnotify_update_iref(conn, want_iref);
+}
+
+static bool fsnotify_conn_watches_children(
+					struct fsnotify_mark_connector *conn)
+{
+	if (conn->type != FSNOTIFY_OBJ_TYPE_INODE)
+		return false;
+
+	return fsnotify_inode_watches_children(fsnotify_conn_inode(conn));
+}
+
+static void fsnotify_conn_set_children_dentry_flags(
+					struct fsnotify_mark_connector *conn)
+{
+	if (conn->type != FSNOTIFY_OBJ_TYPE_INODE)
+		return;
+
+	fsnotify_set_children_dentry_flags(fsnotify_conn_inode(conn));
 }
 
 /*
  * Calculate mask of events for a list of marks. The caller must make sure
- * connector and connector->inode cannot disappear under us.  Callers achieve
+ * connector and connector->obj cannot disappear under us.  Callers achieve
  * this by holding a mark->lock or mark->group->mark_mutex for a mark on this
  * list.
  */
 void fsnotify_recalc_mask(struct fsnotify_mark_connector *conn)
 {
+	bool update_children;
+
 	if (!conn)
 		return;
 
 	spin_lock(&conn->lock);
+	update_children = !fsnotify_conn_watches_children(conn);
 	__fsnotify_recalc_mask(conn);
+	update_children &= fsnotify_conn_watches_children(conn);
 	spin_unlock(&conn->lock);
-	if (conn->flags & FSNOTIFY_OBJ_TYPE_INODE)
-		__fsnotify_update_child_dentry_flags(conn->inode);
+	/*
+	 * Set children's PARENT_WATCHED flags only if parent started watching.
+	 * When parent stops watching, we clear false positive PARENT_WATCHED
+	 * flags lazily in __fsnotify_parent().
+	 */
+	if (update_children)
+		fsnotify_conn_set_children_dentry_flags(conn);
 }
 
 /* Free all connectors queued for freeing once SRCU period ends */
@@ -171,25 +327,39 @@ static void fsnotify_connector_destroy_workfn(struct work_struct *work)
 	}
 }
 
-static struct inode *fsnotify_detach_connector_from_object(
-					struct fsnotify_mark_connector *conn)
+static void *fsnotify_detach_connector_from_object(
+					struct fsnotify_mark_connector *conn,
+					unsigned int *type)
 {
+	fsnotify_connp_t *connp = fsnotify_object_connp(conn->obj, conn->type);
+	struct super_block *sb = fsnotify_connector_sb(conn);
 	struct inode *inode = NULL;
 
-	if (conn->flags & FSNOTIFY_OBJ_TYPE_INODE) {
-		inode = conn->inode;
-		rcu_assign_pointer(inode->i_fsnotify_marks, NULL);
+	*type = conn->type;
+	if (conn->type == FSNOTIFY_OBJ_TYPE_DETACHED)
+		return NULL;
+
+	if (conn->type == FSNOTIFY_OBJ_TYPE_INODE) {
+		inode = fsnotify_conn_inode(conn);
 		inode->i_fsnotify_mask = 0;
-		conn->inode = NULL;
-		conn->flags &= ~FSNOTIFY_OBJ_TYPE_INODE;
-	} else if (conn->flags & FSNOTIFY_OBJ_TYPE_VFSMOUNT) {
-		rcu_assign_pointer(real_mount(conn->mnt)->mnt_fsnotify_marks,
-				   NULL);
-		real_mount(conn->mnt)->mnt_fsnotify_mask = 0;
-		conn->mnt = NULL;
-		conn->flags &= ~FSNOTIFY_OBJ_TYPE_VFSMOUNT;
+
+		/* Unpin inode when detaching from connector */
+		if (!(conn->flags & FSNOTIFY_CONN_FLAG_HAS_IREF))
+			inode = NULL;
+	} else if (conn->type == FSNOTIFY_OBJ_TYPE_VFSMOUNT) {
+		fsnotify_conn_mount(conn)->mnt_fsnotify_mask = 0;
+	} else if (conn->type == FSNOTIFY_OBJ_TYPE_SB) {
+		fsnotify_conn_sb(conn)->s_fsnotify_mask = 0;
+	} else if (conn->type == FSNOTIFY_OBJ_TYPE_MNTNS) {
+		fsnotify_conn_mntns(conn)->n_fsnotify_mask = 0;
 	}
 
+	rcu_assign_pointer(*connp, NULL);
+	conn->obj = NULL;
+	conn->type = FSNOTIFY_OBJ_TYPE_DETACHED;
+	if (sb)
+		fsnotify_update_sb_watchers(sb, conn);
+
 	return inode;
 }
 
@@ -203,15 +373,27 @@ static void fsnotify_final_mark_destroy(struct fsnotify_mark *mark)
 	fsnotify_put_group(group);
 }
 
+/* Drop object reference originally held by a connector */
+static void fsnotify_drop_object(unsigned int type, void *objp)
+{
+	if (!objp)
+		return;
+	/* Currently only inode references are passed to be dropped */
+	if (WARN_ON_ONCE(type != FSNOTIFY_OBJ_TYPE_INODE))
+		return;
+	fsnotify_put_inode_ref(objp);
+}
+
 void fsnotify_put_mark(struct fsnotify_mark *mark)
 {
-	struct fsnotify_mark_connector *conn;
-	struct inode *inode = NULL;
+	struct fsnotify_mark_connector *conn = READ_ONCE(mark->connector);
+	void *objp = NULL;
+	unsigned int type = FSNOTIFY_OBJ_TYPE_DETACHED;
 	bool free_conn = false;
 
 	/* Catch marks that were actually never attached to object */
-	if (!mark->connector) {
-		if (atomic_dec_and_test(&mark->refcnt))
+	if (!conn) {
+		if (refcount_dec_and_test(&mark->refcnt))
 			fsnotify_final_mark_destroy(mark);
 		return;
 	}
@@ -220,28 +402,33 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
 	 * We have to be careful so that traversals of obj_list under lock can
 	 * safely grab mark reference.
 	 */
-	if (!atomic_dec_and_lock(&mark->refcnt, &mark->connector->lock))
+	if (!refcount_dec_and_lock(&mark->refcnt, &conn->lock))
 		return;
 
-	conn = mark->connector;
 	hlist_del_init_rcu(&mark->obj_list);
 	if (hlist_empty(&conn->list)) {
-		inode = fsnotify_detach_connector_from_object(conn);
+		objp = fsnotify_detach_connector_from_object(conn, &type);
 		free_conn = true;
 	} else {
-		__fsnotify_recalc_mask(conn);
+		struct super_block *sb = fsnotify_connector_sb(conn);
+
+		/* Update watched objects after detaching mark */
+		if (sb)
+			fsnotify_update_sb_watchers(sb, conn);
+		objp = __fsnotify_recalc_mask(conn);
+		type = conn->type;
 	}
-	mark->connector = NULL;
+	WRITE_ONCE(mark->connector, NULL);
 	spin_unlock(&conn->lock);
 
-	iput(inode);
+	fsnotify_drop_object(type, objp);
 
 	if (free_conn) {
 		spin_lock(&destroy_lock);
 		conn->destroy_next = connector_destroy_list;
 		connector_destroy_list = conn;
 		spin_unlock(&destroy_lock);
-		queue_work(system_unbound_wq, &connector_reaper_work);
+		queue_work(system_dfl_wq, &connector_reaper_work);
 	}
 	/*
 	 * Note that we didn't update flags telling whether inode cares about
@@ -252,36 +439,68 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
 	spin_lock(&destroy_lock);
 	list_add(&mark->g_list, &destroy_list);
 	spin_unlock(&destroy_lock);
-	queue_delayed_work(system_unbound_wq, &reaper_work,
+	queue_delayed_work(system_dfl_wq, &reaper_work,
 			   FSNOTIFY_REAPER_DELAY);
 }
+EXPORT_SYMBOL_GPL(fsnotify_put_mark);
 
-bool fsnotify_prepare_user_wait(struct fsnotify_iter_info *iter_info)
+/*
+ * Get mark reference when we found the mark via lockless traversal of object
+ * list. Mark can be already removed from the list by now and on its way to be
+ * destroyed once SRCU period ends.
+ *
+ * Also pin the group so it doesn't disappear under us.
+ */
+static bool fsnotify_get_mark_safe(struct fsnotify_mark *mark)
 {
-	struct fsnotify_group *group;
+	if (!mark)
+		return true;
+
+	if (refcount_inc_not_zero(&mark->refcnt)) {
+		spin_lock(&mark->lock);
+		if (mark->flags & FSNOTIFY_MARK_FLAG_ATTACHED) {
+			/* mark is attached, group is still alive then */
+			atomic_inc(&mark->group->user_waits);
+			spin_unlock(&mark->lock);
+			return true;
+		}
+		spin_unlock(&mark->lock);
+		fsnotify_put_mark(mark);
+	}
+	return false;
+}
 
-	if (WARN_ON_ONCE(!iter_info->inode_mark && !iter_info->vfsmount_mark))
-		return false;
+/*
+ * Puts marks and wakes up group destruction if necessary.
+ *
+ * Pairs with fsnotify_get_mark_safe()
+ */
+static void fsnotify_put_mark_wake(struct fsnotify_mark *mark)
+{
+	if (mark) {
+		struct fsnotify_group *group = mark->group;
 
-	if (iter_info->inode_mark)
-		group = iter_info->inode_mark->group;
-	else
-		group = iter_info->vfsmount_mark->group;
+		fsnotify_put_mark(mark);
+		/*
+		 * We abuse notification_waitq on group shutdown for waiting for
+		 * all marks pinned when waiting for userspace.
+		 */
+		if (atomic_dec_and_test(&group->user_waits) && group->shutdown)
+			wake_up(&group->notification_waitq);
+	}
+}
 
-	/*
-	 * Since acquisition of mark reference is an atomic op as well, we can
-	 * be sure this inc is seen before any effect of refcount increment.
-	 */
-	atomic_inc(&group->user_waits);
+bool fsnotify_prepare_user_wait(struct fsnotify_iter_info *iter_info)
+	__releases(&fsnotify_mark_srcu)
+{
+	int type;
 
-	if (iter_info->inode_mark) {
+	fsnotify_foreach_iter_type(type) {
 		/* This can fail if mark is being removed */
-		if (!fsnotify_get_mark_safe(iter_info->inode_mark))
-			goto out_wait;
-	}
-	if (iter_info->vfsmount_mark) {
-		if (!fsnotify_get_mark_safe(iter_info->vfsmount_mark))
-			goto out_inode;
+		if (!fsnotify_get_mark_safe(iter_info->marks[type])) {
+			__release(&fsnotify_mark_srcu);
+			goto fail;
+		}
 	}
 
 	/*
@@ -292,34 +511,21 @@ bool fsnotify_prepare_user_wait(struct fsnotify_iter_info *iter_info)
 	srcu_read_unlock(&fsnotify_mark_srcu, iter_info->srcu_idx);
 
 	return true;
-out_inode:
-	if (iter_info->inode_mark)
-		fsnotify_put_mark(iter_info->inode_mark);
-out_wait:
-	if (atomic_dec_and_test(&group->user_waits) && group->shutdown)
-		wake_up(&group->notification_waitq);
+
+fail:
+	for (type--; type >= 0; type--)
+		fsnotify_put_mark_wake(iter_info->marks[type]);
 	return false;
 }
 
 void fsnotify_finish_user_wait(struct fsnotify_iter_info *iter_info)
+	__acquires(&fsnotify_mark_srcu)
 {
-	struct fsnotify_group *group = NULL;
+	int type;
 
 	iter_info->srcu_idx = srcu_read_lock(&fsnotify_mark_srcu);
-	if (iter_info->inode_mark) {
-		group = iter_info->inode_mark->group;
-		fsnotify_put_mark(iter_info->inode_mark);
-	}
-	if (iter_info->vfsmount_mark) {
-		group = iter_info->vfsmount_mark->group;
-		fsnotify_put_mark(iter_info->vfsmount_mark);
-	}
-	/*
-	 * We abuse notification_waitq on group shutdown for waiting for all
-	 * marks pinned when waiting for userspace.
-	 */
-	if (atomic_dec_and_test(&group->user_waits) && group->shutdown)
-		wake_up(&group->notification_waitq);
+	fsnotify_foreach_iter_type(type)
+		fsnotify_put_mark_wake(iter_info->marks[type]);
 }
 
 /*
@@ -334,11 +540,9 @@ void fsnotify_finish_user_wait(struct fsnotify_iter_info *iter_info)
  */
 void fsnotify_detach_mark(struct fsnotify_mark *mark)
 {
-	struct fsnotify_group *group = mark->group;
-
-	WARN_ON_ONCE(!mutex_is_locked(&group->mark_mutex));
+	fsnotify_group_assert_locked(mark->group);
 	WARN_ON_ONCE(!srcu_read_lock_held(&fsnotify_mark_srcu) &&
-		     atomic_read(&mark->refcnt) < 1 +
+		     refcount_read(&mark->refcnt) < 1 +
 			!!(mark->flags & FSNOTIFY_MARK_FLAG_ATTACHED));
 
 	spin_lock(&mark->lock);
@@ -351,8 +555,6 @@ void fsnotify_detach_mark(struct fsnotify_mark *mark)
 	list_del_init(&mark->g_list);
 	spin_unlock(&mark->lock);
 
-	atomic_dec(&group->num_marks);
-
 	/* Drop mark reference acquired in fsnotify_add_mark_locked() */
 	fsnotify_put_mark(mark);
 }
@@ -390,11 +592,12 @@ void fsnotify_free_mark(struct fsnotify_mark *mark)
 void fsnotify_destroy_mark(struct fsnotify_mark *mark,
 			   struct fsnotify_group *group)
 {
-	mutex_lock_nested(&group->mark_mutex, SINGLE_DEPTH_NESTING);
+	fsnotify_group_lock(group);
 	fsnotify_detach_mark(mark);
-	mutex_unlock(&group->mark_mutex);
+	fsnotify_group_unlock(group);
 	fsnotify_free_mark(mark);
 }
+EXPORT_SYMBOL_GPL(fsnotify_destroy_mark);
 
 /*
  * Sorting function for lists of fsnotify marks.
@@ -432,10 +635,28 @@ int fsnotify_compare_groups(struct fsnotify_group *a, struct fsnotify_group *b)
 	return -1;
 }
 
-static int fsnotify_attach_connector_to_object(
-				struct fsnotify_mark_connector __rcu **connp,
-				struct inode *inode,
-				struct vfsmount *mnt)
+static int fsnotify_attach_info_to_sb(struct super_block *sb)
+{
+	struct fsnotify_sb_info *sbinfo;
+
+	/* sb info is freed on fsnotify_sb_delete() */
+	sbinfo = kzalloc(sizeof(*sbinfo), GFP_KERNEL);
+	if (!sbinfo)
+		return -ENOMEM;
+
+	/*
+	 * cmpxchg() provides the barrier so that callers of fsnotify_sb_info()
+	 * will observe an initialized structure
+	 */
+	if (cmpxchg(&sb->s_fsnotify_info, NULL, sbinfo)) {
+		/* Someone else created sbinfo for us */
+		kfree(sbinfo);
+	}
+	return 0;
+}
+
+static int fsnotify_attach_connector_to_object(fsnotify_connp_t *connp,
+					       void *obj, unsigned int obj_type)
 {
 	struct fsnotify_mark_connector *conn;
 
@@ -444,24 +665,19 @@ static int fsnotify_attach_connector_to_object(
 		return -ENOMEM;
 	spin_lock_init(&conn->lock);
 	INIT_HLIST_HEAD(&conn->list);
-	if (inode) {
-		conn->flags = FSNOTIFY_OBJ_TYPE_INODE;
-		conn->inode = igrab(inode);
-	} else {
-		conn->flags = FSNOTIFY_OBJ_TYPE_VFSMOUNT;
-		conn->mnt = mnt;
-	}
+	conn->flags = 0;
+	conn->prio = 0;
+	conn->type = obj_type;
+	conn->obj = obj;
+
 	/*
 	 * cmpxchg() provides the barrier so that readers of *connp can see
 	 * only initialized structure
 	 */
 	if (cmpxchg(connp, NULL, conn)) {
 		/* Someone else created list structure for us */
-		if (inode)
-			iput(inode);
 		kmem_cache_free(fsnotify_mark_connector_cachep, conn);
 	}
-
 	return 0;
 }
 
@@ -472,7 +688,7 @@ static int fsnotify_attach_connector_to_object(
  * they are sure list cannot go away under them.
  */
 static struct fsnotify_mark_connector *fsnotify_grab_connector(
-				struct fsnotify_mark_connector __rcu **connp)
+						fsnotify_connp_t *connp)
 {
 	struct fsnotify_mark_connector *conn;
 	int idx;
@@ -482,8 +698,7 @@ static struct fsnotify_mark_connector *fsnotify_grab_connector(
 	if (!conn)
 		goto out;
 	spin_lock(&conn->lock);
-	if (!(conn->flags & (FSNOTIFY_OBJ_TYPE_INODE |
-			     FSNOTIFY_OBJ_TYPE_VFSMOUNT))) {
+	if (conn->type == FSNOTIFY_OBJ_TYPE_DETACHED) {
 		spin_unlock(&conn->lock);
 		srcu_read_unlock(&fsnotify_mark_srcu, idx);
 		return NULL;
@@ -499,28 +714,36 @@ out:
  * to which group and for which inodes. These marks are ordered according to
  * priority, highest number first, and then by the group's location in memory.
  */
-static int fsnotify_add_mark_list(struct fsnotify_mark *mark,
-				  struct inode *inode, struct vfsmount *mnt,
-				  int allow_dups)
+static int fsnotify_add_mark_list(struct fsnotify_mark *mark, void *obj,
+				  unsigned int obj_type, int add_flags)
 {
+	struct super_block *sb = fsnotify_object_sb(obj, obj_type);
 	struct fsnotify_mark *lmark, *last = NULL;
 	struct fsnotify_mark_connector *conn;
-	struct fsnotify_mark_connector __rcu **connp;
+	fsnotify_connp_t *connp;
 	int cmp;
 	int err = 0;
 
-	if (WARN_ON(!inode && !mnt))
+	if (WARN_ON(!fsnotify_valid_obj_type(obj_type)))
 		return -EINVAL;
-	if (inode)
-		connp = &inode->i_fsnotify_marks;
-	else
-		connp = &real_mount(mnt)->mnt_fsnotify_marks;
+
+	/*
+	 * Attach the sb info before attaching a connector to any object on sb.
+	 * The sb info will remain attached as long as sb lives.
+	 */
+	if (sb && !fsnotify_sb_info(sb)) {
+		err = fsnotify_attach_info_to_sb(sb);
+		if (err)
+			return err;
+	}
+
+	connp = fsnotify_object_connp(obj, obj_type);
 restart:
 	spin_lock(&mark->lock);
 	conn = fsnotify_grab_connector(connp);
 	if (!conn) {
 		spin_unlock(&mark->lock);
-		err = fsnotify_attach_connector_to_object(connp, inode, mnt);
+		err = fsnotify_attach_connector_to_object(connp, obj, obj_type);
 		if (err)
 			return err;
 		goto restart;
@@ -538,7 +761,7 @@ restart:
 
 		if ((lmark->group == mark->group) &&
 		    (lmark->flags & FSNOTIFY_MARK_FLAG_ATTACHED) &&
-		    !allow_dups) {
+		    !(mark->group->flags & FSNOTIFY_GROUP_DUPS)) {
 			err = -EEXIST;
 			goto out_err;
 		}
@@ -554,7 +777,14 @@ restart:
 	/* mark should be the last entry.  last is the current last entry */
 	hlist_add_behind_rcu(&mark->obj_list, &last->obj_list);
 added:
-	mark->connector = conn;
+	if (sb)
+		fsnotify_update_sb_watchers(sb, conn);
+	/*
+	 * Since connector is attached to object using cmpxchg() we are
+	 * guaranteed that connector initialization is fully visible by anyone
+	 * seeing mark->connector set.
+	 */
+	WRITE_ONCE(mark->connector, conn);
 out_err:
 	spin_unlock(&conn->lock);
 	spin_unlock(&mark->lock);
@@ -566,15 +796,14 @@ out_err:
  * These marks may be used for the fsnotify backend to determine which
  * event types should be delivered to which group.
  */
-int fsnotify_add_mark_locked(struct fsnotify_mark *mark, struct inode *inode,
-			     struct vfsmount *mnt, int allow_dups)
+int fsnotify_add_mark_locked(struct fsnotify_mark *mark,
+			     void *obj, unsigned int obj_type,
+			     int add_flags)
 {
 	struct fsnotify_group *group = mark->group;
 	int ret = 0;
 
-	BUG_ON(inode && mnt);
-	BUG_ON(!inode && !mnt);
-	BUG_ON(!mutex_is_locked(&group->mark_mutex));
+	fsnotify_group_assert_locked(group);
 
 	/*
 	 * LOCKING ORDER!!!!
@@ -586,51 +815,54 @@ int fsnotify_add_mark_locked(struct fsnotify_mark *mark, struct inode *inode,
 	mark->flags |= FSNOTIFY_MARK_FLAG_ALIVE | FSNOTIFY_MARK_FLAG_ATTACHED;
 
 	list_add(&mark->g_list, &group->marks_list);
-	atomic_inc(&group->num_marks);
 	fsnotify_get_mark(mark); /* for g_list */
 	spin_unlock(&mark->lock);
 
-	ret = fsnotify_add_mark_list(mark, inode, mnt, allow_dups);
+	ret = fsnotify_add_mark_list(mark, obj, obj_type, add_flags);
 	if (ret)
 		goto err;
 
-	if (mark->mask)
-		fsnotify_recalc_mask(mark->connector);
+	fsnotify_recalc_mask(mark->connector);
 
 	return ret;
 err:
+	spin_lock(&mark->lock);
 	mark->flags &= ~(FSNOTIFY_MARK_FLAG_ALIVE |
 			 FSNOTIFY_MARK_FLAG_ATTACHED);
 	list_del_init(&mark->g_list);
-	atomic_dec(&group->num_marks);
+	spin_unlock(&mark->lock);
 
 	fsnotify_put_mark(mark);
 	return ret;
 }
 
-int fsnotify_add_mark(struct fsnotify_mark *mark, struct inode *inode,
-		      struct vfsmount *mnt, int allow_dups)
+int fsnotify_add_mark(struct fsnotify_mark *mark, void *obj,
+		      unsigned int obj_type, int add_flags)
 {
 	int ret;
 	struct fsnotify_group *group = mark->group;
 
-	mutex_lock(&group->mark_mutex);
-	ret = fsnotify_add_mark_locked(mark, inode, mnt, allow_dups);
-	mutex_unlock(&group->mark_mutex);
+	fsnotify_group_lock(group);
+	ret = fsnotify_add_mark_locked(mark, obj, obj_type, add_flags);
+	fsnotify_group_unlock(group);
 	return ret;
 }
+EXPORT_SYMBOL_GPL(fsnotify_add_mark);
 
 /*
  * Given a list of marks, find the mark associated with given group. If found
  * take a reference to that mark and return it, else return NULL.
  */
-struct fsnotify_mark *fsnotify_find_mark(
-				struct fsnotify_mark_connector __rcu **connp,
-				struct fsnotify_group *group)
+struct fsnotify_mark *fsnotify_find_mark(void *obj, unsigned int obj_type,
+					 struct fsnotify_group *group)
 {
+	fsnotify_connp_t *connp = fsnotify_object_connp(obj, obj_type);
 	struct fsnotify_mark_connector *conn;
 	struct fsnotify_mark *mark;
 
+	if (!connp)
+		return NULL;
+
 	conn = fsnotify_grab_connector(connp);
 	if (!conn)
 		return NULL;
@@ -646,17 +878,18 @@ struct fsnotify_mark *fsnotify_find_mark(
 	spin_unlock(&conn->lock);
 	return NULL;
 }
+EXPORT_SYMBOL_GPL(fsnotify_find_mark);
 
-/* Clear any marks in a group with given type */
+/* Clear any marks in a group with given type mask */
 void fsnotify_clear_marks_by_group(struct fsnotify_group *group,
-				   unsigned int type)
+				   unsigned int obj_type)
 {
 	struct fsnotify_mark *lmark, *mark;
 	LIST_HEAD(to_free);
 	struct list_head *head = &to_free;
 
 	/* Skip selection step if we want to clear all marks. */
-	if (type == FSNOTIFY_OBJ_ALL_TYPES) {
+	if (obj_type == FSNOTIFY_OBJ_TYPE_ANY) {
 		head = &group->marks_list;
 		goto clear;
 	}
@@ -669,35 +902,36 @@ void fsnotify_clear_marks_by_group(struct fsnotify_group *group,
 	 * move marks to free to to_free list in one go and then free marks in
 	 * to_free list one by one.
 	 */
-	mutex_lock_nested(&group->mark_mutex, SINGLE_DEPTH_NESTING);
+	fsnotify_group_lock(group);
 	list_for_each_entry_safe(mark, lmark, &group->marks_list, g_list) {
-		if (mark->connector->flags & type)
+		if (mark->connector->type == obj_type)
 			list_move(&mark->g_list, &to_free);
 	}
-	mutex_unlock(&group->mark_mutex);
+	fsnotify_group_unlock(group);
 
 clear:
 	while (1) {
-		mutex_lock_nested(&group->mark_mutex, SINGLE_DEPTH_NESTING);
+		fsnotify_group_lock(group);
 		if (list_empty(head)) {
-			mutex_unlock(&group->mark_mutex);
+			fsnotify_group_unlock(group);
 			break;
 		}
 		mark = list_first_entry(head, struct fsnotify_mark, g_list);
 		fsnotify_get_mark(mark);
 		fsnotify_detach_mark(mark);
-		mutex_unlock(&group->mark_mutex);
+		fsnotify_group_unlock(group);
 		fsnotify_free_mark(mark);
 		fsnotify_put_mark(mark);
 	}
 }
 
-/* Destroy all marks attached to inode / vfsmount */
-void fsnotify_destroy_marks(struct fsnotify_mark_connector __rcu **connp)
+/* Destroy all marks attached to an object via connector */
+void fsnotify_destroy_marks(fsnotify_connp_t *connp)
 {
 	struct fsnotify_mark_connector *conn;
 	struct fsnotify_mark *mark, *old_mark = NULL;
-	struct inode *inode;
+	void *objp;
+	unsigned int type;
 
 	conn = fsnotify_grab_connector(connp);
 	if (!conn)
@@ -723,11 +957,11 @@ void fsnotify_destroy_marks(struct fsnotify_mark_connector __rcu **connp)
 	 * mark references get dropped. It would lead to strange results such
 	 * as delaying inode deletion or blocking unmount.
 	 */
-	inode = fsnotify_detach_connector_from_object(conn);
+	objp = fsnotify_detach_connector_from_object(conn, &type);
 	spin_unlock(&conn->lock);
 	if (old_mark)
 		fsnotify_put_mark(old_mark);
-	iput(inode);
+	fsnotify_drop_object(type, objp);
 }
 
 /*
@@ -738,10 +972,12 @@ void fsnotify_init_mark(struct fsnotify_mark *mark,
 {
 	memset(mark, 0, sizeof(*mark));
 	spin_lock_init(&mark->lock);
-	atomic_set(&mark->refcnt, 1);
+	refcount_set(&mark->refcnt, 1);
 	fsnotify_get_group(group);
 	mark->group = group;
+	WRITE_ONCE(mark->connector, NULL);
 }
+EXPORT_SYMBOL_GPL(fsnotify_init_mark);
 
 /*
  * Destroy all marks in destroy_list, waits for SRCU period to finish before
@@ -770,3 +1006,4 @@ void fsnotify_wait_marks_destroyed(void)
 {
 	flush_delayed_work(&reaper_work);
 }
+EXPORT_SYMBOL_GPL(fsnotify_wait_marks_destroyed);