Merge branch 'fsnotify' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs

Pull fsnotify updates from Jan Kara:

 - fixes of use-after-tree issues when handling fanotify permission
   events from Miklos

 - refcount_t conversions from Elena

 - fixes of ENOMEM handling in dnotify and fsnotify from me

* 'fsnotify' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  fsnotify: convert fsnotify_mark.refcnt from atomic_t to refcount_t
  fanotify: clean up CONFIG_FANOTIFY_ACCESS_PERMISSIONS ifdefs
  fsnotify: clean up fsnotify()
  fanotify: fix fsnotify_prepare_user_wait() failure
  fsnotify: fix pinning group in fsnotify_prepare_user_wait()
  fsnotify: pin both inode and vfsmount mark
  fsnotify: clean up fsnotify_prepare/finish_user_wait()
  fsnotify: convert fsnotify_group.refcnt from atomic_t to refcount_t
  fsnotify: Protect bail out path of fsnotify_add_mark_locked() properly
  dnotify: Handle errors from fsnotify_add_mark_locked() in fcntl_dirnotify()

fs/notify/dnotify/dnotify.c

@@ -319,7 +319,11 @@ int fcntl_dirnotify(int fd, struct file *filp, unsigned long arg)
 		dn_mark = container_of(fsn_mark, struct dnotify_mark, fsn_mark);
 		spin_lock(&fsn_mark->lock);
 	} else {
-		fsnotify_add_mark_locked(new_fsn_mark, inode, NULL, 0);
+		error = fsnotify_add_mark_locked(new_fsn_mark, inode, NULL, 0);
+		if (error) {
+			mutex_unlock(&dnotify_group->mark_mutex);
+			goto out_err;
+		}
 		spin_lock(&new_fsn_mark->lock);
 		fsn_mark = new_fsn_mark;
 		dn_mark = new_dn_mark;
@@ -345,6 +349,7 @@ int fcntl_dirnotify(int fd, struct file *filp, unsigned long arg)
 		 */
 		if (dn_mark == new_dn_mark)
 			destroy = 1;
+		error = 0;
 		goto out;
 	}
 

fs/notify/fanotify/fanotify.c

@@ -36,15 +36,13 @@ static int fanotify_merge(struct list_head *list, struct fsnotify_event *event)
 
 	pr_debug("%s: list=%p event=%p\n", __func__, list, event);
 
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 	/*
 	 * Don't merge a permission event with any other event so that we know
 	 * the event structure we have created in fanotify_handle_event() is the
 	 * one we should check for permission response.
 	 */
-	if (event->mask & FAN_ALL_PERM_EVENTS)
+	if (fanotify_is_perm_event(event->mask))
 		return 0;
-#endif
 
 	list_for_each_entry_reverse(test_event, list, list) {
 		if (should_merge(test_event, event)) {
@@ -56,7 +54,6 @@ static int fanotify_merge(struct list_head *list, struct fsnotify_event *event)
 	return 0;
 }
 
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 static int fanotify_get_response(struct fsnotify_group *group,
 				 struct fanotify_perm_event_info *event,
 				 struct fsnotify_iter_info *iter_info)
@@ -65,19 +62,8 @@ static int fanotify_get_response(struct fsnotify_group *group,
 
 	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
 
-	/*
-	 * fsnotify_prepare_user_wait() fails if we race with mark deletion.
-	 * Just let the operation pass in that case.
-	 */
-	if (!fsnotify_prepare_user_wait(iter_info)) {
-		event->response = FAN_ALLOW;
-		goto out;
-	}
-
 	wait_event(group->fanotify_data.access_waitq, event->response);
 
-	fsnotify_finish_user_wait(iter_info);
-out:
 	/* userspace responded, convert to something usable */
 	switch (event->response) {
 	case FAN_ALLOW:
@@ -94,7 +80,6 @@ static int fanotify_get_response(struct fsnotify_group *group,
 	
 	return ret;
 }
-#endif
 
 static bool fanotify_should_send_event(struct fsnotify_mark *inode_mark,
 				       struct fsnotify_mark *vfsmnt_mark,
@@ -153,8 +138,7 @@ struct fanotify_event_info *fanotify_alloc_event(struct inode *inode, u32 mask,
 {
 	struct fanotify_event_info *event;
 
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
-	if (mask & FAN_ALL_PERM_EVENTS) {
+	if (fanotify_is_perm_event(mask)) {
 		struct fanotify_perm_event_info *pevent;
 
 		pevent = kmem_cache_alloc(fanotify_perm_event_cachep,
@@ -165,7 +149,6 @@ struct fanotify_event_info *fanotify_alloc_event(struct inode *inode, u32 mask,
 		pevent->response = 0;
 		goto init;
 	}
-#endif
 	event = kmem_cache_alloc(fanotify_event_cachep, GFP_KERNEL);
 	if (!event)
 		return NULL;
@@ -212,9 +195,19 @@ static int fanotify_handle_event(struct fsnotify_group *group,
 	pr_debug("%s: group=%p inode=%p mask=%x\n", __func__, group, inode,
 		 mask);
 
+	if (fanotify_is_perm_event(mask)) {
+		/*
+		 * fsnotify_prepare_user_wait() fails if we race with mark
+		 * deletion.  Just let the operation pass in that case.
+		 */
+		if (!fsnotify_prepare_user_wait(iter_info))
+			return 0;
+	}
+
 	event = fanotify_alloc_event(inode, mask, data);
+	ret = -ENOMEM;
 	if (unlikely(!event))
-		return -ENOMEM;
+		goto finish;
 
 	fsn_event = &event->fse;
 	ret = fsnotify_add_event(group, fsn_event, fanotify_merge);
@@ -224,16 +217,16 @@ static int fanotify_handle_event(struct fsnotify_group *group,
 		/* Our event wasn't used in the end. Free it. */
 		fsnotify_destroy_event(group, fsn_event);
 
-		return 0;
-	}
-
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
-	if (mask & FAN_ALL_PERM_EVENTS) {
+		ret = 0;
+	} else if (fanotify_is_perm_event(mask)) {
 		ret = fanotify_get_response(group, FANOTIFY_PE(fsn_event),
 					    iter_info);
 		fsnotify_destroy_event(group, fsn_event);
 	}
-#endif
+finish:
+	if (fanotify_is_perm_event(mask))
+		fsnotify_finish_user_wait(iter_info);
+
 	return ret;
 }
 
@@ -253,13 +246,11 @@ static void fanotify_free_event(struct fsnotify_event *fsn_event)
 	event = FANOTIFY_E(fsn_event);
 	path_put(&event->path);
 	put_pid(event->tgid);
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
-	if (fsn_event->mask & FAN_ALL_PERM_EVENTS) {
+	if (fanotify_is_perm_event(fsn_event->mask)) {
 		kmem_cache_free(fanotify_perm_event_cachep,
 				FANOTIFY_PE(fsn_event));
 		return;
 	}
-#endif
 	kmem_cache_free(fanotify_event_cachep, event);
 }
 

fs/notify/fanotify/fanotify.h

@@ -22,7 +22,6 @@ struct fanotify_event_info {
 	struct pid *tgid;
 };
 
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 /*
  * Structure for permission fanotify events. It gets allocated and freed in
  * fanotify_handle_event() since we wait there for user response. When the
@@ -41,7 +40,12 @@ FANOTIFY_PE(struct fsnotify_event *fse)
 {
 	return container_of(fse, struct fanotify_perm_event_info, fae.fse);
 }
-#endif
+
+static inline bool fanotify_is_perm_event(u32 mask)
+{
+	return IS_ENABLED(CONFIG_FANOTIFY_ACCESS_PERMISSIONS) &&
+		mask & FAN_ALL_PERM_EVENTS;
+}
 
 static inline struct fanotify_event_info *FANOTIFY_E(struct fsnotify_event *fse)
 {

fs/notify/fanotify/fanotify_user.c

@@ -143,7 +143,6 @@ static int fill_event_metadata(struct fsnotify_group *group,
 	return ret;
 }
 
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 static struct fanotify_perm_event_info *dequeue_event(
 				struct fsnotify_group *group, int fd)
 {
@@ -200,7 +199,6 @@ static int process_access_response(struct fsnotify_group *group,
 
 	return 0;
 }
-#endif
 
 static ssize_t copy_event_to_user(struct fsnotify_group *group,
 				  struct fsnotify_event *event,
@@ -222,10 +220,8 @@ static ssize_t copy_event_to_user(struct fsnotify_group *group,
 			 fanotify_event_metadata.event_len))
 		goto out_close_fd;
 
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
-	if (event->mask & FAN_ALL_PERM_EVENTS)
+	if (fanotify_is_perm_event(event->mask))
 		FANOTIFY_PE(event)->fd = fd;
-#endif
 
 	if (fd != FAN_NOFD)
 		fd_install(fd, f);
@@ -310,10 +306,9 @@ static ssize_t fanotify_read(struct file *file, char __user *buf,
 		 * Permission events get queued to wait for response.  Other
 		 * events can be destroyed now.
 		 */
-		if (!(kevent->mask & FAN_ALL_PERM_EVENTS)) {
+		if (!fanotify_is_perm_event(kevent->mask)) {
 			fsnotify_destroy_event(group, kevent);
 		} else {
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 			if (ret <= 0) {
 				FANOTIFY_PE(kevent)->response = FAN_DENY;
 				wake_up(&group->fanotify_data.access_waitq);
@@ -323,7 +318,6 @@ static ssize_t fanotify_read(struct file *file, char __user *buf,
 					&group->fanotify_data.access_list);
 				spin_unlock(&group->notification_lock);
 			}
-#endif
 		}
 		if (ret < 0)
 			break;
@@ -339,11 +333,13 @@ static ssize_t fanotify_read(struct file *file, char __user *buf,
 
 static ssize_t fanotify_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
 {
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 	struct fanotify_response response = { .fd = -1, .response = -1 };
 	struct fsnotify_group *group;
 	int ret;
 
+	if (!IS_ENABLED(CONFIG_FANOTIFY_ACCESS_PERMISSIONS))
+		return -EINVAL;
+
 	group = file->private_data;
 
 	if (count > sizeof(response))
@@ -359,16 +355,11 @@ static ssize_t fanotify_write(struct file *file, const char __user *buf, size_t
 		count = ret;
 
 	return count;
-#else
-	return -EINVAL;
-#endif
 }
 
 static int fanotify_release(struct inode *ignored, struct file *file)
 {
 	struct fsnotify_group *group = file->private_data;
-
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 	struct fanotify_perm_event_info *event, *next;
 	struct fsnotify_event *fsn_event;
 
@@ -404,14 +395,14 @@ static int fanotify_release(struct inode *ignored, struct file *file)
 			spin_unlock(&group->notification_lock);
 			fsnotify_destroy_event(group, fsn_event);
 			spin_lock(&group->notification_lock);
-		} else
+		} else {
 			FANOTIFY_PE(fsn_event)->response = FAN_ALLOW;
+		}
 	}
 	spin_unlock(&group->notification_lock);
 
 	/* Response for all permission events it set, wakeup waiters */
 	wake_up(&group->fanotify_data.access_waitq);
-#endif
 
 	/* matches the fanotify_init->fsnotify_alloc_group */
 	fsnotify_destroy_group(group);
@@ -769,10 +760,8 @@ SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
 	if (force_o_largefile())
 		event_f_flags |= O_LARGEFILE;
 	group->fanotify_data.f_flags = event_f_flags;
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 	init_waitqueue_head(&group->fanotify_data.access_waitq);
 	INIT_LIST_HEAD(&group->fanotify_data.access_list);
-#endif
 	switch (flags & FAN_ALL_CLASS_BITS) {
 	case FAN_CLASS_NOTIF:
 		group->priority = FS_PRIO_0;
@@ -826,6 +815,7 @@ SYSCALL_DEFINE5(fanotify_mark, int, fanotify_fd, unsigned int, flags,
 	struct fsnotify_group *group;
 	struct fd f;
 	struct path path;
+	u32 valid_mask = FAN_ALL_EVENTS | FAN_EVENT_ON_CHILD;
 	int ret;
 
 	pr_debug("%s: fanotify_fd=%d flags=%x dfd=%d pathname=%p mask=%llx\n",
@@ -856,11 +846,10 @@ SYSCALL_DEFINE5(fanotify_mark, int, fanotify_fd, unsigned int, flags,
 		mask &= ~FAN_ONDIR;
 	}
 
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
-	if (mask & ~(FAN_ALL_EVENTS | FAN_ALL_PERM_EVENTS | FAN_EVENT_ON_CHILD))
-#else
-	if (mask & ~(FAN_ALL_EVENTS | FAN_EVENT_ON_CHILD))
-#endif
+	if (IS_ENABLED(CONFIG_FANOTIFY_ACCESS_PERMISSIONS))
+		valid_mask |= FAN_ALL_PERM_EVENTS;
+
+	if (mask & ~valid_mask)
 		return -EINVAL;
 
 	f = fdget(fanotify_fd);
@@ -950,10 +939,10 @@ static int __init fanotify_user_setup(void)
 {
 	fanotify_mark_cache = KMEM_CACHE(fsnotify_mark, SLAB_PANIC);
 	fanotify_event_cachep = KMEM_CACHE(fanotify_event_info, SLAB_PANIC);
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
-	fanotify_perm_event_cachep = KMEM_CACHE(fanotify_perm_event_info,
-						SLAB_PANIC);
-#endif
+	if (IS_ENABLED(CONFIG_FANOTIFY_ACCESS_PERMISSIONS)) {
+		fanotify_perm_event_cachep =
+			KMEM_CACHE(fanotify_perm_event_info, SLAB_PANIC);
+	}
 
 	return 0;
 }

fs/notify/fsnotify.c

@@ -243,6 +243,29 @@ static int send_to_group(struct inode *to_tell,
 					file_name, cookie, iter_info);
 }
 
+static struct fsnotify_mark *fsnotify_first_mark(struct fsnotify_mark_connector **connp)
+{
+	struct fsnotify_mark_connector *conn;
+	struct hlist_node *node = NULL;
+
+	conn = srcu_dereference(*connp, &fsnotify_mark_srcu);
+	if (conn)
+		node = srcu_dereference(conn->list.first, &fsnotify_mark_srcu);
+
+	return hlist_entry_safe(node, struct fsnotify_mark, obj_list);
+}
+
+static struct fsnotify_mark *fsnotify_next_mark(struct fsnotify_mark *mark)
+{
+	struct hlist_node *node = NULL;
+
+	if (mark)
+		node = srcu_dereference(mark->obj_list.next,
+					&fsnotify_mark_srcu);
+
+	return hlist_entry_safe(node, struct fsnotify_mark, obj_list);
+}
+
 /*
  * This is the main call to fsnotify.  The VFS calls into hook specific functions
  * in linux/fsnotify.h.  Those functions then in turn call here.  Here will call
@@ -252,11 +275,7 @@ static int send_to_group(struct inode *to_tell,
 int fsnotify(struct inode *to_tell, __u32 mask, const void *data, int data_is,
 	     const unsigned char *file_name, u32 cookie)
 {
-	struct hlist_node *inode_node = NULL, *vfsmount_node = NULL;
-	struct fsnotify_mark *inode_mark = NULL, *vfsmount_mark = NULL;
-	struct fsnotify_group *inode_group, *vfsmount_group;
-	struct fsnotify_mark_connector *inode_conn, *vfsmount_conn;
-	struct fsnotify_iter_info iter_info;
+	struct fsnotify_iter_info iter_info = {};
 	struct mount *mnt;
 	int ret = 0;
 	/* global tests shouldn't care about events on child only the specific event */
@@ -291,26 +310,16 @@ int fsnotify(struct inode *to_tell, __u32 mask, const void *data, int data_is,
 
 	if ((mask & FS_MODIFY) ||
 	    (test_mask & to_tell->i_fsnotify_mask)) {
-		inode_conn = srcu_dereference(to_tell->i_fsnotify_marks,
-					      &fsnotify_mark_srcu);
-		if (inode_conn)
-			inode_node = srcu_dereference(inode_conn->list.first,
-						      &fsnotify_mark_srcu);
+		iter_info.inode_mark =
+			fsnotify_first_mark(&to_tell->i_fsnotify_marks);
 	}
 
 	if (mnt && ((mask & FS_MODIFY) ||
 		    (test_mask & mnt->mnt_fsnotify_mask))) {
-		inode_conn = srcu_dereference(to_tell->i_fsnotify_marks,
-					      &fsnotify_mark_srcu);
-		if (inode_conn)
-			inode_node = srcu_dereference(inode_conn->list.first,
-						      &fsnotify_mark_srcu);
-		vfsmount_conn = srcu_dereference(mnt->mnt_fsnotify_marks,
-					         &fsnotify_mark_srcu);
-		if (vfsmount_conn)
-			vfsmount_node = srcu_dereference(
-						vfsmount_conn->list.first,
-						&fsnotify_mark_srcu);
+		iter_info.inode_mark =
+			fsnotify_first_mark(&to_tell->i_fsnotify_marks);
+		iter_info.vfsmount_mark =
+			fsnotify_first_mark(&mnt->mnt_fsnotify_marks);
 	}
 
 	/*
@@ -318,39 +327,19 @@ int fsnotify(struct inode *to_tell, __u32 mask, const void *data, int data_is,
 	 * ignore masks are properly reflected for mount mark notifications.
 	 * That's why this traversal is so complicated...
 	 */
-	while (inode_node || vfsmount_node) {
-		inode_group = NULL;
-		inode_mark = NULL;
-		vfsmount_group = NULL;
-		vfsmount_mark = NULL;
-
-		if (inode_node) {
-			inode_mark = hlist_entry(srcu_dereference(inode_node, &fsnotify_mark_srcu),
-						 struct fsnotify_mark, obj_list);
-			inode_group = inode_mark->group;
-		}
-
-		if (vfsmount_node) {
-			vfsmount_mark = hlist_entry(srcu_dereference(vfsmount_node, &fsnotify_mark_srcu),
-						    struct fsnotify_mark, obj_list);
-			vfsmount_group = vfsmount_mark->group;
-		}
-
-		if (inode_group && vfsmount_group) {
-			int cmp = fsnotify_compare_groups(inode_group,
-							  vfsmount_group);
-			if (cmp > 0) {
-				inode_group = NULL;
+	while (iter_info.inode_mark || iter_info.vfsmount_mark) {
+		struct fsnotify_mark *inode_mark = iter_info.inode_mark;
+		struct fsnotify_mark *vfsmount_mark = iter_info.vfsmount_mark;
+
+		if (inode_mark && vfsmount_mark) {
+			int cmp = fsnotify_compare_groups(inode_mark->group,
+							  vfsmount_mark->group);
+			if (cmp > 0)
 				inode_mark = NULL;
-			} else if (cmp < 0) {
-				vfsmount_group = NULL;
+			else if (cmp < 0)
 				vfsmount_mark = NULL;
-			}
 		}
 
-		iter_info.inode_mark = inode_mark;
-		iter_info.vfsmount_mark = vfsmount_mark;
-
 		ret = send_to_group(to_tell, inode_mark, vfsmount_mark, mask,
 				    data, data_is, cookie, file_name,
 				    &iter_info);
@@ -358,12 +347,12 @@ int fsnotify(struct inode *to_tell, __u32 mask, const void *data, int data_is,
 		if (ret && (mask & ALL_FSNOTIFY_PERM_EVENTS))
 			goto out;
 
-		if (inode_group)
-			inode_node = srcu_dereference(inode_node->next,
-						      &fsnotify_mark_srcu);
-		if (vfsmount_group)
-			vfsmount_node = srcu_dereference(vfsmount_node->next,
-							 &fsnotify_mark_srcu);
+		if (inode_mark)
+			iter_info.inode_mark =
+				fsnotify_next_mark(iter_info.inode_mark);
+		if (vfsmount_mark)
+			iter_info.vfsmount_mark =
+				fsnotify_next_mark(iter_info.vfsmount_mark);
 	}
 	ret = 0;
 out:

fs/notify/group.c

@@ -107,7 +107,7 @@ void fsnotify_destroy_group(struct fsnotify_group *group)
  */
 void fsnotify_get_group(struct fsnotify_group *group)
 {
-	atomic_inc(&group->refcnt);
+	refcount_inc(&group->refcnt);
 }
 
 /*
@@ -115,7 +115,7 @@ void fsnotify_get_group(struct fsnotify_group *group)
  */
 void fsnotify_put_group(struct fsnotify_group *group)
 {
-	if (atomic_dec_and_test(&group->refcnt))
+	if (refcount_dec_and_test(&group->refcnt))
 		fsnotify_final_destroy_group(group);
 }
 
@@ -131,7 +131,7 @@ struct fsnotify_group *fsnotify_alloc_group(const struct fsnotify_ops *ops)
 		return ERR_PTR(-ENOMEM);
 
 	/* set to 0 when there a no external references to this group */
-	atomic_set(&group->refcnt, 1);
+	refcount_set(&group->refcnt, 1);
 	atomic_set(&group->num_marks, 0);
 	atomic_set(&group->user_waits, 0);
 

fs/notify/inotify/inotify_user.c

@@ -376,7 +376,7 @@ static struct inotify_inode_mark *inotify_idr_find_locked(struct fsnotify_group
 
 		fsnotify_get_mark(fsn_mark);
 		/* One ref for being in the idr, one ref we just took */
-		BUG_ON(atomic_read(&fsn_mark->refcnt) < 2);
+		BUG_ON(refcount_read(&fsn_mark->refcnt) < 2);
 	}
 
 	return i_mark;
@@ -446,7 +446,7 @@ static void inotify_remove_from_idr(struct fsnotify_group *group,
 	 * One ref for being in the idr
 	 * one ref grabbed by inotify_idr_find
 	 */
-	if (unlikely(atomic_read(&i_mark->fsn_mark.refcnt) < 2)) {
+	if (unlikely(refcount_read(&i_mark->fsn_mark.refcnt) < 2)) {
 		printk(KERN_ERR "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p\n",
 			 __func__, i_mark, i_mark->wd, i_mark->fsn_mark.group);
 		/* we can't really recover with bad ref cnting.. */

fs/notify/mark.c

@@ -105,18 +105,8 @@ 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);
-}
-
-/*
- * 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.
- */
-static bool fsnotify_get_mark_safe(struct fsnotify_mark *mark)
-{
-	return atomic_inc_not_zero(&mark->refcnt);
+	WARN_ON_ONCE(!refcount_read(&mark->refcnt));
+	refcount_inc(&mark->refcnt);
 }
 
 static void __fsnotify_recalc_mask(struct fsnotify_mark_connector *conn)
@@ -211,7 +201,7 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
 
 	/* Catch marks that were actually never attached to object */
 	if (!mark->connector) {
-		if (atomic_dec_and_test(&mark->refcnt))
+		if (refcount_dec_and_test(&mark->refcnt))
 			fsnotify_final_mark_destroy(mark);
 		return;
 	}
@@ -220,7 +210,7 @@ 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, &mark->connector->lock))
 		return;
 
 	conn = mark->connector;
@@ -256,32 +246,60 @@ void fsnotify_put_mark(struct fsnotify_mark *mark)
 			   FSNOTIFY_REAPER_DELAY);
 }
 
-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 (WARN_ON_ONCE(!iter_info->inode_mark && !iter_info->vfsmount_mark))
-		return false;
-
-	if (iter_info->inode_mark)
-		group = iter_info->inode_mark->group;
-	else
-		group = iter_info->vfsmount_mark->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;
+}
 
-	/*
-	 * 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);
+/*
+ * 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) {
-		/* This can fail if mark is being removed */
-		if (!fsnotify_get_mark_safe(iter_info->inode_mark))
-			goto out_wait;
+		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);
 	}
-	if (iter_info->vfsmount_mark) {
-		if (!fsnotify_get_mark_safe(iter_info->vfsmount_mark))
-			goto out_inode;
+}
+
+bool fsnotify_prepare_user_wait(struct fsnotify_iter_info *iter_info)
+{
+	/* This can fail if mark is being removed */
+	if (!fsnotify_get_mark_safe(iter_info->inode_mark))
+		return false;
+	if (!fsnotify_get_mark_safe(iter_info->vfsmount_mark)) {
+		fsnotify_put_mark_wake(iter_info->inode_mark);
+		return false;
 	}
 
 	/*
@@ -292,34 +310,13 @@ 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);
-	return false;
 }
 
 void fsnotify_finish_user_wait(struct fsnotify_iter_info *iter_info)
 {
-	struct fsnotify_group *group = NULL;
-
 	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_put_mark_wake(iter_info->inode_mark);
+	fsnotify_put_mark_wake(iter_info->vfsmount_mark);
 }
 
 /*
@@ -338,7 +335,7 @@ void fsnotify_detach_mark(struct fsnotify_mark *mark)
 
 	WARN_ON_ONCE(!mutex_is_locked(&group->mark_mutex));
 	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);
@@ -599,9 +596,11 @@ int fsnotify_add_mark_locked(struct fsnotify_mark *mark, struct inode *inode,
 
 	return ret;
 err:
+	spin_lock(&mark->lock);
 	mark->flags &= ~(FSNOTIFY_MARK_FLAG_ALIVE |
 			 FSNOTIFY_MARK_FLAG_ATTACHED);
 	list_del_init(&mark->g_list);
+	spin_unlock(&mark->lock);
 	atomic_dec(&group->num_marks);
 
 	fsnotify_put_mark(mark);
@@ -738,7 +737,7 @@ 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;
 }

include/linux/fsnotify_backend.h

@@ -18,6 +18,7 @@
 #include <linux/types.h>
 #include <linux/atomic.h>
 #include <linux/user_namespace.h>
+#include <linux/refcount.h>
 
 /*
  * IN_* from inotfy.h lines up EXACTLY with FS_*, this is so we can easily
@@ -136,7 +137,7 @@ struct fsnotify_group {
 	 * inotify_init() and the refcnt will hit 0 only when that fd has been
 	 * closed.
 	 */
-	atomic_t refcnt;		/* things with interest in this group */
+	refcount_t refcnt;		/* things with interest in this group */
 
 	const struct fsnotify_ops *ops;	/* how this group handles things */
 
@@ -183,11 +184,9 @@ struct fsnotify_group {
 #endif
 #ifdef CONFIG_FANOTIFY
 		struct fanotify_group_private_data {
-#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
 			/* allows a group to block waiting for a userspace response */
 			struct list_head access_list;
 			wait_queue_head_t access_waitq;
-#endif /* CONFIG_FANOTIFY_ACCESS_PERMISSIONS */
 			int f_flags;
 			unsigned int max_marks;
 			struct user_struct *user;
@@ -244,7 +243,7 @@ struct fsnotify_mark {
 	__u32 mask;
 	/* We hold one for presence in g_list. Also one ref for each 'thing'
 	 * in kernel that found and may be using this mark. */
-	atomic_t refcnt;
+	refcount_t refcnt;
 	/* Group this mark is for. Set on mark creation, stable until last ref
 	 * is dropped */
 	struct fsnotify_group *group;

kernel/audit_tree.c

@@ -1008,7 +1008,7 @@ static void audit_tree_freeing_mark(struct fsnotify_mark *entry, struct fsnotify
 	 * We are guaranteed to have at least one reference to the mark from
 	 * either the inode or the caller of fsnotify_destroy_mark().
 	 */
-	BUG_ON(atomic_read(&entry->refcnt) < 1);
+	BUG_ON(refcount_read(&entry->refcnt) < 1);
 }
 
 static const struct fsnotify_ops audit_tree_ops = {