Merge commit 'fsnotify/for-next'

12 files changed, 197 insertions, 157 deletions

diff --git a/fs/compat.c b/fs/compat.c
index 94502dab972a..1465b801afde 100644
--- a/fs/compat.c
+++ b/fs/compat.c
@@ -1177,11 +1177,10 @@ out:
 	if (iov != iovstack)
 		kfree(iov);
 	if ((ret + (type == READ)) > 0) {
-		struct dentry *dentry = file->f_path.dentry;
 		if (type == READ)
-			fsnotify_access(dentry);
+			fsnotify_access(file);
 		else
-			fsnotify_modify(dentry);
+			fsnotify_modify(file);
 	}
 	return ret;
 }
diff --git a/fs/exec.c b/fs/exec.c
index fb4f3cdda78c..9f974080712c 100644
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -128,7 +128,7 @@ SYSCALL_DEFINE1(uselib, const char __user *, library)
 	if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
 		goto exit;
 
-	fsnotify_open(file->f_path.dentry);
+	fsnotify_open(file);
 
 	error = -ENOEXEC;
 	if(file->f_op) {
@@ -663,7 +663,7 @@ struct file *open_exec(const char *name)
 	if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
 		goto exit;
 
-	fsnotify_open(file->f_path.dentry);
+	fsnotify_open(file);
 
 	err = deny_write_access(file);
 	if (err)
diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c
index 23341c1063bc..5c8f67eb049d 100644
--- a/fs/nfsd/vfs.c
+++ b/fs/nfsd/vfs.c
@@ -948,7 +948,7 @@ nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
 		nfsdstats.io_read += host_err;
 		*count = host_err;
 		err = 0;
-		fsnotify_access(file->f_path.dentry);
+		fsnotify_access(file);
 	} else 
 		err = nfserrno(host_err);
 out:
@@ -1059,7 +1059,7 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
 		goto out_nfserr;
 	*cnt = host_err;
 	nfsdstats.io_write += host_err;
-	fsnotify_modify(file->f_path.dentry);
+	fsnotify_modify(file);
 
 	/* clear setuid/setgid flag after write */
 	if (inode->i_mode & (S_ISUID | S_ISGID))
diff --git a/fs/notify/dnotify/dnotify.c b/fs/notify/dnotify/dnotify.c
index 828a889be909..8c568a09138e 100644
--- a/fs/notify/dnotify/dnotify.c
+++ b/fs/notify/dnotify/dnotify.c
@@ -132,7 +132,8 @@ static int dnotify_handle_event(struct fsnotify_group *group,
  * userspace notification for that pair.
  */
 static bool dnotify_should_send_event(struct fsnotify_group *group,
-				      struct inode *inode, __u32 mask)
+				      struct inode *inode, __u32 mask,
+				      void *data, int data_type)
 {
 	struct fsnotify_mark_entry *entry;
 	bool send;
@@ -361,7 +362,7 @@ int fcntl_dirnotify(int fd, struct file *filp, unsigned long arg)
 		dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry);
 		spin_lock(&entry->lock);
 	} else {
-		fsnotify_add_mark(new_entry, dnotify_group, inode);
+		fsnotify_add_mark(new_entry, dnotify_group, inode, 0);
 		spin_lock(&new_entry->lock);
 		entry = new_entry;
 		dnentry = new_dnentry;
@@ -431,8 +432,7 @@ static int __init dnotify_init(void)
 	dnotify_struct_cache = KMEM_CACHE(dnotify_struct, SLAB_PANIC);
 	dnotify_mark_entry_cache = KMEM_CACHE(dnotify_mark_entry, SLAB_PANIC);
 
-	dnotify_group = fsnotify_obtain_group(DNOTIFY_GROUP_NUM,
-					      0, &dnotify_fsnotify_ops);
+	dnotify_group = fsnotify_obtain_group(0, &dnotify_fsnotify_ops);
 	if (IS_ERR(dnotify_group))
 		panic("unable to allocate fsnotify group for dnotify\n");
 	return 0;
diff --git a/fs/notify/fsnotify.c b/fs/notify/fsnotify.c
index 037e878e03fc..d4b02729598e 100644
--- a/fs/notify/fsnotify.c
+++ b/fs/notify/fsnotify.c
@@ -77,13 +77,16 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 }
 
 /* Notify this dentry's parent about a child's events. */
-void __fsnotify_parent(struct dentry *dentry, __u32 mask)
+void __fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask)
 {
 	struct dentry *parent;
 	struct inode *p_inode;
 	bool send = false;
 	bool should_update_children = false;
 
+	if (file)
+		dentry = file->f_path.dentry;
+
 	if (!(dentry->d_flags & DCACHE_FSNOTIFY_PARENT_WATCHED))
 		return;
 
@@ -114,8 +117,12 @@ void __fsnotify_parent(struct dentry *dentry, __u32 mask)
 		 * specifies these are events which came from a child. */
 		mask |= FS_EVENT_ON_CHILD;
 
-		fsnotify(p_inode, mask, dentry->d_inode, FSNOTIFY_EVENT_INODE,
-			 dentry->d_name.name, 0);
+		if (file)
+			fsnotify(p_inode, mask, file, FSNOTIFY_EVENT_FILE,
+				 dentry->d_name.name, 0);
+		else
+			fsnotify(p_inode, mask, dentry->d_inode, FSNOTIFY_EVENT_INODE,
+				 dentry->d_name.name, 0);
 		dput(parent);
 	}
 
@@ -156,7 +163,8 @@ void fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is, const
 	idx = srcu_read_lock(&fsnotify_grp_srcu);
 	list_for_each_entry_rcu(group, &fsnotify_groups, group_list) {
 		if (test_mask & group->mask) {
-			if (!group->ops->should_send_event(group, to_tell, mask))
+			if (!group->ops->should_send_event(group, to_tell, mask,
+							   data, data_is))
 				continue;
 			if (!event) {
 				event = fsnotify_create_event(to_tell, mask, data,
diff --git a/fs/notify/group.c b/fs/notify/group.c
index 0e1677144bc5..62fb8961a57b 100644
--- a/fs/notify/group.c
+++ b/fs/notify/group.c
@@ -78,15 +78,6 @@ void fsnotify_recalc_group_mask(struct fsnotify_group *group)
 }
 
 /*
- * Take a reference to a group so things found under the fsnotify_grp_mutex
- * can't get freed under us
- */
-static void fsnotify_get_group(struct fsnotify_group *group)
-{
-	atomic_inc(&group->refcnt);
-}
-
-/*
  * Final freeing of a group
  */
 void fsnotify_final_destroy_group(struct fsnotify_group *group)
@@ -171,50 +162,23 @@ void fsnotify_put_group(struct fsnotify_group *group)
 }
 
 /*
- * Simply run the fsnotify_groups list and find a group which matches
- * the given parameters.  If a group is found we take a reference to that
- * group.
- */
-static struct fsnotify_group *fsnotify_find_group(unsigned int group_num, __u32 mask,
-						  const struct fsnotify_ops *ops)
-{
-	struct fsnotify_group *group_iter;
-	struct fsnotify_group *group = NULL;
-
-	BUG_ON(!mutex_is_locked(&fsnotify_grp_mutex));
-
-	list_for_each_entry_rcu(group_iter, &fsnotify_groups, group_list) {
-		if (group_iter->group_num == group_num) {
-			if ((group_iter->mask == mask) &&
-			    (group_iter->ops == ops)) {
-				fsnotify_get_group(group_iter);
-				group = group_iter;
-			} else
-				group = ERR_PTR(-EEXIST);
-		}
-	}
-	return group;
-}
-
-/*
  * Either finds an existing group which matches the group_num, mask, and ops or
  * creates a new group and adds it to the global group list.  In either case we
  * take a reference for the group returned.
  */
-struct fsnotify_group *fsnotify_obtain_group(unsigned int group_num, __u32 mask,
+struct fsnotify_group *fsnotify_obtain_group(__u32 mask,
 					     const struct fsnotify_ops *ops)
 {
-	struct fsnotify_group *group, *tgroup;
+	struct fsnotify_group *group;
 
 	/* very low use, simpler locking if we just always alloc */
-	group = kmalloc(sizeof(struct fsnotify_group), GFP_KERNEL);
+	group = kzalloc(sizeof(struct fsnotify_group), GFP_KERNEL);
 	if (!group)
 		return ERR_PTR(-ENOMEM);
 
 	atomic_set(&group->refcnt, 1);
 
 	group->on_group_list = 0;
-	group->group_num = group_num;
 	group->mask = mask;
 
 	mutex_init(&group->notification_mutex);
@@ -230,14 +194,6 @@ struct fsnotify_group *fsnotify_obtain_group(unsigned int group_num, __u32 mask,
 	group->ops = ops;
 
 	mutex_lock(&fsnotify_grp_mutex);
-	tgroup = fsnotify_find_group(group_num, mask, ops);
-	if (tgroup) {
-		/* group already exists */
-		mutex_unlock(&fsnotify_grp_mutex);
-		/* destroy the new one we made */
-		fsnotify_put_group(group);
-		return tgroup;
-	}
 
 	/* group not found, add a new one */
 	list_add_rcu(&group->group_list, &fsnotify_groups);
diff --git a/fs/notify/inode_mark.c b/fs/notify/inode_mark.c
index c8a07c65482b..3f3261b11295 100644
--- a/fs/notify/inode_mark.c
+++ b/fs/notify/inode_mark.c
@@ -283,12 +283,20 @@ struct fsnotify_mark_entry *fsnotify_find_mark_entry(struct fsnotify_group *grou
 	return NULL;
 }
 
+void fsnotify_duplicate_mark(struct fsnotify_mark_entry *new, struct fsnotify_mark_entry *old)
+{
+	assert_spin_locked(&old->lock);
+	new->inode = old->inode;
+	new->group = old->group;
+	new->mask = old->mask;
+	new->free_mark = old->free_mark;
+}
+
 /*
  * Nothing fancy, just initialize lists and locks and counters.
  */
 void fsnotify_init_mark(struct fsnotify_mark_entry *entry,
 			void (*free_mark)(struct fsnotify_mark_entry *entry))
-
 {
 	spin_lock_init(&entry->lock);
 	atomic_set(&entry->refcnt, 1);
@@ -305,9 +313,10 @@ void fsnotify_init_mark(struct fsnotify_mark_entry *entry,
  * event types should be delivered to which group and for which inodes.
  */
 int fsnotify_add_mark(struct fsnotify_mark_entry *entry,
-		      struct fsnotify_group *group, struct inode *inode)
+		      struct fsnotify_group *group, struct inode *inode,
+		      int allow_dups)
 {
-	struct fsnotify_mark_entry *lentry;
+	struct fsnotify_mark_entry *lentry = NULL;
 	int ret = 0;
 
 	inode = igrab(inode);
@@ -327,7 +336,8 @@ int fsnotify_add_mark(struct fsnotify_mark_entry *entry,
 	entry->group = group;
 	entry->inode = inode;
 
-	lentry = fsnotify_find_mark_entry(group, inode);
+	if (!allow_dups)
+		lentry = fsnotify_find_mark_entry(group, inode);
 	if (!lentry) {
 		hlist_add_head(&entry->i_list, &inode->i_fsnotify_mark_entries);
 		list_add(&entry->g_list, &group->mark_entries);
diff --git a/fs/notify/inotify/inotify_fsnotify.c b/fs/notify/inotify/inotify_fsnotify.c
index c9ee67b442e1..027e07bf30ce 100644
--- a/fs/notify/inotify/inotify_fsnotify.c
+++ b/fs/notify/inotify/inotify_fsnotify.c
@@ -31,6 +31,60 @@
 
 #include "inotify.h"
 
+/*
+ * Check if 2 events contain the same information.  We do not compare private data
+ * but at this moment that isn't a problem for any know fsnotify listeners.
+ */
+static bool event_compare(struct fsnotify_event *old, struct fsnotify_event *new)
+{
+	if ((old->mask == new->mask) &&
+	    (old->to_tell == new->to_tell) &&
+	    (old->data_type == new->data_type) &&
+	    (old->name_len == new->name_len)) {
+		switch (old->data_type) {
+		case (FSNOTIFY_EVENT_INODE):
+			/* remember, after old was put on the wait_q we aren't
+			 * allowed to look at the inode any more, only thing
+			 * left to check was if the file_name is the same */
+			if (old->name_len &&
+			    !strcmp(old->file_name, new->file_name))
+				return true;
+			break;
+		case (FSNOTIFY_EVENT_PATH):
+			if ((old->path.mnt == new->path.mnt) &&
+			    (old->path.dentry == new->path.dentry))
+				return true;
+			break;
+		case (FSNOTIFY_EVENT_NONE):
+			if (old->mask & FS_Q_OVERFLOW)
+				return true;
+			else if (old->mask & FS_IN_IGNORED)
+				return false;
+			return true;
+		};
+	}
+	return false;
+}
+
+static int inotify_merge(struct list_head *list, struct fsnotify_event *event)
+{
+	struct fsnotify_event_holder *last_holder;
+	struct fsnotify_event *last_event;
+	int ret = 0;
+
+	/* and the list better be locked by something too */
+	spin_lock(&event->lock);
+
+	last_holder = list_entry(list->prev, struct fsnotify_event_holder, event_list);
+	last_event = last_holder->event;
+	if (event_compare(last_event, event))
+		ret = -EEXIST;
+
+	spin_unlock(&event->lock);
+
+	return ret;
+}
+
 static int inotify_handle_event(struct fsnotify_group *group, struct fsnotify_event *event)
 {
 	struct fsnotify_mark_entry *entry;
@@ -61,7 +115,7 @@ static int inotify_handle_event(struct fsnotify_group *group, struct fsnotify_ev
 	fsn_event_priv->group = group;
 	event_priv->wd = wd;
 
-	ret = fsnotify_add_notify_event(group, event, fsn_event_priv);
+	ret = fsnotify_add_notify_event(group, event, fsn_event_priv, inotify_merge);
 	if (ret) {
 		inotify_free_event_priv(fsn_event_priv);
 		/* EEXIST says we tail matched, EOVERFLOW isn't something
@@ -85,7 +139,8 @@ static void inotify_freeing_mark(struct fsnotify_mark_entry *entry, struct fsnot
 	inotify_ignored_and_remove_idr(entry, group);
 }
 
-static bool inotify_should_send_event(struct fsnotify_group *group, struct inode *inode, __u32 mask)
+static bool inotify_should_send_event(struct fsnotify_group *group, struct inode *inode,
+				      __u32 mask, void *data, int data_type)
 {
 	struct fsnotify_mark_entry *entry;
 	bool send;
diff --git a/fs/notify/inotify/inotify_user.c b/fs/notify/inotify/inotify_user.c
index dcd2040d330c..0728e83eeff8 100644
--- a/fs/notify/inotify/inotify_user.c
+++ b/fs/notify/inotify/inotify_user.c
@@ -55,12 +55,6 @@ int inotify_max_user_watches __read_mostly;
 static struct kmem_cache *inotify_inode_mark_cachep __read_mostly;
 struct kmem_cache *event_priv_cachep __read_mostly;
 
-/*
- * When inotify registers a new group it increments this and uses that
- * value as an offset to set the fsnotify group "name" and priority.
- */
-static atomic_t inotify_grp_num;
-
 #ifdef CONFIG_SYSCTL
 
 #include <linux/sysctl.h>
@@ -441,7 +435,7 @@ void inotify_ignored_and_remove_idr(struct fsnotify_mark_entry *entry,
 	fsn_event_priv->group = group;
 	event_priv->wd = ientry->wd;
 
-	ret = fsnotify_add_notify_event(group, ignored_event, fsn_event_priv);
+	ret = fsnotify_add_notify_event(group, ignored_event, fsn_event_priv, NULL);
 	if (ret)
 		inotify_free_event_priv(fsn_event_priv);
 
@@ -459,7 +453,9 @@ skip_send_ignore:
 /* ding dong the mark is dead */
 static void inotify_free_mark(struct fsnotify_mark_entry *entry)
 {
-	struct inotify_inode_mark_entry *ientry = (struct inotify_inode_mark_entry *)entry;
+	struct inotify_inode_mark_entry *ientry;
+
+	ientry = container_of(entry, struct inotify_inode_mark_entry, fsn_entry);
 
 	kmem_cache_free(inotify_inode_mark_cachep, ientry);
 }
@@ -572,7 +568,7 @@ retry:
 	fsnotify_get_mark(&tmp_ientry->fsn_entry);
 
 	/* we are on the idr, now get on the inode */
-	ret = fsnotify_add_mark(&tmp_ientry->fsn_entry, group, inode);
+	ret = fsnotify_add_mark(&tmp_ientry->fsn_entry, group, inode, 0);
 	if (ret) {
 		/* we failed to get on the inode, get off the idr */
 		inotify_remove_from_idr(group, tmp_ientry);
@@ -626,11 +622,8 @@ retry:
 static struct fsnotify_group *inotify_new_group(struct user_struct *user, unsigned int max_events)
 {
 	struct fsnotify_group *group;
-	unsigned int grp_num;
 
-	/* fsnotify_obtain_group took a reference to group, we put this when we kill the file in the end */
-	grp_num = (INOTIFY_GROUP_NUM - atomic_inc_return(&inotify_grp_num));
-	group = fsnotify_obtain_group(grp_num, 0, &inotify_fsnotify_ops);
+	group = fsnotify_obtain_group(0, &inotify_fsnotify_ops);
 	if (IS_ERR(group))
 		return group;
 
diff --git a/fs/notify/notification.c b/fs/notify/notification.c
index 3816d5750dd5..dafd0b7687b8 100644
--- a/fs/notify/notification.c
+++ b/fs/notify/notification.c
@@ -56,7 +56,7 @@ static struct kmem_cache *fsnotify_event_holder_cachep;
  * it is needed.  It's refcnt is set 1 at kernel init time and will never
  * get set to 0 so it will never get 'freed'
  */
-static struct fsnotify_event q_overflow_event;
+static struct fsnotify_event *q_overflow_event;
 static atomic_t fsnotify_sync_cookie = ATOMIC_INIT(0);
 
 /**
@@ -104,7 +104,8 @@ struct fsnotify_event_holder *fsnotify_alloc_event_holder(void)
 
 void fsnotify_destroy_event_holder(struct fsnotify_event_holder *holder)
 {
-	kmem_cache_free(fsnotify_event_holder_cachep, holder);
+	if (holder)
+		kmem_cache_free(fsnotify_event_holder_cachep, holder);
 }
 
 /*
@@ -129,53 +130,17 @@ struct fsnotify_event_private_data *fsnotify_remove_priv_from_event(struct fsnot
 }
 
 /*
- * Check if 2 events contain the same information.  We do not compare private data
- * but at this moment that isn't a problem for any know fsnotify listeners.
- */
-static bool event_compare(struct fsnotify_event *old, struct fsnotify_event *new)
-{
-	if ((old->mask == new->mask) &&
-	    (old->to_tell == new->to_tell) &&
-	    (old->data_type == new->data_type) &&
-	    (old->name_len == new->name_len)) {
-		switch (old->data_type) {
-		case (FSNOTIFY_EVENT_INODE):
-			/* remember, after old was put on the wait_q we aren't
-			 * allowed to look at the inode any more, only thing
-			 * left to check was if the file_name is the same */
-			if (old->name_len &&
-			    !strcmp(old->file_name, new->file_name))
-				return true;
-			break;
-		case (FSNOTIFY_EVENT_PATH):
-			if ((old->path.mnt == new->path.mnt) &&
-			    (old->path.dentry == new->path.dentry))
-				return true;
-			break;
-		case (FSNOTIFY_EVENT_NONE):
-			if (old->mask & FS_Q_OVERFLOW)
-				return true;
-			else if (old->mask & FS_IN_IGNORED)
-				return false;
-			return false;
-		};
-	}
-	return false;
-}
-
-/*
  * Add an event to the group notification queue.  The group can later pull this
  * event off the queue to deal with.  If the event is successfully added to the
  * group's notification queue, a reference is taken on event.
  */
 int fsnotify_add_notify_event(struct fsnotify_group *group, struct fsnotify_event *event,
-			      struct fsnotify_event_private_data *priv)
+			      struct fsnotify_event_private_data *priv,
+			      int (*merge)(struct list_head *, struct fsnotify_event *))
 {
 	struct fsnotify_event_holder *holder = NULL;
 	struct list_head *list = &group->notification_list;
-	struct fsnotify_event_holder *last_holder;
-	struct fsnotify_event *last_event;
-	int ret = 0;
+	int rc = 0;
 
 	/*
 	 * There is one fsnotify_event_holder embedded inside each fsnotify_event.
@@ -195,12 +160,24 @@ alloc_holder:
 	mutex_lock(&group->notification_mutex);
 
 	if (group->q_len >= group->max_events) {
-		event = &q_overflow_event;
-		ret = -EOVERFLOW;
+		event = q_overflow_event;
+		rc = -EOVERFLOW;
 		/* sorry, no private data on the overflow event */
 		priv = NULL;
 	}
 
+	if (!list_empty(list) && merge) {
+		int ret;
+
+		ret = merge(list, event);
+		if (ret) {
+			mutex_unlock(&group->notification_mutex);
+			if (holder != &event->holder)
+				fsnotify_destroy_event_holder(holder);
+			return ret;
+		}
+	}
+
 	spin_lock(&event->lock);
 
 	if (list_empty(&event->holder.event_list)) {
@@ -215,18 +192,6 @@ alloc_holder:
 		goto alloc_holder;
 	}
 
-	if (!list_empty(list)) {
-		last_holder = list_entry(list->prev, struct fsnotify_event_holder, event_list);
-		last_event = last_holder->event;
-		if (event_compare(last_event, event)) {
-			spin_unlock(&event->lock);
-			mutex_unlock(&group->notification_mutex);
-			if (holder != &event->holder)
-				fsnotify_destroy_event_holder(holder);
-			return -EEXIST;
-		}
-	}
-
 	group->q_len++;
 	holder->event = event;
 
@@ -238,7 +203,7 @@ alloc_holder:
 	mutex_unlock(&group->notification_mutex);
 
 	wake_up(&group->notification_waitq);
-	return ret;
+	return rc;
 }
 
 /*
@@ -314,25 +279,77 @@ void fsnotify_flush_notify(struct fsnotify_group *group)
 
 static void initialize_event(struct fsnotify_event *event)
 {
-	event->holder.event = NULL;
 	INIT_LIST_HEAD(&event->holder.event_list);
 	atomic_set(&event->refcnt, 1);
 
 	spin_lock_init(&event->lock);
 
-	event->path.dentry = NULL;
-	event->path.mnt = NULL;
-	event->inode = NULL;
-	event->data_type = FSNOTIFY_EVENT_NONE;
-
 	INIT_LIST_HEAD(&event->private_data_list);
+}
+
+/*
+ * Caller damn well better be holding whatever mutex is protecting the
+ * old_holder->event_list and the new_event must be a clean event which
+ * cannot be found anywhere else in the kernel.
+ */
+int fsnotify_replace_event(struct fsnotify_event_holder *old_holder,
+			   struct fsnotify_event *new_event)
+{
+	struct fsnotify_event *old_event = old_holder->event;
+	struct fsnotify_event_holder *new_holder = &new_event->holder;
 
-	event->to_tell = NULL;
+	enum event_spinlock_class {
+		SPINLOCK_OLD,
+		SPINLOCK_NEW,
+	};
 
-	event->file_name = NULL;
-	event->name_len = 0;
+	/*
+	 * if the new_event's embedded holder is in use someone
+	 * screwed up and didn't give us a clean new event.
+	 */
+	BUG_ON(!list_empty(&new_holder->event_list));
+
+	spin_lock_nested(&old_event->lock, SPINLOCK_OLD);
+	spin_lock_nested(&new_event->lock, SPINLOCK_NEW);
+
+	new_holder->event = new_event;
+	list_replace_init(&old_holder->event_list, &new_holder->event_list);
+
+	spin_unlock(&new_event->lock);
+	spin_unlock(&old_event->lock);
+
+	/* event == holder means we are referenced through the in event holder */
+	if (old_holder != &old_event->holder)
+		fsnotify_destroy_event_holder(old_holder);
+
+	fsnotify_get_event(new_event); /* on the list take reference */
+	fsnotify_put_event(old_event); /* off the list, drop reference */
+
+	return 0;
+}
+
+struct fsnotify_event *fsnotify_clone_event(struct fsnotify_event *old_event)
+{
+	struct fsnotify_event *event;
 
-	event->sync_cookie = 0;
+	event = kmem_cache_alloc(fsnotify_event_cachep, GFP_KERNEL);
+	if (!event)
+		return NULL;
+
+	memcpy(event, old_event, sizeof(*event));
+	initialize_event(event);
+
+	if (event->name_len) {
+		event->file_name = kstrdup(old_event->file_name, GFP_KERNEL);
+		if (!event->file_name) {
+			kmem_cache_free(fsnotify_event_cachep, event);
+			return NULL;
+		}
+	}
+	if (event->data_type == FSNOTIFY_EVENT_PATH)
+		path_get(&event->path);
+
+	return event;
 }
 
 /*
@@ -353,7 +370,7 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask,
 {
 	struct fsnotify_event *event;
 
-	event = kmem_cache_alloc(fsnotify_event_cachep, gfp);
+	event = kmem_cache_zalloc(fsnotify_event_cachep, gfp);
 	if (!event)
 		return NULL;
 
@@ -370,6 +387,7 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask,
 
 	event->sync_cookie = cookie;
 	event->to_tell = to_tell;
+	event->data_type = data_type;
 
 	switch (data_type) {
 	case FSNOTIFY_EVENT_FILE: {
@@ -386,12 +404,10 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask,
 		event->path.dentry = path->dentry;
 		event->path.mnt = path->mnt;
 		path_get(&event->path);
-		event->data_type = FSNOTIFY_EVENT_PATH;
 		break;
 	}
 	case FSNOTIFY_EVENT_INODE:
 		event->inode = data;
-		event->data_type = FSNOTIFY_EVENT_INODE;
 		break;
 	case FSNOTIFY_EVENT_NONE:
 		event->inode = NULL;
@@ -412,8 +428,11 @@ __init int fsnotify_notification_init(void)
 	fsnotify_event_cachep = KMEM_CACHE(fsnotify_event, SLAB_PANIC);
 	fsnotify_event_holder_cachep = KMEM_CACHE(fsnotify_event_holder, SLAB_PANIC);
 
-	initialize_event(&q_overflow_event);
-	q_overflow_event.mask = FS_Q_OVERFLOW;
+	q_overflow_event = fsnotify_create_event(NULL, FS_Q_OVERFLOW, NULL,
+						 FSNOTIFY_EVENT_NONE, NULL, 0,
+						 GFP_KERNEL);
+	if (!q_overflow_event)
+		panic("unable to allocate fsnotify q_overflow_event\n");
 
 	return 0;
 }
diff --git a/fs/open.c b/fs/open.c
index 40d1fa25f5aa..6f37e62b6ea6 100644
--- a/fs/open.c
+++ b/fs/open.c
@@ -1040,7 +1040,7 @@ long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
 				put_unused_fd(fd);
 				fd = PTR_ERR(f);
 			} else {
-				fsnotify_open(f->f_path.dentry);
+				fsnotify_open(f);
 				fd_install(fd, f);
 			}
 		}
diff --git a/fs/read_write.c b/fs/read_write.c
index 6c8c55dec2bc..c7479e91848f 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -293,7 +293,7 @@ ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
 		else
 			ret = do_sync_read(file, buf, count, pos);
 		if (ret > 0) {
-			fsnotify_access(file->f_path.dentry);
+			fsnotify_access(file);
 			add_rchar(current, ret);
 		}
 		inc_syscr(current);
@@ -348,7 +348,7 @@ ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_
 		else
 			ret = do_sync_write(file, buf, count, pos);
 		if (ret > 0) {
-			fsnotify_modify(file->f_path.dentry);
+			fsnotify_modify(file);
 			add_wchar(current, ret);
 		}
 		inc_syscw(current);
@@ -656,9 +656,9 @@ out:
 		kfree(iov);
 	if ((ret + (type == READ)) > 0) {
 		if (type == READ)
-			fsnotify_access(file->f_path.dentry);
+			fsnotify_access(file);
 		else
-			fsnotify_modify(file->f_path.dentry);
+			fsnotify_modify(file);
 	}
 	return ret;
 }