Merge branch 'for-linus' of git://git.infradead.org/users/eparis/notify

* 'for-linus' of git://git.infradead.org/users/eparis/notify:
  fsnotify: allow groups to set freeing_mark to null
  inotify/dnotify: should_send_event shouldn't match on FS_EVENT_ON_CHILD
  dnotify: do not bother to lock entry->lock when reading mask
  dnotify: do not use ?true:false when assigning to a bool
  fsnotify: move events should indicate the event was on a child
  inotify: reimplement inotify using fsnotify
  fsnotify: handle filesystem unmounts with fsnotify marks
  fsnotify: fsnotify marks on inodes pin them in core
  fsnotify: allow groups to add private data to events
  fsnotify: add correlations between events
  fsnotify: include pathnames with entries when possible
  fsnotify: generic notification queue and waitq
  dnotify: reimplement dnotify using fsnotify
  fsnotify: parent event notification
  fsnotify: add marks to inodes so groups can interpret how to handle those inodes
  fsnotify: unified filesystem notification backend

16 files changed, 2285 insertions, 554 deletions

diff --git a/fs/inode.c b/fs/inode.c
index bca0c618fdb..ca337014ae2 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -22,6 +22,7 @@
 #include <linux/cdev.h>
 #include <linux/bootmem.h>
 #include <linux/inotify.h>
+#include <linux/fsnotify.h>
 #include <linux/mount.h>
 #include <linux/async.h>
 
@@ -189,6 +190,10 @@ struct inode *inode_init_always(struct super_block *sb, struct inode *inode)
 	inode->i_private = NULL;
 	inode->i_mapping = mapping;
 
+#ifdef CONFIG_FSNOTIFY
+	inode->i_fsnotify_mask = 0;
+#endif
+
 	return inode;
 
 out_free_security:
@@ -221,6 +226,7 @@ void destroy_inode(struct inode *inode)
 	BUG_ON(inode_has_buffers(inode));
 	ima_inode_free(inode);
 	security_inode_free(inode);
+	fsnotify_inode_delete(inode);
 	if (inode->i_sb->s_op->destroy_inode)
 		inode->i_sb->s_op->destroy_inode(inode);
 	else
@@ -252,6 +258,9 @@ void inode_init_once(struct inode *inode)
 	INIT_LIST_HEAD(&inode->inotify_watches);
 	mutex_init(&inode->inotify_mutex);
 #endif
+#ifdef CONFIG_FSNOTIFY
+	INIT_HLIST_HEAD(&inode->i_fsnotify_mark_entries);
+#endif
 }
 EXPORT_SYMBOL(inode_init_once);
 
@@ -398,6 +407,7 @@ int invalidate_inodes(struct super_block *sb)
 	mutex_lock(&iprune_mutex);
 	spin_lock(&inode_lock);
 	inotify_unmount_inodes(&sb->s_inodes);
+	fsnotify_unmount_inodes(&sb->s_inodes);
 	busy = invalidate_list(&sb->s_inodes, &throw_away);
 	spin_unlock(&inode_lock);
 
diff --git a/fs/notify/Kconfig b/fs/notify/Kconfig
index 50914d7303c..31dac7e3b0f 100644
--- a/fs/notify/Kconfig
+++ b/fs/notify/Kconfig
@@ -1,2 +1,15 @@
+config FSNOTIFY
+	bool "Filesystem notification backend"
+	default y
+	---help---
+	   fsnotify is a backend for filesystem notification.  fsnotify does
+	   not provide any userspace interface but does provide the basis
+	   needed for other notification schemes such as dnotify, inotify,
+	   and fanotify.
+
+	   Say Y here to enable fsnotify suport.
+
+	   If unsure, say Y.
+
 source "fs/notify/dnotify/Kconfig"
 source "fs/notify/inotify/Kconfig"
diff --git a/fs/notify/Makefile b/fs/notify/Makefile
index 5a95b6010ce..0922cc826c4 100644
--- a/fs/notify/Makefile
+++ b/fs/notify/Makefile
@@ -1,2 +1,4 @@
+obj-$(CONFIG_FSNOTIFY)		+= fsnotify.o notification.o group.o inode_mark.o
+
 obj-y			+= dnotify/
 obj-y			+= inotify/
diff --git a/fs/notify/dnotify/Kconfig b/fs/notify/dnotify/Kconfig
index 26adf5dfa64..904ff8d5405 100644
--- a/fs/notify/dnotify/Kconfig
+++ b/fs/notify/dnotify/Kconfig
@@ -1,5 +1,6 @@
 config DNOTIFY
 	bool "Dnotify support"
+	depends on FSNOTIFY
 	default y
 	help
 	  Dnotify is a directory-based per-fd file change notification system
diff --git a/fs/notify/dnotify/dnotify.c b/fs/notify/dnotify/dnotify.c
index b0aa2cde80b..828a889be90 100644
--- a/fs/notify/dnotify/dnotify.c
+++ b/fs/notify/dnotify/dnotify.c
@@ -3,6 +3,9 @@
  *
  * Copyright (C) 2000,2001,2002 Stephen Rothwell
  *
+ * Copyright (C) 2009 Eric Paris <Red Hat Inc>
+ * dnotify was largly rewritten to use the new fsnotify infrastructure
+ *
  * 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
@@ -21,24 +24,173 @@
 #include <linux/spinlock.h>
 #include <linux/slab.h>
 #include <linux/fdtable.h>
+#include <linux/fsnotify_backend.h>
 
 int dir_notify_enable __read_mostly = 1;
 
-static struct kmem_cache *dn_cache __read_mostly;
+static struct kmem_cache *dnotify_struct_cache __read_mostly;
+static struct kmem_cache *dnotify_mark_entry_cache __read_mostly;
+static struct fsnotify_group *dnotify_group __read_mostly;
+static DEFINE_MUTEX(dnotify_mark_mutex);
+
+/*
+ * dnotify will attach one of these to each inode (i_fsnotify_mark_entries) which
+ * is being watched by dnotify.  If multiple userspace applications are watching
+ * the same directory with dnotify their information is chained in dn
+ */
+struct dnotify_mark_entry {
+	struct fsnotify_mark_entry fsn_entry;
+	struct dnotify_struct *dn;
+};
 
-static void redo_inode_mask(struct inode *inode)
+/*
+ * When a process starts or stops watching an inode the set of events which
+ * dnotify cares about for that inode may change.  This function runs the
+ * list of everything receiving dnotify events about this directory and calculates
+ * the set of all those events.  After it updates what dnotify is interested in
+ * it calls the fsnotify function so it can update the set of all events relevant
+ * to this inode.
+ */
+static void dnotify_recalc_inode_mask(struct fsnotify_mark_entry *entry)
 {
-	unsigned long new_mask;
+	__u32 new_mask, old_mask;
 	struct dnotify_struct *dn;
+	struct dnotify_mark_entry *dnentry  = container_of(entry,
+							   struct dnotify_mark_entry,
+							   fsn_entry);
+
+	assert_spin_locked(&entry->lock);
 
+	old_mask = entry->mask;
 	new_mask = 0;
-	for (dn = inode->i_dnotify; dn != NULL; dn = dn->dn_next)
-		new_mask |= dn->dn_mask & ~DN_MULTISHOT;
-	inode->i_dnotify_mask = new_mask;
+	for (dn = dnentry->dn; dn != NULL; dn = dn->dn_next)
+		new_mask |= (dn->dn_mask & ~FS_DN_MULTISHOT);
+	entry->mask = new_mask;
+
+	if (old_mask == new_mask)
+		return;
+
+	if (entry->inode)
+		fsnotify_recalc_inode_mask(entry->inode);
+}
+
+/*
+ * Mains fsnotify call where events are delivered to dnotify.
+ * Find the dnotify mark on the relevant inode, run the list of dnotify structs
+ * on that mark and determine which of them has expressed interest in receiving
+ * events of this type.  When found send the correct process and signal and
+ * destroy the dnotify struct if it was not registered to receive multiple
+ * events.
+ */
+static int dnotify_handle_event(struct fsnotify_group *group,
+				struct fsnotify_event *event)
+{
+	struct fsnotify_mark_entry *entry = NULL;
+	struct dnotify_mark_entry *dnentry;
+	struct inode *to_tell;
+	struct dnotify_struct *dn;
+	struct dnotify_struct **prev;
+	struct fown_struct *fown;
+
+	to_tell = event->to_tell;
+
+	spin_lock(&to_tell->i_lock);
+	entry = fsnotify_find_mark_entry(group, to_tell);
+	spin_unlock(&to_tell->i_lock);
+
+	/* unlikely since we alreay passed dnotify_should_send_event() */
+	if (unlikely(!entry))
+		return 0;
+	dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry);
+
+	spin_lock(&entry->lock);
+	prev = &dnentry->dn;
+	while ((dn = *prev) != NULL) {
+		if ((dn->dn_mask & event->mask) == 0) {
+			prev = &dn->dn_next;
+			continue;
+		}
+		fown = &dn->dn_filp->f_owner;
+		send_sigio(fown, dn->dn_fd, POLL_MSG);
+		if (dn->dn_mask & FS_DN_MULTISHOT)
+			prev = &dn->dn_next;
+		else {
+			*prev = dn->dn_next;
+			kmem_cache_free(dnotify_struct_cache, dn);
+			dnotify_recalc_inode_mask(entry);
+		}
+	}
+
+	spin_unlock(&entry->lock);
+	fsnotify_put_mark(entry);
+
+	return 0;
+}
+
+/*
+ * Given an inode and mask determine if dnotify would be interested in sending
+ * userspace notification for that pair.
+ */
+static bool dnotify_should_send_event(struct fsnotify_group *group,
+				      struct inode *inode, __u32 mask)
+{
+	struct fsnotify_mark_entry *entry;
+	bool send;
+
+	/* !dir_notify_enable should never get here, don't waste time checking
+	if (!dir_notify_enable)
+		return 0; */
+
+	/* not a dir, dnotify doesn't care */
+	if (!S_ISDIR(inode->i_mode))
+		return false;
+
+	spin_lock(&inode->i_lock);
+	entry = fsnotify_find_mark_entry(group, inode);
+	spin_unlock(&inode->i_lock);
+
+	/* no mark means no dnotify watch */
+	if (!entry)
+		return false;
+
+	mask = (mask & ~FS_EVENT_ON_CHILD);
+	send = (mask & entry->mask);
+
+	fsnotify_put_mark(entry); /* matches fsnotify_find_mark_entry */
+
+	return send;
+}
+
+static void dnotify_free_mark(struct fsnotify_mark_entry *entry)
+{
+	struct dnotify_mark_entry *dnentry = container_of(entry,
+							  struct dnotify_mark_entry,
+							  fsn_entry);
+
+	BUG_ON(dnentry->dn);
+
+	kmem_cache_free(dnotify_mark_entry_cache, dnentry);
 }
 
+static struct fsnotify_ops dnotify_fsnotify_ops = {
+	.handle_event = dnotify_handle_event,
+	.should_send_event = dnotify_should_send_event,
+	.free_group_priv = NULL,
+	.freeing_mark = NULL,
+	.free_event_priv = NULL,
+};
+
+/*
+ * Called every time a file is closed.  Looks first for a dnotify mark on the
+ * inode.  If one is found run all of the ->dn entries attached to that
+ * mark for one relevant to this process closing the file and remove that
+ * dnotify_struct.  If that was the last dnotify_struct also remove the
+ * fsnotify_mark_entry.
+ */
 void dnotify_flush(struct file *filp, fl_owner_t id)
 {
+	struct fsnotify_mark_entry *entry;
+	struct dnotify_mark_entry *dnentry;
 	struct dnotify_struct *dn;
 	struct dnotify_struct **prev;
 	struct inode *inode;
@@ -46,145 +198,243 @@ void dnotify_flush(struct file *filp, fl_owner_t id)
 	inode = filp->f_path.dentry->d_inode;
 	if (!S_ISDIR(inode->i_mode))
 		return;
+
 	spin_lock(&inode->i_lock);
-	prev = &inode->i_dnotify;
+	entry = fsnotify_find_mark_entry(dnotify_group, inode);
+	spin_unlock(&inode->i_lock);
+	if (!entry)
+		return;
+	dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry);
+
+	mutex_lock(&dnotify_mark_mutex);
+
+	spin_lock(&entry->lock);
+	prev = &dnentry->dn;
 	while ((dn = *prev) != NULL) {
 		if ((dn->dn_owner == id) && (dn->dn_filp == filp)) {
 			*prev = dn->dn_next;
-			redo_inode_mask(inode);
-			kmem_cache_free(dn_cache, dn);
+			kmem_cache_free(dnotify_struct_cache, dn);
+			dnotify_recalc_inode_mask(entry);
 			break;
 		}
 		prev = &dn->dn_next;
 	}
-	spin_unlock(&inode->i_lock);
+
+	spin_unlock(&entry->lock);
+
+	/* nothing else could have found us thanks to the dnotify_mark_mutex */
+	if (dnentry->dn == NULL)
+		fsnotify_destroy_mark_by_entry(entry);
+
+	fsnotify_recalc_group_mask(dnotify_group);
+
+	mutex_unlock(&dnotify_mark_mutex);
+
+	fsnotify_put_mark(entry);
+}
+
+/* this conversion is done only at watch creation */
+static __u32 convert_arg(unsigned long arg)
+{
+	__u32 new_mask = FS_EVENT_ON_CHILD;
+
+	if (arg & DN_MULTISHOT)
+		new_mask |= FS_DN_MULTISHOT;
+	if (arg & DN_DELETE)
+		new_mask |= (FS_DELETE | FS_MOVED_FROM);
+	if (arg & DN_MODIFY)
+		new_mask |= FS_MODIFY;
+	if (arg & DN_ACCESS)
+		new_mask |= FS_ACCESS;
+	if (arg & DN_ATTRIB)
+		new_mask |= FS_ATTRIB;
+	if (arg & DN_RENAME)
+		new_mask |= FS_DN_RENAME;
+	if (arg & DN_CREATE)
+		new_mask |= (FS_CREATE | FS_MOVED_TO);
+
+	return new_mask;
 }
 
+/*
+ * If multiple processes watch the same inode with dnotify there is only one
+ * dnotify mark in inode->i_fsnotify_mark_entries but we chain a dnotify_struct
+ * onto that mark.  This function either attaches the new dnotify_struct onto
+ * that list, or it |= the mask onto an existing dnofiy_struct.
+ */
+static int attach_dn(struct dnotify_struct *dn, struct dnotify_mark_entry *dnentry,
+		     fl_owner_t id, int fd, struct file *filp, __u32 mask)
+{
+	struct dnotify_struct *odn;
+
+	odn = dnentry->dn;
+	while (odn != NULL) {
+		/* adding more events to existing dnofiy_struct? */
+		if ((odn->dn_owner == id) && (odn->dn_filp == filp)) {
+			odn->dn_fd = fd;
+			odn->dn_mask |= mask;
+			return -EEXIST;
+		}
+		odn = odn->dn_next;
+	}
+
+	dn->dn_mask = mask;
+	dn->dn_fd = fd;
+	dn->dn_filp = filp;
+	dn->dn_owner = id;
+	dn->dn_next = dnentry->dn;
+	dnentry->dn = dn;
+
+	return 0;
+}
+
+/*
+ * When a process calls fcntl to attach a dnotify watch to a directory it ends
+ * up here.  Allocate both a mark for fsnotify to add and a dnotify_struct to be
+ * attached to the fsnotify_mark.
+ */
 int fcntl_dirnotify(int fd, struct file *filp, unsigned long arg)
 {
+	struct dnotify_mark_entry *new_dnentry, *dnentry;
+	struct fsnotify_mark_entry *new_entry, *entry;
 	struct dnotify_struct *dn;
-	struct dnotify_struct *odn;
-	struct dnotify_struct **prev;
 	struct inode *inode;
 	fl_owner_t id = current->files;
 	struct file *f;
-	int error = 0;
+	int destroy = 0, error = 0;
+	__u32 mask;
+
+	/* we use these to tell if we need to kfree */
+	new_entry = NULL;
+	dn = NULL;
+
+	if (!dir_notify_enable) {
+		error = -EINVAL;
+		goto out_err;
+	}
 
+	/* a 0 mask means we are explicitly removing the watch */
 	if ((arg & ~DN_MULTISHOT) == 0) {
 		dnotify_flush(filp, id);
-		return 0;
+		error = 0;
+		goto out_err;
 	}
-	if (!dir_notify_enable)
-		return -EINVAL;
+
+	/* dnotify only works on directories */
 	inode = filp->f_path.dentry->d_inode;
-	if (!S_ISDIR(inode->i_mode))
-		return -ENOTDIR;
-	dn = kmem_cache_alloc(dn_cache, GFP_KERNEL);
-	if (dn == NULL)
-		return -ENOMEM;
-	spin_lock(&inode->i_lock);
-	prev = &inode->i_dnotify;
-	while ((odn = *prev) != NULL) {
-		if ((odn->dn_owner == id) && (odn->dn_filp == filp)) {
-			odn->dn_fd = fd;
-			odn->dn_mask |= arg;
-			inode->i_dnotify_mask |= arg & ~DN_MULTISHOT;
-			goto out_free;
-		}
-		prev = &odn->dn_next;
+	if (!S_ISDIR(inode->i_mode)) {
+		error = -ENOTDIR;
+		goto out_err;
 	}
 
-	rcu_read_lock();
-	f = fcheck(fd);
-	rcu_read_unlock();
-	/* we'd lost the race with close(), sod off silently */
-	/* note that inode->i_lock prevents reordering problems
-	 * between accesses to descriptor table and ->i_dnotify */
-	if (f != filp)
-		goto out_free;
+	/* expect most fcntl to add new rather than augment old */
+	dn = kmem_cache_alloc(dnotify_struct_cache, GFP_KERNEL);
+	if (!dn) {
+		error = -ENOMEM;
+		goto out_err;
+	}
 
-	error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
-	if (error)
-		goto out_free;
+	/* new fsnotify mark, we expect most fcntl calls to add a new mark */
+	new_dnentry = kmem_cache_alloc(dnotify_mark_entry_cache, GFP_KERNEL);
+	if (!new_dnentry) {
+		error = -ENOMEM;
+		goto out_err;
+	}
 
-	dn->dn_mask = arg;
-	dn->dn_fd = fd;
-	dn->dn_filp = filp;
-	dn->dn_owner = id;
-	inode->i_dnotify_mask |= arg & ~DN_MULTISHOT;
-	dn->dn_next = inode->i_dnotify;
-	inode->i_dnotify = dn;
-	spin_unlock(&inode->i_lock);
-	return 0;
+	/* convert the userspace DN_* "arg" to the internal FS_* defines in fsnotify */
+	mask = convert_arg(arg);
 
-out_free:
-	spin_unlock(&inode->i_lock);
-	kmem_cache_free(dn_cache, dn);
-	return error;
-}
+	/* set up the new_entry and new_dnentry */
+	new_entry = &new_dnentry->fsn_entry;
+	fsnotify_init_mark(new_entry, dnotify_free_mark);
+	new_entry->mask = mask;
+	new_dnentry->dn = NULL;
 
-void __inode_dir_notify(struct inode *inode, unsigned long event)
-{
-	struct dnotify_struct *	dn;
-	struct dnotify_struct **prev;
-	struct fown_struct *	fown;
-	int			changed = 0;
+	/* this is needed to prevent the fcntl/close race described below */
+	mutex_lock(&dnotify_mark_mutex);
 
+	/* add the new_entry or find an old one. */
 	spin_lock(&inode->i_lock);
-	prev = &inode->i_dnotify;
-	while ((dn = *prev) != NULL) {
-		if ((dn->dn_mask & event) == 0) {
-			prev = &dn->dn_next;
-			continue;
-		}
-		fown = &dn->dn_filp->f_owner;
-		send_sigio(fown, dn->dn_fd, POLL_MSG);
-		if (dn->dn_mask & DN_MULTISHOT)
-			prev = &dn->dn_next;
-		else {
-			*prev = dn->dn_next;
-			changed = 1;
-			kmem_cache_free(dn_cache, dn);
-		}
-	}
-	if (changed)
-		redo_inode_mask(inode);
+	entry = fsnotify_find_mark_entry(dnotify_group, inode);
 	spin_unlock(&inode->i_lock);
-}
-
-EXPORT_SYMBOL(__inode_dir_notify);
+	if (entry) {
+		dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry);
+		spin_lock(&entry->lock);
+	} else {
+		fsnotify_add_mark(new_entry, dnotify_group, inode);
+		spin_lock(&new_entry->lock);
+		entry = new_entry;
+		dnentry = new_dnentry;
+		/* we used new_entry, so don't free it */
+		new_entry = NULL;
+	}
 
-/*
- * This is hopelessly wrong, but unfixable without API changes.  At
- * least it doesn't oops the kernel...
- *
- * To safely access ->d_parent we need to keep d_move away from it.  Use the
- * dentry's d_lock for this.
- */
-void dnotify_parent(struct dentry *dentry, unsigned long event)
-{
-	struct dentry *parent;
+	rcu_read_lock();
+	f = fcheck(fd);
+	rcu_read_unlock();
 
-	if (!dir_notify_enable)
-		return;
+	/* if (f != filp) means that we lost a race and another task/thread
+	 * actually closed the fd we are still playing with before we grabbed
+	 * the dnotify_mark_mutex and entry->lock.  Since closing the fd is the
+	 * only time we clean up the mark entries we need to get our mark off
+	 * the list. */
+	if (f != filp) {
+		/* if we added ourselves, shoot ourselves, it's possible that
+		 * the flush actually did shoot this entry.  That's fine too
+		 * since multiple calls to destroy_mark is perfectly safe, if
+		 * we found a dnentry already attached to the inode, just sod
+		 * off silently as the flush at close time dealt with it.
+		 */
+		if (dnentry == new_dnentry)
+			destroy = 1;
+		goto out;
+	}
 
-	spin_lock(&dentry->d_lock);
-	parent = dentry->d_parent;
-	if (parent->d_inode->i_dnotify_mask & event) {
-		dget(parent);
-		spin_unlock(&dentry->d_lock);
-		__inode_dir_notify(parent->d_inode, event);
-		dput(parent);
-	} else {
-		spin_unlock(&dentry->d_lock);
+	error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
+	if (error) {
+		/* if we added, we must shoot */
+		if (dnentry == new_dnentry)
+			destroy = 1;
+		goto out;
 	}
+
+	error = attach_dn(dn, dnentry, id, fd, filp, mask);
+	/* !error means that we attached the dn to the dnentry, so don't free it */
+	if (!error)
+		dn = NULL;
+	/* -EEXIST means that we didn't add this new dn and used an old one.
+	 * that isn't an error (and the unused dn should be freed) */
+	else if (error == -EEXIST)
+		error = 0;
+
+	dnotify_recalc_inode_mask(entry);
+out:
+	spin_unlock(&entry->lock);
+
+	if (destroy)
+		fsnotify_destroy_mark_by_entry(entry);
+
+	fsnotify_recalc_group_mask(dnotify_group);
+
+	mutex_unlock(&dnotify_mark_mutex);
+	fsnotify_put_mark(entry);
+out_err:
+	if (new_entry)
+		fsnotify_put_mark(new_entry);
+	if (dn)
+		kmem_cache_free(dnotify_struct_cache, dn);
+	return error;
 }
-EXPORT_SYMBOL_GPL(dnotify_parent);
 
 static int __init dnotify_init(void)
 {
-	dn_cache = kmem_cache_create("dnotify_cache",
-		sizeof(struct dnotify_struct), 0, SLAB_PANIC, NULL);
+	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);
+	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
new file mode 100644
index 00000000000..ec2f7bd7681
--- /dev/null
+++ b/fs/notify/fsnotify.c
@@ -0,0 +1,186 @@
+/*
+ *  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.
+ */
+
+#include <linux/dcache.h>
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/srcu.h>
+
+#include <linux/fsnotify_backend.h>
+#include "fsnotify.h"
+
+/*
+ * Clear all of the marks on an inode when it is being evicted from core
+ */
+void __fsnotify_inode_delete(struct inode *inode)
+{
+	fsnotify_clear_marks_by_inode(inode);
+}
+EXPORT_SYMBOL_GPL(__fsnotify_inode_delete);
+
+/*
+ * Given an inode, first check if we care what happens to our children.  Inotify
+ * and dnotify both tell their parents about events.  If we care about any event
+ * on a child we run all of our children and set a dentry flag saying that the
+ * parent cares.  Thus when an event happens on a child it can quickly tell if
+ * if there is a need to find a parent and send the event to the parent.
+ */
+void __fsnotify_update_child_dentry_flags(struct inode *inode)
+{
+	struct dentry *alias;
+	int watched;
+
+	if (!S_ISDIR(inode->i_mode))
+		return;
+
+	/* determine if the children should tell inode about their events */
+	watched = fsnotify_inode_watches_children(inode);
+
+	spin_lock(&dcache_lock);
+	/* run all of the dentries associated with this inode.  Since this is a
+	 * directory, there damn well better only be one item on this list */
+	list_for_each_entry(alias, &inode->i_dentry, d_alias) {
+		struct dentry *child;
+
+		/* run all of the children of the original inode and fix their
+		 * d_flags to indicate parental interest (their parent is the
+		 * original inode) */
+		list_for_each_entry(child, &alias->d_subdirs, d_u.d_child) {
+			if (!child->d_inode)
+				continue;
+
+			spin_lock(&child->d_lock);
+			if (watched)
+				child->d_flags |= DCACHE_FSNOTIFY_PARENT_WATCHED;
+			else
+				child->d_flags &= ~DCACHE_FSNOTIFY_PARENT_WATCHED;
+			spin_unlock(&child->d_lock);
+		}
+	}
+	spin_unlock(&dcache_lock);
+}
+
+/* Notify this dentry's parent about a child's events. */
+void __fsnotify_parent(struct dentry *dentry, __u32 mask)
+{
+	struct dentry *parent;
+	struct inode *p_inode;
+	bool send = false;
+	bool should_update_children = false;
+
+	if (!(dentry->d_flags & DCACHE_FSNOTIFY_PARENT_WATCHED))
+		return;
+
+	spin_lock(&dentry->d_lock);
+	parent = dentry->d_parent;
+	p_inode = parent->d_inode;
+
+	if (fsnotify_inode_watches_children(p_inode)) {
+		if (p_inode->i_fsnotify_mask & mask) {
+			dget(parent);
+			send = true;
+		}
+	} else {
+		/*
+		 * The parent doesn't care about events on it's children but
+		 * at least one child thought it did.  We need to run all the
+		 * children and update their d_flags to let them know p_inode
+		 * doesn't care about them any more.
+		 */
+		dget(parent);
+		should_update_children = true;
+	}
+
+	spin_unlock(&dentry->d_lock);
+
+	if (send) {
+		/* we are notifying a parent so come up with the new mask which
+		 * 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);
+		dput(parent);
+	}
+
+	if (unlikely(should_update_children)) {
+		__fsnotify_update_child_dentry_flags(p_inode);
+		dput(parent);
+	}
+}
+EXPORT_SYMBOL_GPL(__fsnotify_parent);
+
+/*
+ * 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
+ * out to all of the registered fsnotify_group.  Those groups can then use the
+ * notification event in whatever means they feel necessary.
+ */
+void fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is, const char *file_name, u32 cookie)
+{
+	struct fsnotify_group *group;
+	struct fsnotify_event *event = NULL;
+	int idx;
+	/* global tests shouldn't care about events on child only the specific event */
+	__u32 test_mask = (mask & ~FS_EVENT_ON_CHILD);
+
+	if (list_empty(&fsnotify_groups))
+		return;
+
+	if (!(test_mask & fsnotify_mask))
+		return;
+
+	if (!(test_mask & to_tell->i_fsnotify_mask))
+		return;
+	/*
+	 * SRCU!!  the groups list is very very much read only and the path is
+	 * very hot.  The VAST majority of events are not going to need to do
+	 * anything other than walk the list so it's crazy to pre-allocate.
+	 */
+	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))
+				continue;
+			if (!event) {
+				event = fsnotify_create_event(to_tell, mask, data, data_is, file_name, cookie);
+				/* shit, we OOM'd and now we can't tell, maybe
+				 * someday someone else will want to do something
+				 * here */
+				if (!event)
+					break;
+			}
+			group->ops->handle_event(group, event);
+		}
+	}
+	srcu_read_unlock(&fsnotify_grp_srcu, idx);
+	/*
+	 * fsnotify_create_event() took a reference so the event can't be cleaned
+	 * up while we are still trying to add it to lists, drop that one.
+	 */
+	if (event)
+		fsnotify_put_event(event);
+}
+EXPORT_SYMBOL_GPL(fsnotify);
+
+static __init int fsnotify_init(void)
+{
+	return init_srcu_struct(&fsnotify_grp_srcu);
+}
+subsys_initcall(fsnotify_init);
diff --git a/fs/notify/fsnotify.h b/fs/notify/fsnotify.h
new file mode 100644
index 00000000000..4dc240824b2
--- /dev/null
+++ b/fs/notify/fsnotify.h
@@ -0,0 +1,34 @@
+#ifndef __FS_NOTIFY_FSNOTIFY_H_
+#define __FS_NOTIFY_FSNOTIFY_H_
+
+#include <linux/list.h>
+#include <linux/fsnotify.h>
+#include <linux/srcu.h>
+#include <linux/types.h>
+
+/* protects reads of fsnotify_groups */
+extern struct srcu_struct fsnotify_grp_srcu;
+/* all groups which receive fsnotify events */
+extern struct list_head fsnotify_groups;
+/* all bitwise OR of all event types (FS_*) for all fsnotify_groups */
+extern __u32 fsnotify_mask;
+
+/* destroy all events sitting in this groups notification queue */
+extern void fsnotify_flush_notify(struct fsnotify_group *group);
+
+/* final kfree of a group */
+extern void fsnotify_final_destroy_group(struct fsnotify_group *group);
+
+/* run the list of all marks associated with inode and flag them to be freed */
+extern void fsnotify_clear_marks_by_inode(struct inode *inode);
+/*
+ * update the dentry->d_flags of all of inode's children to indicate if inode cares
+ * about events that happen to its children.
+ */
+extern void __fsnotify_update_child_dentry_flags(struct inode *inode);
+
+/* allocate and destroy and event holder to attach events to notification/access queues */
+extern struct fsnotify_event_holder *fsnotify_alloc_event_holder(void);
+extern void fsnotify_destroy_event_holder(struct fsnotify_event_holder *holder);
+
+#endif	/* __FS_NOTIFY_FSNOTIFY_H_ */
diff --git a/fs/notify/group.c b/fs/notify/group.c
new file mode 100644
index 00000000000..0e1677144bc
--- /dev/null
+++ b/fs/notify/group.c
@@ -0,0 +1,254 @@
+/*
+ *  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.
+ */
+
+#include <linux/list.h>
+#include <linux/mutex.h>
+#include <linux/slab.h>
+#include <linux/srcu.h>
+#include <linux/rculist.h>
+#include <linux/wait.h>
+
+#include <linux/fsnotify_backend.h>
+#include "fsnotify.h"
+
+#include <asm/atomic.h>
+
+/* protects writes to fsnotify_groups and fsnotify_mask */
+static DEFINE_MUTEX(fsnotify_grp_mutex);
+/* protects reads while running the fsnotify_groups list */
+struct srcu_struct fsnotify_grp_srcu;
+/* all groups registered to receive filesystem notifications */
+LIST_HEAD(fsnotify_groups);
+/* bitwise OR of all events (FS_*) interesting to some group on this system */
+__u32 fsnotify_mask;
+
+/*
+ * When a new group registers or changes it's set of interesting events
+ * this function updates the fsnotify_mask to contain all interesting events
+ */
+void fsnotify_recalc_global_mask(void)
+{
+	struct fsnotify_group *group;
+	__u32 mask = 0;
+	int idx;
+
+	idx = srcu_read_lock(&fsnotify_grp_srcu);
+	list_for_each_entry_rcu(group, &fsnotify_groups, group_list)
+		mask |= group->mask;
+	srcu_read_unlock(&fsnotify_grp_srcu, idx);
+	fsnotify_mask = mask;
+}
+
+/*
+ * Update the group->mask by running all of the marks associated with this
+ * group and finding the bitwise | of all of the mark->mask.  If we change
+ * the group->mask we need to update the global mask of events interesting
+ * to the system.
+ */
+void fsnotify_recalc_group_mask(struct fsnotify_group *group)
+{
+	__u32 mask = 0;
+	__u32 old_mask = group->mask;
+	struct fsnotify_mark_entry *entry;
+
+	spin_lock(&group->mark_lock);
+	list_for_each_entry(entry, &group->mark_entries, g_list)
+		mask |= entry->mask;
+	spin_unlock(&group->mark_lock);
+
+	group->mask = mask;
+
+	if (old_mask != mask)
+		fsnotify_recalc_global_mask();
+}
+
+/*
+ * 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)
+{
+	/* clear the notification queue of all events */
+	fsnotify_flush_notify(group);
+
+	if (group->ops->free_group_priv)
+		group->ops->free_group_priv(group);
+
+	kfree(group);
+}
+
+/*
+ * Trying to get rid of a group.  We need to first get rid of any outstanding
+ * allocations and then free the group.  Remember that fsnotify_clear_marks_by_group
+ * could miss marks that are being freed by inode and those marks could still
+ * hold a reference to this group (via group->num_marks)  If we get into that
+ * situtation, the fsnotify_final_destroy_group will get called when that final
+ * mark is freed.
+ */
+static void fsnotify_destroy_group(struct fsnotify_group *group)
+{
+	/* clear all inode mark entries for this group */
+	fsnotify_clear_marks_by_group(group);
+
+	/* past the point of no return, matches the initial value of 1 */
+	if (atomic_dec_and_test(&group->num_marks))
+		fsnotify_final_destroy_group(group);
+}
+
+/*
+ * Remove this group from the global list of groups that will get events
+ * this can be done even if there are still references and things still using
+ * this group.  This just stops the group from getting new events.
+ */
+static void __fsnotify_evict_group(struct fsnotify_group *group)
+{
+	BUG_ON(!mutex_is_locked(&fsnotify_grp_mutex));
+
+	if (group->on_group_list)
+		list_del_rcu(&group->group_list);
+	group->on_group_list = 0;
+}
+
+/*
+ * Called when a group is no longer interested in getting events.  This can be
+ * used if a group is misbehaving or if for some reason a group should no longer
+ * get any filesystem events.
+ */
+void fsnotify_evict_group(struct fsnotify_group *group)
+{
+	mutex_lock(&fsnotify_grp_mutex);
+	__fsnotify_evict_group(group);
+	mutex_unlock(&fsnotify_grp_mutex);
+}
+
+/*
+ * Drop a reference to a group.  Free it if it's through.
+ */
+void fsnotify_put_group(struct fsnotify_group *group)
+{
+	if (!atomic_dec_and_mutex_lock(&group->refcnt, &fsnotify_grp_mutex))
+		return;
+
+	/*
+	 * OK, now we know that there's no other users *and* we hold mutex,
+	 * so no new references will appear
+	 */
+	__fsnotify_evict_group(group);
+
+	/*
+	 * now it's off the list, so the only thing we might care about is
+	 * srcu access....
+	 */
+	mutex_unlock(&fsnotify_grp_mutex);
+	synchronize_srcu(&fsnotify_grp_srcu);
+
+	/* and now it is really dead. _Nothing_ could be seeing it */
+	fsnotify_recalc_global_mask();
+	fsnotify_destroy_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,
+					     const struct fsnotify_ops *ops)
+{
+	struct fsnotify_group *group, *tgroup;
+
+	/* very low use, simpler locking if we just always alloc */
+	group = kmalloc(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);
+	INIT_LIST_HEAD(&group->notification_list);
+	init_waitqueue_head(&group->notification_waitq);
+	group->q_len = 0;
+	group->max_events = UINT_MAX;
+
+	spin_lock_init(&group->mark_lock);
+	atomic_set(&group->num_marks, 0);
+	INIT_LIST_HEAD(&group->mark_entries);
+
+	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);
+	group->on_group_list = 1;
+	/* being on the fsnotify_groups list holds one num_marks */
+	atomic_inc(&group->num_marks);
+
+	mutex_unlock(&fsnotify_grp_mutex);
+
+	if (mask)
+		fsnotify_recalc_global_mask();
+
+	return group;
+}
diff --git a/fs/notify/inode_mark.c b/fs/notify/inode_mark.c
new file mode 100644
index 00000000000..c8a07c65482
--- /dev/null
+++ b/fs/notify/inode_mark.c
@@ -0,0 +1,426 @@
+/*
+ *  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.
+ */
+
+/*
+ * fsnotify inode mark locking/lifetime/and refcnting
+ *
+ * REFCNT:
+ * The mark->refcnt tells how many "things" in the kernel currently are
+ * referencing this object.  The object typically will live inside the kernel
+ * with a refcnt of 2, one for each list it is on (i_list, g_list).  Any task
+ * which can find this object holding the appropriete locks, can take a reference
+ * and the object itself is guarenteed to survive until the reference is dropped.
+ *
+ * LOCKING:
+ * There are 3 spinlocks involved with fsnotify inode marks and they MUST
+ * be taken in order as follows:
+ *
+ * entry->lock
+ * group->mark_lock
+ * inode->i_lock
+ *
+ * entry->lock protects 2 things, entry->group and entry->inode.  You must hold
+ * that lock to dereference either of these things (they could be NULL even with
+ * the lock)
+ *
+ * group->mark_lock protects the mark_entries list anchored inside a given group
+ * and each entry is hooked via the g_list.  It also sorta protects the
+ * free_g_list, which when used is anchored by a private list on the stack of the
+ * task which held the group->mark_lock.
+ *
+ * inode->i_lock protects the i_fsnotify_mark_entries list anchored inside a
+ * given inode and each entry is hooked via the i_list. (and sorta the
+ * free_i_list)
+ *
+ *
+ * LIFETIME:
+ * Inode marks survive between when they are added to an inode and when their
+ * refcnt==0.
+ *
+ * The inode mark can be cleared for a number of different reasons including:
+ * - The inode is unlinked for the last time.  (fsnotify_inode_remove)
+ * - The inode is being evicted from cache. (fsnotify_inode_delete)
+ * - The fs the inode is on is unmounted.  (fsnotify_inode_delete/fsnotify_unmount_inodes)
+ * - Something explicitly requests that it be removed.  (fsnotify_destroy_mark_by_entry)
+ * - The fsnotify_group associated with the mark is going away and all such marks
+ *   need to be cleaned up. (fsnotify_clear_marks_by_group)
+ *
+ * Worst case we are given an inode and need to clean up all the marks on that
+ * inode.  We take i_lock and walk the i_fsnotify_mark_entries safely.  For each
+ * mark on the list we take a reference (so the mark can't disappear under us).
+ * We remove that mark form the inode's list of marks and we add this mark to a
+ * private list anchored on the stack using i_free_list;  At this point we no
+ * longer fear anything finding the mark using the inode's list of marks.
+ *
+ * We can safely and locklessly run the private list on the stack of everything
+ * we just unattached from the original inode.  For each mark on the private list
+ * we grab the mark-> and can thus dereference mark->group and mark->inode.  If
+ * we see the group and inode are not NULL we take those locks.  Now holding all
+ * 3 locks we can completely remove the mark from other tasks finding it in the
+ * future.  Remember, 10 things might already be referencing this mark, but they
+ * better be holding a ref.  We drop our reference we took before we unhooked it
+ * from the inode.  When the ref hits 0 we can free the mark.
+ *
+ * Very similarly for freeing by group, except we use free_g_list.
+ *
+ * This has the very interesting property of being able to run concurrently with
+ * any (or all) other directions.
+ */
+
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+#include <linux/writeback.h> /* for inode_lock */
+
+#include <asm/atomic.h>
+
+#include <linux/fsnotify_backend.h>
+#include "fsnotify.h"
+
+void fsnotify_get_mark(struct fsnotify_mark_entry *entry)
+{
+	atomic_inc(&entry->refcnt);
+}
+
+void fsnotify_put_mark(struct fsnotify_mark_entry *entry)
+{
+	if (atomic_dec_and_test(&entry->refcnt))
+		entry->free_mark(entry);
+}
+
+/*
+ * Recalculate the mask of events relevant to a given inode locked.
+ */
+static void fsnotify_recalc_inode_mask_locked(struct inode *inode)
+{
+	struct fsnotify_mark_entry *entry;
+	struct hlist_node *pos;
+	__u32 new_mask = 0;
+
+	assert_spin_locked(&inode->i_lock);
+
+	hlist_for_each_entry(entry, pos, &inode->i_fsnotify_mark_entries, i_list)
+		new_mask |= entry->mask;
+	inode->i_fsnotify_mask = new_mask;
+}
+
+/*
+ * Recalculate the inode->i_fsnotify_mask, or the mask of all FS_* event types
+ * any notifier is interested in hearing for this inode.
+ */
+void fsnotify_recalc_inode_mask(struct inode *inode)
+{
+	spin_lock(&inode->i_lock);
+	fsnotify_recalc_inode_mask_locked(inode);
+	spin_unlock(&inode->i_lock);
+
+	__fsnotify_update_child_dentry_flags(inode);
+}
+
+/*
+ * Any time a mark is getting freed we end up here.
+ * The caller had better be holding a reference to this mark so we don't actually
+ * do the final put under the entry->lock
+ */
+void fsnotify_destroy_mark_by_entry(struct fsnotify_mark_entry *entry)
+{
+	struct fsnotify_group *group;
+	struct inode *inode;
+
+	spin_lock(&entry->lock);
+
+	group = entry->group;
+	inode = entry->inode;
+
+	BUG_ON(group && !inode);
+	BUG_ON(!group && inode);
+
+	/* if !group something else already marked this to die */
+	if (!group) {
+		spin_unlock(&entry->lock);
+		return;
+	}
+
+	/* 1 from caller and 1 for being on i_list/g_list */
+	BUG_ON(atomic_read(&entry->refcnt) < 2);
+
+	spin_lock(&group->mark_lock);
+	spin_lock(&inode->i_lock);
+
+	hlist_del_init(&entry->i_list);
+	entry->inode = NULL;
+
+	list_del_init(&entry->g_list);
+	entry->group = NULL;
+
+	fsnotify_put_mark(entry); /* for i_list and g_list */
+
+	/*
+	 * this mark is now off the inode->i_fsnotify_mark_entries list and we
+	 * hold the inode->i_lock, so this is the perfect time to update the
+	 * inode->i_fsnotify_mask
+	 */
+	fsnotify_recalc_inode_mask_locked(inode);
+
+	spin_unlock(&inode->i_lock);
+	spin_unlock(&group->mark_lock);
+	spin_unlock(&entry->lock);
+
+	/*
+	 * Some groups like to know that marks are being freed.  This is a
+	 * callback to the group function to let it know that this entry
+	 * is being freed.
+	 */
+	if (group->ops->freeing_mark)
+		group->ops->freeing_mark(entry, group);
+
+	/*
+	 * __fsnotify_update_child_dentry_flags(inode);
+	 *
+	 * I really want to call that, but we can't, we have no idea if the inode
+	 * still exists the second we drop the entry->lock.
+	 *
+	 * The next time an event arrive to this inode from one of it's children
+	 * __fsnotify_parent will see that the inode doesn't care about it's
+	 * children and will update all of these flags then.  So really this
+	 * is just a lazy update (and could be a perf win...)
+	 */
+
+
+	iput(inode);
+
+	/*
+	 * it's possible that this group tried to destroy itself, but this
+	 * this mark was simultaneously being freed by inode.  If that's the
+	 * case, we finish freeing the group here.
+	 */
+	if (unlikely(atomic_dec_and_test(&group->num_marks)))
+		fsnotify_final_destroy_group(group);
+}
+
+/*
+ * Given a group, destroy all of the marks associated with that group.
+ */
+void fsnotify_clear_marks_by_group(struct fsnotify_group *group)
+{
+	struct fsnotify_mark_entry *lentry, *entry;
+	LIST_HEAD(free_list);
+
+	spin_lock(&group->mark_lock);
+	list_for_each_entry_safe(entry, lentry, &group->mark_entries, g_list) {
+		list_add(&entry->free_g_list, &free_list);
+		list_del_init(&entry->g_list);
+		fsnotify_get_mark(entry);
+	}
+	spin_unlock(&group->mark_lock);
+
+	list_for_each_entry_safe(entry, lentry, &free_list, free_g_list) {
+		fsnotify_destroy_mark_by_entry(entry);
+		fsnotify_put_mark(entry);
+	}
+}
+
+/*
+ * Given an inode, destroy all of the marks associated with that inode.
+ */
+void fsnotify_clear_marks_by_inode(struct inode *inode)
+{
+	struct fsnotify_mark_entry *entry, *lentry;
+	struct hlist_node *pos, *n;
+	LIST_HEAD(free_list);
+
+	spin_lock(&inode->i_lock);
+	hlist_for_each_entry_safe(entry, pos, n, &inode->i_fsnotify_mark_entries, i_list) {
+		list_add(&entry->free_i_list, &free_list);
+		hlist_del_init(&entry->i_list);
+		fsnotify_get_mark(entry);
+	}
+	spin_unlock(&inode->i_lock);
+
+	list_for_each_entry_safe(entry, lentry, &free_list, free_i_list) {
+		fsnotify_destroy_mark_by_entry(entry);
+		fsnotify_put_mark(entry);
+	}
+}
+
+/*
+ * given a group and inode, find the mark associated with that combination.
+ * if found take a reference to that mark and return it, else return NULL
+ */
+struct fsnotify_mark_entry *fsnotify_find_mark_entry(struct fsnotify_group *group,
+						     struct inode *inode)
+{
+	struct fsnotify_mark_entry *entry;
+	struct hlist_node *pos;
+
+	assert_spin_locked(&inode->i_lock);
+
+	hlist_for_each_entry(entry, pos, &inode->i_fsnotify_mark_entries, i_list) {
+		if (entry->group == group) {
+			fsnotify_get_mark(entry);
+			return entry;
+		}
+	}
+	return NULL;
+}
+
+/*
+ * 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);
+	INIT_HLIST_NODE(&entry->i_list);
+	entry->group = NULL;
+	entry->mask = 0;
+	entry->inode = NULL;
+	entry->free_mark = free_mark;
+}
+
+/*
+ * Attach an initialized mark entry to a given group and inode.
+ * These marks may be used for the fsnotify backend to determine which
+ * 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_mark_entry *lentry;
+	int ret = 0;
+
+	inode = igrab(inode);
+	if (unlikely(!inode))
+		return -EINVAL;
+
+	/*
+	 * LOCKING ORDER!!!!
+	 * entry->lock
+	 * group->mark_lock
+	 * inode->i_lock
+	 */
+	spin_lock(&entry->lock);
+	spin_lock(&group->mark_lock);
+	spin_lock(&inode->i_lock);
+
+	entry->group = group;
+	entry->inode = inode;
+
+	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);
+
+		fsnotify_get_mark(entry); /* for i_list and g_list */
+
+		atomic_inc(&group->num_marks);
+
+		fsnotify_recalc_inode_mask_locked(inode);
+	}
+
+	spin_unlock(&inode->i_lock);
+	spin_unlock(&group->mark_lock);
+	spin_unlock(&entry->lock);
+
+	if (lentry) {
+		ret = -EEXIST;
+		iput(inode);
+		fsnotify_put_mark(lentry);
+	} else {
+		__fsnotify_update_child_dentry_flags(inode);
+	}
+
+	return ret;
+}
+
+/**
+ * fsnotify_unmount_inodes - an sb is unmounting.  handle any watched inodes.
+ * @list: list of inodes being unmounted (sb->s_inodes)
+ *
+ * Called with inode_lock held, protecting the unmounting super block's list
+ * of inodes, and with iprune_mutex held, keeping shrink_icache_memory() at bay.
+ * We temporarily drop inode_lock, however, and CAN block.
+ */
+void fsnotify_unmount_inodes(struct list_head *list)
+{
+	struct inode *inode, *next_i, *need_iput = NULL;
+
+	list_for_each_entry_safe(inode, next_i, list, i_sb_list) {
+		struct inode *need_iput_tmp;
+
+		/*
+		 * We cannot __iget() an inode in state I_CLEAR, I_FREEING,
+		 * I_WILL_FREE, or I_NEW which is fine because by that point
+		 * the inode cannot have any associated watches.
+		 */
+		if (inode->i_state & (I_CLEAR|I_FREEING|I_WILL_FREE|I_NEW))
+			continue;
+
+		/*
+		 * If i_count is zero, the inode cannot have any watches and
+		 * doing an __iget/iput with MS_ACTIVE clear would actually
+		 * evict all inodes with zero i_count from icache which is
+		 * unnecessarily violent and may in fact be illegal to do.
+		 */
+		if (!atomic_read(&inode->i_count))
+			continue;
+
+		need_iput_tmp = need_iput;
+		need_iput = NULL;
+
+		/* In case fsnotify_inode_delete() drops a reference. */
+		if (inode != need_iput_tmp)
+			__iget(inode);
+		else
+			need_iput_tmp = NULL;
+
+		/* In case the dropping of a reference would nuke next_i. */
+		if ((&next_i->i_sb_list != list) &&
+		    atomic_read(&next_i->i_count) &&
+		    !(next_i->i_state & (I_CLEAR | I_FREEING | I_WILL_FREE))) {
+			__iget(next_i);
+			need_iput = next_i;
+		}
+
+		/*
+		 * We can safely drop inode_lock here because we hold
+		 * references on both inode and next_i.  Also no new inodes
+		 * will be added since the umount has begun.  Finally,
+		 * iprune_mutex keeps shrink_icache_memory() away.
+		 */
+		spin_unlock(&inode_lock);
+
+		if (need_iput_tmp)
+			iput(need_iput_tmp);
+
+		/* for each watch, send FS_UNMOUNT and then remove it */
+		fsnotify(inode, FS_UNMOUNT, inode, FSNOTIFY_EVENT_INODE, NULL, 0);
+
+		fsnotify_inode_delete(inode);
+
+		iput(inode);
+
+		spin_lock(&inode_lock);
+	}
+}
diff --git a/fs/notify/inotify/Kconfig b/fs/notify/inotify/Kconfig
index 44679284102..5356884289a 100644
--- a/fs/notify/inotify/Kconfig
+++ b/fs/notify/inotify/Kconfig
@@ -1,26 +1,30 @@
 config INOTIFY
 	bool "Inotify file change notification support"
-	default y
+	default n
 	---help---
-	  Say Y here to enable inotify support.  Inotify is a file change
-	  notification system and a replacement for dnotify.  Inotify fixes
-	  numerous shortcomings in dnotify and introduces several new features
-	  including multiple file events, one-shot support, and unmount
-	  notification.
+	  Say Y here to enable legacy in kernel inotify support.  Inotify is a
+	  file change notification system.  It is a replacement for dnotify.
+	  This option only provides the legacy inotify in kernel API.  There
+	  are no in tree kernel users of this interface since it is deprecated.
+	  You only need this if you are loading an out of tree kernel module
+	  that uses inotify.
 
 	  For more information, see <file:Documentation/filesystems/inotify.txt>
 
-	  If unsure, say Y.
+	  If unsure, say N.
 
 config INOTIFY_USER
 	bool "Inotify support for userspace"
-	depends on INOTIFY
+	depends on FSNOTIFY
 	default y
 	---help---
 	  Say Y here to enable inotify support for userspace, including the
 	  associated system calls.  Inotify allows monitoring of both files and
 	  directories via a single open fd.  Events are read from the file
 	  descriptor, which is also select()- and poll()-able.
+	  Inotify fixes numerous shortcomings in dnotify and introduces several
+	  new features including multiple file events, one-shot support, and
+	  unmount notification.
 
 	  For more information, see <file:Documentation/filesystems/inotify.txt>
 
diff --git a/fs/notify/inotify/Makefile b/fs/notify/inotify/Makefile
index e290f3bb9d8..94382817136 100644
--- a/fs/notify/inotify/Makefile
+++ b/fs/notify/inotify/Makefile
@@ -1,2 +1,2 @@
 obj-$(CONFIG_INOTIFY)		+= inotify.o
-obj-$(CONFIG_INOTIFY_USER)	+= inotify_user.o
+obj-$(CONFIG_INOTIFY_USER)	+= inotify_fsnotify.o inotify_user.o
diff --git a/fs/notify/inotify/inotify.c b/fs/notify/inotify/inotify.c
index 220c13f0d73..40b1cf914cc 100644
--- a/fs/notify/inotify/inotify.c
+++ b/fs/notify/inotify/inotify.c
@@ -32,6 +32,7 @@
 #include <linux/list.h>
 #include <linux/writeback.h>
 #include <linux/inotify.h>
+#include <linux/fsnotify_backend.h>
 
 static atomic_t inotify_cookie;
 
@@ -905,6 +906,25 @@ EXPORT_SYMBOL_GPL(inotify_rm_watch);
  */
 static int __init inotify_setup(void)
 {
+	BUILD_BUG_ON(IN_ACCESS != FS_ACCESS);
+	BUILD_BUG_ON(IN_MODIFY != FS_MODIFY);
+	BUILD_BUG_ON(IN_ATTRIB != FS_ATTRIB);
+	BUILD_BUG_ON(IN_CLOSE_WRITE != FS_CLOSE_WRITE);
+	BUILD_BUG_ON(IN_CLOSE_NOWRITE != FS_CLOSE_NOWRITE);
+	BUILD_BUG_ON(IN_OPEN != FS_OPEN);
+	BUILD_BUG_ON(IN_MOVED_FROM != FS_MOVED_FROM);
+	BUILD_BUG_ON(IN_MOVED_TO != FS_MOVED_TO);
+	BUILD_BUG_ON(IN_CREATE != FS_CREATE);
+	BUILD_BUG_ON(IN_DELETE != FS_DELETE);
+	BUILD_BUG_ON(IN_DELETE_SELF != FS_DELETE_SELF);
+	BUILD_BUG_ON(IN_MOVE_SELF != FS_MOVE_SELF);
+	BUILD_BUG_ON(IN_Q_OVERFLOW != FS_Q_OVERFLOW);
+
+	BUILD_BUG_ON(IN_UNMOUNT != FS_UNMOUNT);
+	BUILD_BUG_ON(IN_ISDIR != FS_IN_ISDIR);
+	BUILD_BUG_ON(IN_IGNORED != FS_IN_IGNORED);
+	BUILD_BUG_ON(IN_ONESHOT != FS_IN_ONESHOT);
+
 	atomic_set(&inotify_cookie, 0);
 
 	return 0;
diff --git a/fs/notify/inotify/inotify.h b/fs/notify/inotify/inotify.h
new file mode 100644
index 00000000000..ea2605a58b8
--- /dev/null
+++ b/fs/notify/inotify/inotify.h
@@ -0,0 +1,21 @@
+#include <linux/fsnotify_backend.h>
+#include <linux/inotify.h>
+#include <linux/slab.h> /* struct kmem_cache */
+
+extern struct kmem_cache *event_priv_cachep;
+
+struct inotify_event_private_data {
+	struct fsnotify_event_private_data fsnotify_event_priv_data;
+	int wd;
+};
+
+struct inotify_inode_mark_entry {
+	/* fsnotify_mark_entry MUST be the first thing */
+	struct fsnotify_mark_entry fsn_entry;
+	int wd;
+};
+
+extern void inotify_destroy_mark_entry(struct fsnotify_mark_entry *entry, struct fsnotify_group *group);
+extern void inotify_free_event_priv(struct fsnotify_event_private_data *event_priv);
+
+extern const struct fsnotify_ops inotify_fsnotify_ops;
diff --git a/fs/notify/inotify/inotify_fsnotify.c b/fs/notify/inotify/inotify_fsnotify.c
new file mode 100644
index 00000000000..7ef75b83247
--- /dev/null
+++ b/fs/notify/inotify/inotify_fsnotify.c
@@ -0,0 +1,138 @@
+/*
+ * fs/inotify_user.c - inotify support for userspace
+ *
+ * Authors:
+ *	John McCutchan	<ttb@tentacle.dhs.org>
+ *	Robert Love	<rml@novell.com>
+ *
+ * Copyright (C) 2005 John McCutchan
+ * Copyright 2006 Hewlett-Packard Development Company, L.P.
+ *
+ * Copyright (C) 2009 Eric Paris <Red Hat Inc>
+ * inotify was largely rewriten to make use of the fsnotify infrastructure
+ *
+ * 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.
+ */
+
+#include <linux/fs.h> /* struct inode */
+#include <linux/fsnotify_backend.h>
+#include <linux/inotify.h>
+#include <linux/path.h> /* struct path */
+#include <linux/slab.h> /* kmem_* */
+#include <linux/types.h>
+
+#include "inotify.h"
+
+static int inotify_handle_event(struct fsnotify_group *group, struct fsnotify_event *event)
+{
+	struct fsnotify_mark_entry *entry;
+	struct inotify_inode_mark_entry *ientry;
+	struct inode *to_tell;
+	struct inotify_event_private_data *event_priv;
+	struct fsnotify_event_private_data *fsn_event_priv;
+	int wd, ret;
+
+	to_tell = event->to_tell;
+
+	spin_lock(&to_tell->i_lock);
+	entry = fsnotify_find_mark_entry(group, to_tell);
+	spin_unlock(&to_tell->i_lock);
+	/* race with watch removal?  We already passes should_send */
+	if (unlikely(!entry))
+		return 0;
+	ientry = container_of(entry, struct inotify_inode_mark_entry,
+			      fsn_entry);
+	wd = ientry->wd;
+
+	event_priv = kmem_cache_alloc(event_priv_cachep, GFP_KERNEL);
+	if (unlikely(!event_priv))
+		return -ENOMEM;
+
+	fsn_event_priv = &event_priv->fsnotify_event_priv_data;
+
+	fsn_event_priv->group = group;
+	event_priv->wd = wd;
+
+	ret = fsnotify_add_notify_event(group, event, fsn_event_priv);
+	/* EEXIST is not an error */
+	if (ret == -EEXIST)
+		ret = 0;
+
+	/* did event_priv get attached? */
+	if (list_empty(&fsn_event_priv->event_list))
+		inotify_free_event_priv(fsn_event_priv);
+
+	/*
+	 * If we hold the entry until after the event is on the queue
+	 * IN_IGNORED won't be able to pass this event in the queue
+	 */
+	fsnotify_put_mark(entry);
+
+	return ret;
+}
+
+static void inotify_freeing_mark(struct fsnotify_mark_entry *entry, struct fsnotify_group *group)
+{
+	inotify_destroy_mark_entry(entry, group);
+}
+
+static bool inotify_should_send_event(struct fsnotify_group *group, struct inode *inode, __u32 mask)
+{
+	struct fsnotify_mark_entry *entry;
+	bool send;
+
+	spin_lock(&inode->i_lock);
+	entry = fsnotify_find_mark_entry(group, inode);
+	spin_unlock(&inode->i_lock);
+	if (!entry)
+		return false;
+
+	mask = (mask & ~FS_EVENT_ON_CHILD);
+	send = (entry->mask & mask);
+
+	/* find took a reference */
+	fsnotify_put_mark(entry);
+
+	return send;
+}
+
+static int idr_callback(int id, void *p, void *data)
+{
+	BUG();
+	return 0;
+}
+
+static void inotify_free_group_priv(struct fsnotify_group *group)
+{
+	/* ideally the idr is empty and we won't hit the BUG in teh callback */
+	idr_for_each(&group->inotify_data.idr, idr_callback, NULL);
+	idr_remove_all(&group->inotify_data.idr);
+	idr_destroy(&group->inotify_data.idr);
+}
+
+void inotify_free_event_priv(struct fsnotify_event_private_data *fsn_event_priv)
+{
+	struct inotify_event_private_data *event_priv;
+
+
+	event_priv = container_of(fsn_event_priv, struct inotify_event_private_data,
+				  fsnotify_event_priv_data);
+
+	kmem_cache_free(event_priv_cachep, event_priv);
+}
+
+const struct fsnotify_ops inotify_fsnotify_ops = {
+	.handle_event = inotify_handle_event,
+	.should_send_event = inotify_should_send_event,
+	.free_group_priv = inotify_free_group_priv,
+	.free_event_priv = inotify_free_event_priv,
+	.freeing_mark = inotify_freeing_mark,
+};
diff --git a/fs/notify/inotify/inotify_user.c b/fs/notify/inotify/inotify_user.c
index 1634319e240..982a412ac5b 100644
--- a/fs/notify/inotify/inotify_user.c
+++ b/fs/notify/inotify/inotify_user.c
@@ -8,6 +8,9 @@
  * Copyright (C) 2005 John McCutchan
  * Copyright 2006 Hewlett-Packard Development Company, L.P.
  *
+ * Copyright (C) 2009 Eric Paris <Red Hat Inc>
+ * inotify was largely rewriten to make use of the fsnotify infrastructure
+ *
  * 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
@@ -19,94 +22,48 @@
  * General Public License for more details.
  */
 
-#include <linux/kernel.h>
-#include <linux/sched.h>
-#include <linux/slab.h>
-#include <linux/fs.h>
 #include <linux/file.h>
-#include <linux/mount.h>
-#include <linux/namei.h>
-#include <linux/poll.h>
-#include <linux/init.h>
-#include <linux/list.h>
+#include <linux/fs.h> /* struct inode */
+#include <linux/fsnotify_backend.h>
+#include <linux/idr.h>
+#include <linux/init.h> /* module_init */
 #include <linux/inotify.h>
+#include <linux/kernel.h> /* roundup() */
+#include <linux/magic.h> /* superblock magic number */
+#include <linux/mount.h> /* mntget */
+#include <linux/namei.h> /* LOOKUP_FOLLOW */
+#include <linux/path.h> /* struct path */
+#include <linux/sched.h> /* struct user */
+#include <linux/slab.h> /* struct kmem_cache */
 #include <linux/syscalls.h>
-#include <linux/magic.h>
+#include <linux/types.h>
+#include <linux/uaccess.h>
+#include <linux/poll.h>
+#include <linux/wait.h>
 
-#include <asm/ioctls.h>
+#include "inotify.h"
 
-static struct kmem_cache *watch_cachep __read_mostly;
-static struct kmem_cache *event_cachep __read_mostly;
+#include <asm/ioctls.h>
 
 static struct vfsmount *inotify_mnt __read_mostly;
 
+/* this just sits here and wastes global memory.  used to just pad userspace messages with zeros */
+static struct inotify_event nul_inotify_event;
+
 /* these are configurable via /proc/sys/fs/inotify/ */
 static int inotify_max_user_instances __read_mostly;
-static int inotify_max_user_watches __read_mostly;
 static int inotify_max_queued_events __read_mostly;
+int inotify_max_user_watches __read_mostly;
 
-/*
- * Lock ordering:
- *
- * inotify_dev->up_mutex (ensures we don't re-add the same watch)
- * 	inode->inotify_mutex (protects inode's watch list)
- * 		inotify_handle->mutex (protects inotify_handle's watch list)
- * 			inotify_dev->ev_mutex (protects device's event queue)
- */
+static struct kmem_cache *inotify_inode_mark_cachep __read_mostly;
+struct kmem_cache *event_priv_cachep __read_mostly;
+static struct fsnotify_event *inotify_ignored_event;
 
 /*
- * Lifetimes of the main data structures:
- *
- * inotify_device: Lifetime is managed by reference count, from
- * sys_inotify_init() until release.  Additional references can bump the count
- * via get_inotify_dev() and drop the count via put_inotify_dev().
- *
- * inotify_user_watch: Lifetime is from create_watch() to the receipt of an
- * IN_IGNORED event from inotify, or when using IN_ONESHOT, to receipt of the
- * first event, or to inotify_destroy().
+ * When inotify registers a new group it increments this and uses that
+ * value as an offset to set the fsnotify group "name" and priority.
  */
-
-/*
- * struct inotify_device - represents an inotify instance
- *
- * This structure is protected by the mutex 'mutex'.
- */
-struct inotify_device {
-	wait_queue_head_t 	wq;		/* wait queue for i/o */
-	struct mutex		ev_mutex;	/* protects event queue */
-	struct mutex		up_mutex;	/* synchronizes watch updates */
-	struct list_head 	events;		/* list of queued events */
-	struct user_struct	*user;		/* user who opened this dev */
-	struct inotify_handle	*ih;		/* inotify handle */
-	struct fasync_struct    *fa;            /* async notification */
-	atomic_t		count;		/* reference count */
-	unsigned int		queue_size;	/* size of the queue (bytes) */
-	unsigned int		event_count;	/* number of pending events */
-	unsigned int		max_events;	/* maximum number of events */
-};
-
-/*
- * struct inotify_kernel_event - An inotify event, originating from a watch and
- * queued for user-space.  A list of these is attached to each instance of the
- * device.  In read(), this list is walked and all events that can fit in the
- * buffer are returned.
- *
- * Protected by dev->ev_mutex of the device in which we are queued.
- */
-struct inotify_kernel_event {
-	struct inotify_event	event;	/* the user-space event */
-	struct list_head        list;	/* entry in inotify_device's list */
-	char			*name;	/* filename, if any */
-};
-
-/*
- * struct inotify_user_watch - our version of an inotify_watch, we add
- * a reference to the associated inotify_device.
- */
-struct inotify_user_watch {
-	struct inotify_device	*dev;	/* associated device */
-	struct inotify_watch	wdata;	/* inotify watch data */
-};
+static atomic_t inotify_grp_num;
 
 #ifdef CONFIG_SYSCTL
 
@@ -149,280 +106,36 @@ ctl_table inotify_table[] = {
 };
 #endif /* CONFIG_SYSCTL */
 
-static inline void get_inotify_dev(struct inotify_device *dev)
-{
-	atomic_inc(&dev->count);
-}
-
-static inline void put_inotify_dev(struct inotify_device *dev)
-{
-	if (atomic_dec_and_test(&dev->count)) {
-		atomic_dec(&dev->user->inotify_devs);
-		free_uid(dev->user);
-		kfree(dev);
-	}
-}
-
-/*
- * free_inotify_user_watch - cleans up the watch and its references
- */
-static void free_inotify_user_watch(struct inotify_watch *w)
-{
-	struct inotify_user_watch *watch;
-	struct inotify_device *dev;
-
-	watch = container_of(w, struct inotify_user_watch, wdata);
-	dev = watch->dev;
-
-	atomic_dec(&dev->user->inotify_watches);
-	put_inotify_dev(dev);
-	kmem_cache_free(watch_cachep, watch);
-}
-
-/*
- * kernel_event - create a new kernel event with the given parameters
- *
- * This function can sleep.
- */
-static struct inotify_kernel_event * kernel_event(s32 wd, u32 mask, u32 cookie,
-						  const char *name)
-{
-	struct inotify_kernel_event *kevent;
-
-	kevent = kmem_cache_alloc(event_cachep, GFP_NOFS);
-	if (unlikely(!kevent))
-		return NULL;
-
-	/* we hand this out to user-space, so zero it just in case */
-	memset(&kevent->event, 0, sizeof(struct inotify_event));
-
-	kevent->event.wd = wd;
-	kevent->event.mask = mask;
-	kevent->event.cookie = cookie;
-
-	INIT_LIST_HEAD(&kevent->list);
-
-	if (name) {
-		size_t len, rem, event_size = sizeof(struct inotify_event);
-
-		/*
-		 * We need to pad the filename so as to properly align an
-		 * array of inotify_event structures.  Because the structure is
-		 * small and the common case is a small filename, we just round
-		 * up to the next multiple of the structure's sizeof.  This is
-		 * simple and safe for all architectures.
-		 */
-		len = strlen(name) + 1;
-		rem = event_size - len;
-		if (len > event_size) {
-			rem = event_size - (len % event_size);
-			if (len % event_size == 0)
-				rem = 0;
-		}
-
-		kevent->name = kmalloc(len + rem, GFP_NOFS);
-		if (unlikely(!kevent->name)) {
-			kmem_cache_free(event_cachep, kevent);
-			return NULL;
-		}
-		memcpy(kevent->name, name, len);
-		if (rem)
-			memset(kevent->name + len, 0, rem);
-		kevent->event.len = len + rem;
-	} else {
-		kevent->event.len = 0;
-		kevent->name = NULL;
-	}
-
-	return kevent;
-}
-
-/*
- * inotify_dev_get_event - return the next event in the given dev's queue
- *
- * Caller must hold dev->ev_mutex.
- */
-static inline struct inotify_kernel_event *
-inotify_dev_get_event(struct inotify_device *dev)
-{
-	return list_entry(dev->events.next, struct inotify_kernel_event, list);
-}
-
-/*
- * inotify_dev_get_last_event - return the last event in the given dev's queue
- *
- * Caller must hold dev->ev_mutex.
- */
-static inline struct inotify_kernel_event *
-inotify_dev_get_last_event(struct inotify_device *dev)
+static inline __u32 inotify_arg_to_mask(u32 arg)
 {
-	if (list_empty(&dev->events))
-		return NULL;
-	return list_entry(dev->events.prev, struct inotify_kernel_event, list);
-}
+	__u32 mask;
 
-/*
- * inotify_dev_queue_event - event handler registered with core inotify, adds
- * a new event to the given device
- *
- * Can sleep (calls kernel_event()).
- */
-static void inotify_dev_queue_event(struct inotify_watch *w, u32 wd, u32 mask,
-				    u32 cookie, const char *name,
-				    struct inode *ignored)
-{
-	struct inotify_user_watch *watch;
-	struct inotify_device *dev;
-	struct inotify_kernel_event *kevent, *last;
+	/* everything should accept their own ignored and cares about children */
+	mask = (FS_IN_IGNORED | FS_EVENT_ON_CHILD);
 
-	watch = container_of(w, struct inotify_user_watch, wdata);
-	dev = watch->dev;
+	/* mask off the flags used to open the fd */
+	mask |= (arg & (IN_ALL_EVENTS | IN_ONESHOT));
 
-	mutex_lock(&dev->ev_mutex);
-
-	/* we can safely put the watch as we don't reference it while
-	 * generating the event
-	 */
-	if (mask & IN_IGNORED || w->mask & IN_ONESHOT)
-		put_inotify_watch(w); /* final put */
-
-	/* coalescing: drop this event if it is a dupe of the previous */
-	last = inotify_dev_get_last_event(dev);
-	if (last && last->event.mask == mask && last->event.wd == wd &&
-			last->event.cookie == cookie) {
-		const char *lastname = last->name;
-
-		if (!name && !lastname)
-			goto out;
-		if (name && lastname && !strcmp(lastname, name))
-			goto out;
-	}
-
-	/* the queue overflowed and we already sent the Q_OVERFLOW event */
-	if (unlikely(dev->event_count > dev->max_events))
-		goto out;
-
-	/* if the queue overflows, we need to notify user space */
-	if (unlikely(dev->event_count == dev->max_events))
-		kevent = kernel_event(-1, IN_Q_OVERFLOW, cookie, NULL);
-	else
-		kevent = kernel_event(wd, mask, cookie, name);
-
-	if (unlikely(!kevent))
-		goto out;
-
-	/* queue the event and wake up anyone waiting */
-	dev->event_count++;
-	dev->queue_size += sizeof(struct inotify_event) + kevent->event.len;
-	list_add_tail(&kevent->list, &dev->events);
-	wake_up_interruptible(&dev->wq);
-	kill_fasync(&dev->fa, SIGIO, POLL_IN);
-
-out:
-	mutex_unlock(&dev->ev_mutex);
-}
-
-/*
- * remove_kevent - cleans up the given kevent
- *
- * Caller must hold dev->ev_mutex.
- */
-static void remove_kevent(struct inotify_device *dev,
-			  struct inotify_kernel_event *kevent)
-{
-	list_del(&kevent->list);
-
-	dev->event_count--;
-	dev->queue_size -= sizeof(struct inotify_event) + kevent->event.len;
-}
-
-/*
- * free_kevent - frees the given kevent.
- */
-static void free_kevent(struct inotify_kernel_event *kevent)
-{
-	kfree(kevent->name);
-	kmem_cache_free(event_cachep, kevent);
-}
-
-/*
- * inotify_dev_event_dequeue - destroy an event on the given device
- *
- * Caller must hold dev->ev_mutex.
- */
-static void inotify_dev_event_dequeue(struct inotify_device *dev)
-{
-	if (!list_empty(&dev->events)) {
-		struct inotify_kernel_event *kevent;
-		kevent = inotify_dev_get_event(dev);
-		remove_kevent(dev, kevent);
-		free_kevent(kevent);
-	}
-}
-
-/*
- * find_inode - resolve a user-given path to a specific inode
- */
-static int find_inode(const char __user *dirname, struct path *path,
-		      unsigned flags)
-{
-	int error;
-
-	error = user_path_at(AT_FDCWD, dirname, flags, path);
-	if (error)
-		return error;
-	/* you can only watch an inode if you have read permissions on it */
-	error = inode_permission(path->dentry->d_inode, MAY_READ);
-	if (error)
-		path_put(path);
-	return error;
+	return mask;
 }
 
-/*
- * create_watch - creates a watch on the given device.
- *
- * Callers must hold dev->up_mutex.
- */
-static int create_watch(struct inotify_device *dev, struct inode *inode,
-			u32 mask)
+static inline u32 inotify_mask_to_arg(__u32 mask)
 {
-	struct inotify_user_watch *watch;
-	int ret;
-
-	if (atomic_read(&dev->user->inotify_watches) >=
-			inotify_max_user_watches)
-		return -ENOSPC;
-
-	watch = kmem_cache_alloc(watch_cachep, GFP_KERNEL);
-	if (unlikely(!watch))
-		return -ENOMEM;
-
-	/* save a reference to device and bump the count to make it official */
-	get_inotify_dev(dev);
-	watch->dev = dev;
-
-	atomic_inc(&dev->user->inotify_watches);
-
-	inotify_init_watch(&watch->wdata);
-	ret = inotify_add_watch(dev->ih, &watch->wdata, inode, mask);
-	if (ret < 0)
-		free_inotify_user_watch(&watch->wdata);
-
-	return ret;
+	return mask & (IN_ALL_EVENTS | IN_ISDIR | IN_UNMOUNT | IN_IGNORED |
+		       IN_Q_OVERFLOW);
 }
 
-/* Device Interface */
-
+/* intofiy userspace file descriptor functions */
 static unsigned int inotify_poll(struct file *file, poll_table *wait)
 {
-	struct inotify_device *dev = file->private_data;
+	struct fsnotify_group *group = file->private_data;
 	int ret = 0;
 
-	poll_wait(file, &dev->wq, wait);
-	mutex_lock(&dev->ev_mutex);
-	if (!list_empty(&dev->events))
+	poll_wait(file, &group->notification_waitq, wait);
+	mutex_lock(&group->notification_mutex);
+	if (!fsnotify_notify_queue_is_empty(group))
 		ret = POLLIN | POLLRDNORM;
-	mutex_unlock(&dev->ev_mutex);
+	mutex_unlock(&group->notification_mutex);
 
 	return ret;
 }
@@ -432,26 +145,29 @@ static unsigned int inotify_poll(struct file *file, poll_table *wait)
  * enough to fit in "count". Return an error pointer if
  * not large enough.
  *
- * Called with the device ev_mutex held.
+ * Called with the group->notification_mutex held.
  */
-static struct inotify_kernel_event *get_one_event(struct inotify_device *dev,
-						  size_t count)
+static struct fsnotify_event *get_one_event(struct fsnotify_group *group,
+					    size_t count)
 {
 	size_t event_size = sizeof(struct inotify_event);
-	struct inotify_kernel_event *kevent;
+	struct fsnotify_event *event;
 
-	if (list_empty(&dev->events))
+	if (fsnotify_notify_queue_is_empty(group))
 		return NULL;
 
-	kevent = inotify_dev_get_event(dev);
-	if (kevent->name)
-		event_size += kevent->event.len;
+	event = fsnotify_peek_notify_event(group);
+
+	event_size += roundup(event->name_len, event_size);
 
 	if (event_size > count)
 		return ERR_PTR(-EINVAL);
 
-	remove_kevent(dev, kevent);
-	return kevent;
+	/* held the notification_mutex the whole time, so this is the
+	 * same event we peeked above */
+	fsnotify_remove_notify_event(group);
+
+	return event;
 }
 
 /*
@@ -460,51 +176,90 @@ static struct inotify_kernel_event *get_one_event(struct inotify_device *dev,
  * We already checked that the event size is smaller than the
  * buffer we had in "get_one_event()" above.
  */
-static ssize_t copy_event_to_user(struct inotify_kernel_event *kevent,
+static ssize_t copy_event_to_user(struct fsnotify_group *group,
+				  struct fsnotify_event *event,
 				  char __user *buf)
 {
+	struct inotify_event inotify_event;
+	struct fsnotify_event_private_data *fsn_priv;
+	struct inotify_event_private_data *priv;
 	size_t event_size = sizeof(struct inotify_event);
+	size_t name_len;
+
+	/* we get the inotify watch descriptor from the event private data */
+	spin_lock(&event->lock);
+	fsn_priv = fsnotify_remove_priv_from_event(group, event);
+	spin_unlock(&event->lock);
+
+	if (!fsn_priv)
+		inotify_event.wd = -1;
+	else {
+		priv = container_of(fsn_priv, struct inotify_event_private_data,
+				    fsnotify_event_priv_data);
+		inotify_event.wd = priv->wd;
+		inotify_free_event_priv(fsn_priv);
+	}
+
+	/* round up event->name_len so it is a multiple of event_size */
+	name_len = roundup(event->name_len, event_size);
+	inotify_event.len = name_len;
+
+	inotify_event.mask = inotify_mask_to_arg(event->mask);
+	inotify_event.cookie = event->sync_cookie;
 
-	if (copy_to_user(buf, &kevent->event, event_size))
+	/* send the main event */
+	if (copy_to_user(buf, &inotify_event, event_size))
 		return -EFAULT;
 
-	if (kevent->name) {
-		buf += event_size;
+	buf += event_size;
 
-		if (copy_to_user(buf, kevent->name, kevent->event.len))
+	/*
+	 * fsnotify only stores the pathname, so here we have to send the pathname
+	 * and then pad that pathname out to a multiple of sizeof(inotify_event)
+	 * with zeros.  I get my zeros from the nul_inotify_event.
+	 */
+	if (name_len) {
+		unsigned int len_to_zero = name_len - event->name_len;
+		/* copy the path name */
+		if (copy_to_user(buf, event->file_name, event->name_len))
 			return -EFAULT;
+		buf += event->name_len;
 
-		event_size += kevent->event.len;
+		/* fill userspace with 0's from nul_inotify_event */
+		if (copy_to_user(buf, &nul_inotify_event, len_to_zero))
+			return -EFAULT;
+		buf += len_to_zero;
+		event_size += name_len;
 	}
+
 	return event_size;
 }
 
 static ssize_t inotify_read(struct file *file, char __user *buf,
 			    size_t count, loff_t *pos)
 {
-	struct inotify_device *dev;
+	struct fsnotify_group *group;
+	struct fsnotify_event *kevent;
 	char __user *start;
 	int ret;
 	DEFINE_WAIT(wait);
 
 	start = buf;
-	dev = file->private_data;
+	group = file->private_data;
 
 	while (1) {
-		struct inotify_kernel_event *kevent;
+		prepare_to_wait(&group->notification_waitq, &wait, TASK_INTERRUPTIBLE);
 
-		prepare_to_wait(&dev->wq, &wait, TASK_INTERRUPTIBLE);
-
-		mutex_lock(&dev->ev_mutex);
-		kevent = get_one_event(dev, count);
-		mutex_unlock(&dev->ev_mutex);
+		mutex_lock(&group->notification_mutex);
+		kevent = get_one_event(group, count);
+		mutex_unlock(&group->notification_mutex);
 
 		if (kevent) {
 			ret = PTR_ERR(kevent);
 			if (IS_ERR(kevent))
 				break;
-			ret = copy_event_to_user(kevent, buf);
-			free_kevent(kevent);
+			ret = copy_event_to_user(group, kevent, buf);
+			fsnotify_put_event(kevent);
 			if (ret < 0)
 				break;
 			buf += ret;
@@ -525,7 +280,7 @@ static ssize_t inotify_read(struct file *file, char __user *buf,
 		schedule();
 	}
 
-	finish_wait(&dev->wq, &wait);
+	finish_wait(&group->notification_waitq, &wait);
 	if (start != buf && ret != -EFAULT)
 		ret = buf - start;
 	return ret;
@@ -533,25 +288,19 @@ static ssize_t inotify_read(struct file *file, char __user *buf,
 
 static int inotify_fasync(int fd, struct file *file, int on)
 {
-	struct inotify_device *dev = file->private_data;
+	struct fsnotify_group *group = file->private_data;
 
-	return fasync_helper(fd, file, on, &dev->fa) >= 0 ? 0 : -EIO;
+	return fasync_helper(fd, file, on, &group->inotify_data.fa) >= 0 ? 0 : -EIO;
 }
 
 static int inotify_release(struct inode *ignored, struct file *file)
 {
-	struct inotify_device *dev = file->private_data;
-
-	inotify_destroy(dev->ih);
+	struct fsnotify_group *group = file->private_data;
 
-	/* destroy all of the events on this device */
-	mutex_lock(&dev->ev_mutex);
-	while (!list_empty(&dev->events))
-		inotify_dev_event_dequeue(dev);
-	mutex_unlock(&dev->ev_mutex);
+	fsnotify_clear_marks_by_group(group);
 
-	/* free this device: the put matching the get in inotify_init() */
-	put_inotify_dev(dev);
+	/* free this group, matching get was inotify_init->fsnotify_obtain_group */
+	fsnotify_put_group(group);
 
 	return 0;
 }
@@ -559,16 +308,27 @@ static int inotify_release(struct inode *ignored, struct file *file)
 static long inotify_ioctl(struct file *file, unsigned int cmd,
 			  unsigned long arg)
 {
-	struct inotify_device *dev;
+	struct fsnotify_group *group;
+	struct fsnotify_event_holder *holder;
+	struct fsnotify_event *event;
 	void __user *p;
 	int ret = -ENOTTY;
+	size_t send_len = 0;
 
-	dev = file->private_data;
+	group = file->private_data;
 	p = (void __user *) arg;
 
 	switch (cmd) {
 	case FIONREAD:
-		ret = put_user(dev->queue_size, (int __user *) p);
+		mutex_lock(&group->notification_mutex);
+		list_for_each_entry(holder, &group->notification_list, event_list) {
+			event = holder->event;
+			send_len += sizeof(struct inotify_event);
+			send_len += roundup(event->name_len,
+					     sizeof(struct inotify_event));
+		}
+		mutex_unlock(&group->notification_mutex);
+		ret = put_user(send_len, (int __user *) p);
 		break;
 	}
 
@@ -576,23 +336,233 @@ static long inotify_ioctl(struct file *file, unsigned int cmd,
 }
 
 static const struct file_operations inotify_fops = {
-	.poll           = inotify_poll,
-	.read           = inotify_read,
-	.fasync         = inotify_fasync,
-	.release        = inotify_release,
-	.unlocked_ioctl = inotify_ioctl,
+	.poll		= inotify_poll,
+	.read		= inotify_read,
+	.fasync		= inotify_fasync,
+	.release	= inotify_release,
+	.unlocked_ioctl	= inotify_ioctl,
 	.compat_ioctl	= inotify_ioctl,
 };
 
-static const struct inotify_operations inotify_user_ops = {
-	.handle_event	= inotify_dev_queue_event,
-	.destroy_watch	= free_inotify_user_watch,
-};
 
+/*
+ * find_inode - resolve a user-given path to a specific inode
+ */
+static int inotify_find_inode(const char __user *dirname, struct path *path, unsigned flags)
+{
+	int error;
+
+	error = user_path_at(AT_FDCWD, dirname, flags, path);
+	if (error)
+		return error;
+	/* you can only watch an inode if you have read permissions on it */
+	error = inode_permission(path->dentry->d_inode, MAY_READ);
+	if (error)
+		path_put(path);
+	return error;
+}
+
+/*
+ * When, for whatever reason, inotify is done with a mark (or what used to be a
+ * watch) we need to remove that watch from the idr and we need to send IN_IGNORED
+ * for the given wd.
+ *
+ * There is a bit of recursion here.  The loop looks like:
+ * 	inotify_destroy_mark_entry -> fsnotify_destroy_mark_by_entry ->
+ *	inotify_freeing_mark -> inotify_destory_mark_entry -> restart
+ * But the loop is broken in 2 places.  fsnotify_destroy_mark_by_entry sets
+ * entry->group = NULL before the call to inotify_freeing_mark, so the if (egroup)
+ * test below will not call back to fsnotify again.  But even if that test wasn't
+ * there this would still be safe since fsnotify_destroy_mark_by_entry() is
+ * safe from recursion.
+ */
+void inotify_destroy_mark_entry(struct fsnotify_mark_entry *entry, struct fsnotify_group *group)
+{
+	struct inotify_inode_mark_entry *ientry;
+	struct inotify_event_private_data *event_priv;
+	struct fsnotify_event_private_data *fsn_event_priv;
+	struct fsnotify_group *egroup;
+	struct idr *idr;
+
+	spin_lock(&entry->lock);
+	egroup = entry->group;
+
+	/* if egroup we aren't really done and something might still send events
+	 * for this inode, on the callback we'll send the IN_IGNORED */
+	if (egroup) {
+		spin_unlock(&entry->lock);
+		fsnotify_destroy_mark_by_entry(entry);
+		return;
+	}
+	spin_unlock(&entry->lock);
+
+	ientry = container_of(entry, struct inotify_inode_mark_entry, fsn_entry);
+
+	event_priv = kmem_cache_alloc(event_priv_cachep, GFP_KERNEL);
+	if (unlikely(!event_priv))
+		goto skip_send_ignore;
+
+	fsn_event_priv = &event_priv->fsnotify_event_priv_data;
+
+	fsn_event_priv->group = group;
+	event_priv->wd = ientry->wd;
+
+	fsnotify_add_notify_event(group, inotify_ignored_event, fsn_event_priv);
+
+	/* did the private data get added? */
+	if (list_empty(&fsn_event_priv->event_list))
+		inotify_free_event_priv(fsn_event_priv);
+
+skip_send_ignore:
+
+	/* remove this entry from the idr */
+	spin_lock(&group->inotify_data.idr_lock);
+	idr = &group->inotify_data.idr;
+	idr_remove(idr, ientry->wd);
+	spin_unlock(&group->inotify_data.idr_lock);
+
+	/* removed from idr, drop that reference */
+	fsnotify_put_mark(entry);
+}
+
+/* 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;
+
+	kmem_cache_free(inotify_inode_mark_cachep, ientry);
+}
+
+static int inotify_update_watch(struct fsnotify_group *group, struct inode *inode, u32 arg)
+{
+	struct fsnotify_mark_entry *entry = NULL;
+	struct inotify_inode_mark_entry *ientry;
+	int ret = 0;
+	int add = (arg & IN_MASK_ADD);
+	__u32 mask;
+	__u32 old_mask, new_mask;
+
+	/* don't allow invalid bits: we don't want flags set */
+	mask = inotify_arg_to_mask(arg);
+	if (unlikely(!mask))
+		return -EINVAL;
+
+	ientry = kmem_cache_alloc(inotify_inode_mark_cachep, GFP_KERNEL);
+	if (unlikely(!ientry))
+		return -ENOMEM;
+	/* we set the mask at the end after attaching it */
+	fsnotify_init_mark(&ientry->fsn_entry, inotify_free_mark);
+	ientry->wd = 0;
+
+find_entry:
+	spin_lock(&inode->i_lock);
+	entry = fsnotify_find_mark_entry(group, inode);
+	spin_unlock(&inode->i_lock);
+	if (entry) {
+		kmem_cache_free(inotify_inode_mark_cachep, ientry);
+		ientry = container_of(entry, struct inotify_inode_mark_entry, fsn_entry);
+	} else {
+		if (atomic_read(&group->inotify_data.user->inotify_watches) >= inotify_max_user_watches) {
+			ret = -ENOSPC;
+			goto out_err;
+		}
+
+		ret = fsnotify_add_mark(&ientry->fsn_entry, group, inode);
+		if (ret == -EEXIST)
+			goto find_entry;
+		else if (ret)
+			goto out_err;
+
+		entry = &ientry->fsn_entry;
+retry:
+		ret = -ENOMEM;
+		if (unlikely(!idr_pre_get(&group->inotify_data.idr, GFP_KERNEL)))
+			goto out_err;
+
+		spin_lock(&group->inotify_data.idr_lock);
+		/* if entry is added to the idr we keep the reference obtained
+		 * through fsnotify_mark_add.  remember to drop this reference
+		 * when entry is removed from idr */
+		ret = idr_get_new_above(&group->inotify_data.idr, entry,
+					++group->inotify_data.last_wd,
+					&ientry->wd);
+		spin_unlock(&group->inotify_data.idr_lock);
+		if (ret) {
+			if (ret == -EAGAIN)
+				goto retry;
+			goto out_err;
+		}
+		atomic_inc(&group->inotify_data.user->inotify_watches);
+	}
+
+	spin_lock(&entry->lock);
+
+	old_mask = entry->mask;
+	if (add) {
+		entry->mask |= mask;
+		new_mask = entry->mask;
+	} else {
+		entry->mask = mask;
+		new_mask = entry->mask;
+	}
+
+	spin_unlock(&entry->lock);
+
+	if (old_mask != new_mask) {
+		/* more bits in old than in new? */
+		int dropped = (old_mask & ~new_mask);
+		/* more bits in this entry than the inode's mask? */
+		int do_inode = (new_mask & ~inode->i_fsnotify_mask);
+		/* more bits in this entry than the group? */
+		int do_group = (new_mask & ~group->mask);
+
+		/* update the inode with this new entry */
+		if (dropped || do_inode)
+			fsnotify_recalc_inode_mask(inode);
+
+		/* update the group mask with the new mask */
+		if (dropped || do_group)
+			fsnotify_recalc_group_mask(group);
+	}
+
+	return ientry->wd;
+
+out_err:
+	/* see this isn't supposed to happen, just kill the watch */
+	if (entry) {
+		fsnotify_destroy_mark_by_entry(entry);
+		fsnotify_put_mark(entry);
+	}
+	return ret;
+}
+
+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);
+	if (IS_ERR(group))
+		return group;
+
+	group->max_events = max_events;
+
+	spin_lock_init(&group->inotify_data.idr_lock);
+	idr_init(&group->inotify_data.idr);
+	group->inotify_data.last_wd = 0;
+	group->inotify_data.user = user;
+	group->inotify_data.fa = NULL;
+
+	return group;
+}
+
+
+/* inotify syscalls */
 SYSCALL_DEFINE1(inotify_init1, int, flags)
 {
-	struct inotify_device *dev;
-	struct inotify_handle *ih;
+	struct fsnotify_group *group;
 	struct user_struct *user;
 	struct file *filp;
 	int fd, ret;
@@ -621,45 +591,27 @@ SYSCALL_DEFINE1(inotify_init1, int, flags)
 		goto out_free_uid;
 	}
 
-	dev = kmalloc(sizeof(struct inotify_device), GFP_KERNEL);
-	if (unlikely(!dev)) {
-		ret = -ENOMEM;
+	/* fsnotify_obtain_group took a reference to group, we put this when we kill the file in the end */
+	group = inotify_new_group(user, inotify_max_queued_events);
+	if (IS_ERR(group)) {
+		ret = PTR_ERR(group);
 		goto out_free_uid;
 	}
 
-	ih = inotify_init(&inotify_user_ops);
-	if (IS_ERR(ih)) {
-		ret = PTR_ERR(ih);
-		goto out_free_dev;
-	}
-	dev->ih = ih;
-	dev->fa = NULL;
-
 	filp->f_op = &inotify_fops;
 	filp->f_path.mnt = mntget(inotify_mnt);
 	filp->f_path.dentry = dget(inotify_mnt->mnt_root);
 	filp->f_mapping = filp->f_path.dentry->d_inode->i_mapping;
 	filp->f_mode = FMODE_READ;
 	filp->f_flags = O_RDONLY | (flags & O_NONBLOCK);
-	filp->private_data = dev;
-
-	INIT_LIST_HEAD(&dev->events);
-	init_waitqueue_head(&dev->wq);
-	mutex_init(&dev->ev_mutex);
-	mutex_init(&dev->up_mutex);
-	dev->event_count = 0;
-	dev->queue_size = 0;
-	dev->max_events = inotify_max_queued_events;
-	dev->user = user;
-	atomic_set(&dev->count, 0);
-
-	get_inotify_dev(dev);
+	filp->private_data = group;
+
 	atomic_inc(&user->inotify_devs);
+
 	fd_install(fd, filp);
 
 	return fd;
-out_free_dev:
-	kfree(dev);
+
 out_free_uid:
 	free_uid(user);
 	put_filp(filp);
@@ -676,8 +628,8 @@ SYSCALL_DEFINE0(inotify_init)
 SYSCALL_DEFINE3(inotify_add_watch, int, fd, const char __user *, pathname,
 		u32, mask)
 {
+	struct fsnotify_group *group;
 	struct inode *inode;
-	struct inotify_device *dev;
 	struct path path;
 	struct file *filp;
 	int ret, fput_needed;
@@ -698,20 +650,20 @@ SYSCALL_DEFINE3(inotify_add_watch, int, fd, const char __user *, pathname,
 	if (mask & IN_ONLYDIR)
 		flags |= LOOKUP_DIRECTORY;
 
-	ret = find_inode(pathname, &path, flags);
-	if (unlikely(ret))
+	ret = inotify_find_inode(pathname, &path, flags);
+	if (ret)
 		goto fput_and_out;
 
-	/* inode held in place by reference to path; dev by fget on fd */
+	/* inode held in place by reference to path; group by fget on fd */
 	inode = path.dentry->d_inode;
-	dev = filp->private_data;
+	group = filp->private_data;
 
-	mutex_lock(&dev->up_mutex);
-	ret = inotify_find_update_watch(dev->ih, inode, mask);
-	if (ret == -ENOENT)
-		ret = create_watch(dev, inode, mask);
-	mutex_unlock(&dev->up_mutex);
+	/* create/update an inode mark */
+	ret = inotify_update_watch(group, inode, mask);
+	if (unlikely(ret))
+		goto path_put_and_out;
 
+path_put_and_out:
 	path_put(&path);
 fput_and_out:
 	fput_light(filp, fput_needed);
@@ -720,9 +672,10 @@ fput_and_out:
 
 SYSCALL_DEFINE2(inotify_rm_watch, int, fd, __s32, wd)
 {
+	struct fsnotify_group *group;
+	struct fsnotify_mark_entry *entry;
 	struct file *filp;
-	struct inotify_device *dev;
-	int ret, fput_needed;
+	int ret = 0, fput_needed;
 
 	filp = fget_light(fd, &fput_needed);
 	if (unlikely(!filp))
@@ -734,10 +687,20 @@ SYSCALL_DEFINE2(inotify_rm_watch, int, fd, __s32, wd)
 		goto out;
 	}
 
-	dev = filp->private_data;
+	group = filp->private_data;
 
-	/* we free our watch data when we get IN_IGNORED */
-	ret = inotify_rm_wd(dev->ih, wd);
+	spin_lock(&group->inotify_data.idr_lock);
+	entry = idr_find(&group->inotify_data.idr, wd);
+	if (unlikely(!entry)) {
+		spin_unlock(&group->inotify_data.idr_lock);
+		ret = -EINVAL;
+		goto out;
+	}
+	fsnotify_get_mark(entry);
+	spin_unlock(&group->inotify_data.idr_lock);
+
+	inotify_destroy_mark_entry(entry, group);
+	fsnotify_put_mark(entry);
 
 out:
 	fput_light(filp, fput_needed);
@@ -753,9 +716,9 @@ inotify_get_sb(struct file_system_type *fs_type, int flags,
 }
 
 static struct file_system_type inotify_fs_type = {
-    .name           = "inotifyfs",
-    .get_sb         = inotify_get_sb,
-    .kill_sb        = kill_anon_super,
+    .name	= "inotifyfs",
+    .get_sb	= inotify_get_sb,
+    .kill_sb	= kill_anon_super,
 };
 
 /*
@@ -775,18 +738,16 @@ static int __init inotify_user_setup(void)
 	if (IS_ERR(inotify_mnt))
 		panic("inotify: kern_mount ret %ld!\n", PTR_ERR(inotify_mnt));
 
+	inotify_inode_mark_cachep = KMEM_CACHE(inotify_inode_mark_entry, SLAB_PANIC);
+	event_priv_cachep = KMEM_CACHE(inotify_event_private_data, SLAB_PANIC);
+	inotify_ignored_event = fsnotify_create_event(NULL, FS_IN_IGNORED, NULL, FSNOTIFY_EVENT_NONE, NULL, 0);
+	if (!inotify_ignored_event)
+		panic("unable to allocate the inotify ignored event\n");
+
 	inotify_max_queued_events = 16384;
 	inotify_max_user_instances = 128;
 	inotify_max_user_watches = 8192;
 
-	watch_cachep = kmem_cache_create("inotify_watch_cache",
-					 sizeof(struct inotify_user_watch),
-					 0, SLAB_PANIC, NULL);
-	event_cachep = kmem_cache_create("inotify_event_cache",
-					 sizeof(struct inotify_kernel_event),
-					 0, SLAB_PANIC, NULL);
-
 	return 0;
 }
-
 module_init(inotify_user_setup);
diff --git a/fs/notify/notification.c b/fs/notify/notification.c
new file mode 100644
index 00000000000..959b73e756f
--- /dev/null
+++ b/fs/notify/notification.c
@@ -0,0 +1,411 @@
+/*
+ *  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.
+ */
+
+/*
+ * Basic idea behind the notification queue: An fsnotify group (like inotify)
+ * sends the userspace notification about events asyncronously some time after
+ * the event happened.  When inotify gets an event it will need to add that
+ * event to the group notify queue.  Since a single event might need to be on
+ * multiple group's notification queues we can't add the event directly to each
+ * queue and instead add a small "event_holder" to each queue.  This event_holder
+ * has a pointer back to the original event.  Since the majority of events are
+ * going to end up on one, and only one, notification queue we embed one
+ * event_holder into each event.  This means we have a single allocation instead
+ * of always needing two.  If the embedded event_holder is already in use by
+ * another group a new event_holder (from fsnotify_event_holder_cachep) will be
+ * allocated and used.
+ */
+
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/list.h>
+#include <linux/module.h>
+#include <linux/mount.h>
+#include <linux/mutex.h>
+#include <linux/namei.h>
+#include <linux/path.h>
+#include <linux/slab.h>
+#include <linux/spinlock.h>
+
+#include <asm/atomic.h>
+
+#include <linux/fsnotify_backend.h>
+#include "fsnotify.h"
+
+static struct kmem_cache *fsnotify_event_cachep;
+static struct kmem_cache *fsnotify_event_holder_cachep;
+/*
+ * This is a magic event we send when the q is too full.  Since it doesn't
+ * hold real event information we just keep one system wide and use it any time
+ * 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 atomic_t fsnotify_sync_cookie = ATOMIC_INIT(0);
+
+/**
+ * fsnotify_get_cookie - return a unique cookie for use in synchronizing events.
+ * Called from fsnotify_move, which is inlined into filesystem modules.
+ */
+u32 fsnotify_get_cookie(void)
+{
+	return atomic_inc_return(&fsnotify_sync_cookie);
+}
+EXPORT_SYMBOL_GPL(fsnotify_get_cookie);
+
+/* return true if the notify queue is empty, false otherwise */
+bool fsnotify_notify_queue_is_empty(struct fsnotify_group *group)
+{
+	BUG_ON(!mutex_is_locked(&group->notification_mutex));
+	return list_empty(&group->notification_list) ? true : false;
+}
+
+void fsnotify_get_event(struct fsnotify_event *event)
+{
+	atomic_inc(&event->refcnt);
+}
+
+void fsnotify_put_event(struct fsnotify_event *event)
+{
+	if (!event)
+		return;
+
+	if (atomic_dec_and_test(&event->refcnt)) {
+		if (event->data_type == FSNOTIFY_EVENT_PATH)
+			path_put(&event->path);
+
+		BUG_ON(!list_empty(&event->private_data_list));
+
+		kfree(event->file_name);
+		kmem_cache_free(fsnotify_event_cachep, event);
+	}
+}
+
+struct fsnotify_event_holder *fsnotify_alloc_event_holder(void)
+{
+	return kmem_cache_alloc(fsnotify_event_holder_cachep, GFP_KERNEL);
+}
+
+void fsnotify_destroy_event_holder(struct fsnotify_event_holder *holder)
+{
+	kmem_cache_free(fsnotify_event_holder_cachep, holder);
+}
+
+/*
+ * Find the private data that the group previously attached to this event when
+ * the group added the event to the notification queue (fsnotify_add_notify_event)
+ */
+struct fsnotify_event_private_data *fsnotify_remove_priv_from_event(struct fsnotify_group *group, struct fsnotify_event *event)
+{
+	struct fsnotify_event_private_data *lpriv;
+	struct fsnotify_event_private_data *priv = NULL;
+
+	assert_spin_locked(&event->lock);
+
+	list_for_each_entry(lpriv, &event->private_data_list, event_list) {
+		if (lpriv->group == group) {
+			priv = lpriv;
+			list_del(&priv->event_list);
+			break;
+		}
+	}
+	return priv;
+}
+
+/*
+ * 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)) {
+		switch (old->data_type) {
+		case (FSNOTIFY_EVENT_INODE):
+			if (old->inode == new->inode)
+				return true;
+			break;
+		case (FSNOTIFY_EVENT_PATH):
+			if ((old->path.mnt == new->path.mnt) &&
+			    (old->path.dentry == new->path.dentry))
+				return true;
+		case (FSNOTIFY_EVENT_NONE):
+			return true;
+		};
+	}
+	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_holder *holder = NULL;
+	struct list_head *list = &group->notification_list;
+	struct fsnotify_event_holder *last_holder;
+	struct fsnotify_event *last_event;
+
+	/* easy to tell if priv was attached to the event */
+	INIT_LIST_HEAD(&priv->event_list);
+
+	/*
+	 * There is one fsnotify_event_holder embedded inside each fsnotify_event.
+	 * Check if we expect to be able to use that holder.  If not alloc a new
+	 * holder.
+	 * For the overflow event it's possible that something will use the in
+	 * event holder before we get the lock so we may need to jump back and
+	 * alloc a new holder, this can't happen for most events...
+	 */
+	if (!list_empty(&event->holder.event_list)) {
+alloc_holder:
+		holder = fsnotify_alloc_event_holder();
+		if (!holder)
+			return -ENOMEM;
+	}
+
+	mutex_lock(&group->notification_mutex);
+
+	if (group->q_len >= group->max_events) {
+		event = &q_overflow_event;
+		/* sorry, no private data on the overflow event */
+		priv = NULL;
+	}
+
+	spin_lock(&event->lock);
+
+	if (list_empty(&event->holder.event_list)) {
+		if (unlikely(holder))
+			fsnotify_destroy_event_holder(holder);
+		holder = &event->holder;
+	} else if (unlikely(!holder)) {
+		/* between the time we checked above and got the lock the in
+		 * event holder was used, go back and get a new one */
+		spin_unlock(&event->lock);
+		mutex_unlock(&group->notification_mutex);
+		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;
+
+	fsnotify_get_event(event);
+	list_add_tail(&holder->event_list, list);
+	if (priv)
+		list_add_tail(&priv->event_list, &event->private_data_list);
+	spin_unlock(&event->lock);
+	mutex_unlock(&group->notification_mutex);
+
+	wake_up(&group->notification_waitq);
+	return 0;
+}
+
+/*
+ * Remove and return the first event from the notification list.  There is a
+ * reference held on this event since it was on the list.  It is the responsibility
+ * of the caller to drop this reference.
+ */
+struct fsnotify_event *fsnotify_remove_notify_event(struct fsnotify_group *group)
+{
+	struct fsnotify_event *event;
+	struct fsnotify_event_holder *holder;
+
+	BUG_ON(!mutex_is_locked(&group->notification_mutex));
+
+	holder = list_first_entry(&group->notification_list, struct fsnotify_event_holder, event_list);
+
+	event = holder->event;
+
+	spin_lock(&event->lock);
+	holder->event = NULL;
+	list_del_init(&holder->event_list);
+	spin_unlock(&event->lock);
+
+	/* event == holder means we are referenced through the in event holder */
+	if (holder != &event->holder)
+		fsnotify_destroy_event_holder(holder);
+
+	group->q_len--;
+
+	return event;
+}
+
+/*
+ * This will not remove the event, that must be done with fsnotify_remove_notify_event()
+ */
+struct fsnotify_event *fsnotify_peek_notify_event(struct fsnotify_group *group)
+{
+	struct fsnotify_event *event;
+	struct fsnotify_event_holder *holder;
+
+	BUG_ON(!mutex_is_locked(&group->notification_mutex));
+
+	holder = list_first_entry(&group->notification_list, struct fsnotify_event_holder, event_list);
+	event = holder->event;
+
+	return event;
+}
+
+/*
+ * Called when a group is being torn down to clean up any outstanding
+ * event notifications.
+ */
+void fsnotify_flush_notify(struct fsnotify_group *group)
+{
+	struct fsnotify_event *event;
+	struct fsnotify_event_private_data *priv;
+
+	mutex_lock(&group->notification_mutex);
+	while (!fsnotify_notify_queue_is_empty(group)) {
+		event = fsnotify_remove_notify_event(group);
+		/* if they don't implement free_event_priv they better not have attached any */
+		if (group->ops->free_event_priv) {
+			spin_lock(&event->lock);
+			priv = fsnotify_remove_priv_from_event(group, event);
+			spin_unlock(&event->lock);
+			if (priv)
+				group->ops->free_event_priv(priv);
+		}
+		fsnotify_put_event(event); /* matches fsnotify_add_notify_event */
+	}
+	mutex_unlock(&group->notification_mutex);
+}
+
+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);
+
+	event->to_tell = NULL;
+
+	event->file_name = NULL;
+	event->name_len = 0;
+
+	event->sync_cookie = 0;
+}
+
+/*
+ * fsnotify_create_event - Allocate a new event which will be sent to each
+ * group's handle_event function if the group was interested in this
+ * particular event.
+ *
+ * @to_tell the inode which is supposed to receive the event (sometimes a
+ *	parent of the inode to which the event happened.
+ * @mask what actually happened.
+ * @data pointer to the object which was actually affected
+ * @data_type flag indication if the data is a file, path, inode, nothing...
+ * @name the filename, if available
+ */
+struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask, void *data,
+					     int data_type, const char *name, u32 cookie)
+{
+	struct fsnotify_event *event;
+
+	event = kmem_cache_alloc(fsnotify_event_cachep, GFP_KERNEL);
+	if (!event)
+		return NULL;
+
+	initialize_event(event);
+
+	if (name) {
+		event->file_name = kstrdup(name, GFP_KERNEL);
+		if (!event->file_name) {
+			kmem_cache_free(fsnotify_event_cachep, event);
+			return NULL;
+		}
+		event->name_len = strlen(event->file_name);
+	}
+
+	event->sync_cookie = cookie;
+	event->to_tell = to_tell;
+
+	switch (data_type) {
+	case FSNOTIFY_EVENT_FILE: {
+		struct file *file = data;
+		struct path *path = &file->f_path;
+		event->path.dentry = path->dentry;
+		event->path.mnt = path->mnt;
+		path_get(&event->path);
+		event->data_type = FSNOTIFY_EVENT_PATH;
+		break;
+	}
+	case FSNOTIFY_EVENT_PATH: {
+		struct path *path = data;
+		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;
+		event->path.dentry = NULL;
+		event->path.mnt = NULL;
+		break;
+	default:
+		BUG();
+	}
+
+	event->mask = mask;
+
+	return event;
+}
+
+__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;
+
+	return 0;
+}
+subsys_initcall(fsnotify_notification_init);
+