Merge commit 'fsnotify/for-next'

Documentation/feature-removal-schedule.txt

@@ -426,6 +426,13 @@ When:	2.6.33
 Why:	Should be implemented in userspace, policy daemon.
 Who:	Johannes Berg <johannes@sipsolutions.net>
 
+---------------------------
+
+What:	CONFIG_INOTIFY
+When:	2.6.33
+Why:	No known users, fsnotify more generic and more easily maintained.
+Who:	Eric Paris <eparis@redhat.com>
+
 ----------------------------
 
 What:	CONFIG_X86_OLD_MCE

fs/compat.c

@@ -1177,11 +1177,10 @@ out:
 	if (iov != iovstack)
 		kfree(iov);
 	if ((ret + (type == READ)) > 0) {
-		struct dentry *dentry = file->f_path.dentry;
 		if (type == READ)
-			fsnotify_access(dentry);
+			fsnotify_access(file);
 		else
-			fsnotify_modify(dentry);
+			fsnotify_modify(file);
 	}
 	return ret;
 }

fs/exec.c

@@ -128,7 +128,7 @@ SYSCALL_DEFINE1(uselib, const char __user *, library)
 	if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
 		goto exit;
 
-	fsnotify_open(file->f_path.dentry);
+	fsnotify_open(file);
 
 	error = -ENOEXEC;
 	if(file->f_op) {
@@ -663,7 +663,7 @@ struct file *open_exec(const char *name)
 	if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
 		goto exit;
 
-	fsnotify_open(file->f_path.dentry);
+	fsnotify_open(file);
 
 	err = deny_write_access(file);
 	if (err)

fs/nfsd/vfs.c

@@ -951,7 +951,7 @@ nfsd_vfs_read(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
 		nfsdstats.io_read += host_err;
 		*count = host_err;
 		err = 0;
-		fsnotify_access(file->f_path.dentry);
+		fsnotify_access(file);
 	} else 
 		err = nfserrno(host_err);
 out:
@@ -1062,7 +1062,7 @@ nfsd_vfs_write(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file *file,
 		goto out_nfserr;
 	*cnt = host_err;
 	nfsdstats.io_write += host_err;
-	fsnotify_modify(file->f_path.dentry);
+	fsnotify_modify(file);
 
 	/* clear setuid/setgid flag after write */
 	if (inode->i_mode & (S_ISUID | S_ISGID))

fs/notify/dnotify/dnotify.c

@@ -132,7 +132,8 @@ static int dnotify_handle_event(struct fsnotify_group *group,
  * userspace notification for that pair.
  */
 static bool dnotify_should_send_event(struct fsnotify_group *group,
-				      struct inode *inode, __u32 mask)
+				      struct inode *inode, __u32 mask,
+				      void *data, int data_type)
 {
 	struct fsnotify_mark_entry *entry;
 	bool send;
@@ -361,7 +362,7 @@ int fcntl_dirnotify(int fd, struct file *filp, unsigned long arg)
 		dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry);
 		spin_lock(&entry->lock);
 	} else {
-		fsnotify_add_mark(new_entry, dnotify_group, inode);
+		fsnotify_add_mark(new_entry, dnotify_group, inode, 0);
 		spin_lock(&new_entry->lock);
 		entry = new_entry;
 		dnentry = new_dnentry;
@@ -431,8 +432,7 @@ static int __init dnotify_init(void)
 	dnotify_struct_cache = KMEM_CACHE(dnotify_struct, SLAB_PANIC);
 	dnotify_mark_entry_cache = KMEM_CACHE(dnotify_mark_entry, SLAB_PANIC);
 
-	dnotify_group = fsnotify_obtain_group(DNOTIFY_GROUP_NUM,
-					      0, &dnotify_fsnotify_ops);
+	dnotify_group = fsnotify_obtain_group(0, &dnotify_fsnotify_ops);
 	if (IS_ERR(dnotify_group))
 		panic("unable to allocate fsnotify group for dnotify\n");
 	return 0;

fs/notify/fsnotify.c

@@ -77,13 +77,16 @@ void __fsnotify_update_child_dentry_flags(struct inode *inode)
 }
 
 /* Notify this dentry's parent about a child's events. */
-void __fsnotify_parent(struct dentry *dentry, __u32 mask)
+void __fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask)
 {
 	struct dentry *parent;
 	struct inode *p_inode;
 	bool send = false;
 	bool should_update_children = false;
 
+	if (file)
+		dentry = file->f_path.dentry;
+
 	if (!(dentry->d_flags & DCACHE_FSNOTIFY_PARENT_WATCHED))
 		return;
 
@@ -114,8 +117,12 @@ void __fsnotify_parent(struct dentry *dentry, __u32 mask)
 		 * specifies these are events which came from a child. */
 		mask |= FS_EVENT_ON_CHILD;
 
-		fsnotify(p_inode, mask, dentry->d_inode, FSNOTIFY_EVENT_INODE,
-			 dentry->d_name.name, 0);
+		if (file)
+			fsnotify(p_inode, mask, file, FSNOTIFY_EVENT_FILE,
+				 dentry->d_name.name, 0);
+		else
+			fsnotify(p_inode, mask, dentry->d_inode, FSNOTIFY_EVENT_INODE,
+				 dentry->d_name.name, 0);
 		dput(parent);
 	}
 
@@ -156,7 +163,8 @@ void fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is, const
 	idx = srcu_read_lock(&fsnotify_grp_srcu);
 	list_for_each_entry_rcu(group, &fsnotify_groups, group_list) {
 		if (test_mask & group->mask) {
-			if (!group->ops->should_send_event(group, to_tell, mask))
+			if (!group->ops->should_send_event(group, to_tell, mask,
+							   data, data_is))
 				continue;
 			if (!event) {
 				event = fsnotify_create_event(to_tell, mask, data,

fs/notify/group.c

@@ -77,15 +77,6 @@ void fsnotify_recalc_group_mask(struct fsnotify_group *group)
 		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
  */
@@ -170,51 +161,24 @@ void fsnotify_put_group(struct fsnotify_group *group)
 	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,
+struct fsnotify_group *fsnotify_obtain_group(__u32 mask,
 					     const struct fsnotify_ops *ops)
 {
-	struct fsnotify_group *group, *tgroup;
+	struct fsnotify_group *group;
 
 	/* very low use, simpler locking if we just always alloc */
-	group = kmalloc(sizeof(struct fsnotify_group), GFP_KERNEL);
+	group = kzalloc(sizeof(struct fsnotify_group), GFP_KERNEL);
 	if (!group)
 		return ERR_PTR(-ENOMEM);
 
 	atomic_set(&group->refcnt, 1);
 
 	group->on_group_list = 0;
-	group->group_num = group_num;
 	group->mask = mask;
 
 	mutex_init(&group->notification_mutex);
@@ -230,14 +194,6 @@ struct fsnotify_group *fsnotify_obtain_group(unsigned int group_num, __u32 mask,
 	group->ops = ops;
 
 	mutex_lock(&fsnotify_grp_mutex);
-	tgroup = fsnotify_find_group(group_num, mask, ops);
-	if (tgroup) {
-		/* group already exists */
-		mutex_unlock(&fsnotify_grp_mutex);
-		/* destroy the new one we made */
-		fsnotify_put_group(group);
-		return tgroup;
-	}
 
 	/* group not found, add a new one */
 	list_add_rcu(&group->group_list, &fsnotify_groups);

fs/notify/inode_mark.c

@@ -283,12 +283,20 @@ struct fsnotify_mark_entry *fsnotify_find_mark_entry(struct fsnotify_group *grou
 	return NULL;
 }
 
+void fsnotify_duplicate_mark(struct fsnotify_mark_entry *new, struct fsnotify_mark_entry *old)
+{
+	assert_spin_locked(&old->lock);
+	new->inode = old->inode;
+	new->group = old->group;
+	new->mask = old->mask;
+	new->free_mark = old->free_mark;
+}
+
 /*
  * Nothing fancy, just initialize lists and locks and counters.
  */
 void fsnotify_init_mark(struct fsnotify_mark_entry *entry,
 			void (*free_mark)(struct fsnotify_mark_entry *entry))
-
 {
 	spin_lock_init(&entry->lock);
 	atomic_set(&entry->refcnt, 1);
@@ -305,9 +313,10 @@ void fsnotify_init_mark(struct fsnotify_mark_entry *entry,
  * event types should be delivered to which group and for which inodes.
  */
 int fsnotify_add_mark(struct fsnotify_mark_entry *entry,
-		      struct fsnotify_group *group, struct inode *inode)
+		      struct fsnotify_group *group, struct inode *inode,
+		      int allow_dups)
 {
-	struct fsnotify_mark_entry *lentry;
+	struct fsnotify_mark_entry *lentry = NULL;
 	int ret = 0;
 
 	inode = igrab(inode);
@@ -324,11 +333,12 @@ int fsnotify_add_mark(struct fsnotify_mark_entry *entry,
 	spin_lock(&group->mark_lock);
 	spin_lock(&inode->i_lock);
 
-	entry->group = group;
-	entry->inode = inode;
-
-	lentry = fsnotify_find_mark_entry(group, inode);
+	if (!allow_dups)
+		lentry = fsnotify_find_mark_entry(group, inode);
 	if (!lentry) {
+		entry->group = group;
+		entry->inode = inode;
+
 		hlist_add_head(&entry->i_list, &inode->i_fsnotify_mark_entries);
 		list_add(&entry->g_list, &group->mark_entries);
 

fs/notify/inotify/inotify_fsnotify.c

@@ -31,6 +31,60 @@
 
 #include "inotify.h"
 
+/*
+ * Check if 2 events contain the same information.  We do not compare private data
+ * but at this moment that isn't a problem for any know fsnotify listeners.
+ */
+static bool event_compare(struct fsnotify_event *old, struct fsnotify_event *new)
+{
+	if ((old->mask == new->mask) &&
+	    (old->to_tell == new->to_tell) &&
+	    (old->data_type == new->data_type) &&
+	    (old->name_len == new->name_len)) {
+		switch (old->data_type) {
+		case (FSNOTIFY_EVENT_INODE):
+			/* remember, after old was put on the wait_q we aren't
+			 * allowed to look at the inode any more, only thing
+			 * left to check was if the file_name is the same */
+			if (old->name_len &&
+			    !strcmp(old->file_name, new->file_name))
+				return true;
+			break;
+		case (FSNOTIFY_EVENT_PATH):
+			if ((old->path.mnt == new->path.mnt) &&
+			    (old->path.dentry == new->path.dentry))
+				return true;
+			break;
+		case (FSNOTIFY_EVENT_NONE):
+			if (old->mask & FS_Q_OVERFLOW)
+				return true;
+			else if (old->mask & FS_IN_IGNORED)
+				return false;
+			return true;
+		};
+	}
+	return false;
+}
+
+static int inotify_merge(struct list_head *list, struct fsnotify_event *event)
+{
+	struct fsnotify_event_holder *last_holder;
+	struct fsnotify_event *last_event;
+	int ret = 0;
+
+	/* and the list better be locked by something too */
+	spin_lock(&event->lock);
+
+	last_holder = list_entry(list->prev, struct fsnotify_event_holder, event_list);
+	last_event = last_holder->event;
+	if (event_compare(last_event, event))
+		ret = -EEXIST;
+
+	spin_unlock(&event->lock);
+
+	return ret;
+}
+
 static int inotify_handle_event(struct fsnotify_group *group, struct fsnotify_event *event)
 {
 	struct fsnotify_mark_entry *entry;
@@ -61,7 +115,7 @@ static int inotify_handle_event(struct fsnotify_group *group, struct fsnotify_ev
 	fsn_event_priv->group = group;
 	event_priv->wd = wd;
 
-	ret = fsnotify_add_notify_event(group, event, fsn_event_priv);
+	ret = fsnotify_add_notify_event(group, event, fsn_event_priv, inotify_merge);
 	if (ret) {
 		inotify_free_event_priv(fsn_event_priv);
 		/* EEXIST says we tail matched, EOVERFLOW isn't something
@@ -85,7 +139,8 @@ static void inotify_freeing_mark(struct fsnotify_mark_entry *entry, struct fsnot
 	inotify_ignored_and_remove_idr(entry, group);
 }
 
-static bool inotify_should_send_event(struct fsnotify_group *group, struct inode *inode, __u32 mask)
+static bool inotify_should_send_event(struct fsnotify_group *group, struct inode *inode,
+				      __u32 mask, void *data, int data_type)
 {
 	struct fsnotify_mark_entry *entry;
 	bool send;

fs/notify/notification.c

@@ -56,7 +56,7 @@ static struct kmem_cache *fsnotify_event_holder_cachep;
  * it is needed.  It's refcnt is set 1 at kernel init time and will never
  * get set to 0 so it will never get 'freed'
  */
-static struct fsnotify_event q_overflow_event;
+static struct fsnotify_event *q_overflow_event;
 static atomic_t fsnotify_sync_cookie = ATOMIC_INIT(0);
 
 /**
@@ -104,7 +104,8 @@ struct fsnotify_event_holder *fsnotify_alloc_event_holder(void)
 
 void fsnotify_destroy_event_holder(struct fsnotify_event_holder *holder)
 {
-	kmem_cache_free(fsnotify_event_holder_cachep, holder);
+	if (holder)
+		kmem_cache_free(fsnotify_event_holder_cachep, holder);
 }
 
 /*
@@ -128,54 +129,18 @@ struct fsnotify_event_private_data *fsnotify_remove_priv_from_event(struct fsnot
 	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) &&
-	    (old->name_len == new->name_len)) {
-		switch (old->data_type) {
-		case (FSNOTIFY_EVENT_INODE):
-			/* remember, after old was put on the wait_q we aren't
-			 * allowed to look at the inode any more, only thing
-			 * left to check was if the file_name is the same */
-			if (old->name_len &&
-			    !strcmp(old->file_name, new->file_name))
-				return true;
-			break;
-		case (FSNOTIFY_EVENT_PATH):
-			if ((old->path.mnt == new->path.mnt) &&
-			    (old->path.dentry == new->path.dentry))
-				return true;
-			break;
-		case (FSNOTIFY_EVENT_NONE):
-			if (old->mask & FS_Q_OVERFLOW)
-				return true;
-			else if (old->mask & FS_IN_IGNORED)
-				return false;
-			return false;
-		};
-	}
-	return false;
-}
-
 /*
  * Add an event to the group notification queue.  The group can later pull this
  * event off the queue to deal with.  If the event is successfully added to the
  * group's notification queue, a reference is taken on event.
  */
 int fsnotify_add_notify_event(struct fsnotify_group *group, struct fsnotify_event *event,
-			      struct fsnotify_event_private_data *priv)
+			      struct fsnotify_event_private_data *priv,
+			      int (*merge)(struct list_head *, struct fsnotify_event *))
 {
 	struct fsnotify_event_holder *holder = NULL;
 	struct list_head *list = &group->notification_list;
-	struct fsnotify_event_holder *last_holder;
-	struct fsnotify_event *last_event;
-	int ret = 0;
+	int rc = 0;
 
 	/*
 	 * There is one fsnotify_event_holder embedded inside each fsnotify_event.
@@ -195,12 +160,24 @@ alloc_holder:
 	mutex_lock(&group->notification_mutex);
 
 	if (group->q_len >= group->max_events) {
-		event = &q_overflow_event;
-		ret = -EOVERFLOW;
+		event = q_overflow_event;
+		rc = -EOVERFLOW;
 		/* sorry, no private data on the overflow event */
 		priv = NULL;
 	}
 
+	if (!list_empty(list) && merge) {
+		int ret;
+
+		ret = merge(list, event);
+		if (ret) {
+			mutex_unlock(&group->notification_mutex);
+			if (holder != &event->holder)
+				fsnotify_destroy_event_holder(holder);
+			return ret;
+		}
+	}
+
 	spin_lock(&event->lock);
 
 	if (list_empty(&event->holder.event_list)) {
@@ -215,18 +192,6 @@ alloc_holder:
 		goto alloc_holder;
 	}
 
-	if (!list_empty(list)) {
-		last_holder = list_entry(list->prev, struct fsnotify_event_holder, event_list);
-		last_event = last_holder->event;
-		if (event_compare(last_event, event)) {
-			spin_unlock(&event->lock);
-			mutex_unlock(&group->notification_mutex);
-			if (holder != &event->holder)
-				fsnotify_destroy_event_holder(holder);
-			return -EEXIST;
-		}
-	}
-
 	group->q_len++;
 	holder->event = event;
 
@@ -238,7 +203,7 @@ alloc_holder:
 	mutex_unlock(&group->notification_mutex);
 
 	wake_up(&group->notification_waitq);
-	return ret;
+	return rc;
 }
 
 /*
@@ -314,25 +279,77 @@ void fsnotify_flush_notify(struct fsnotify_group *group)
 
 static void initialize_event(struct fsnotify_event *event)
 {
-	event->holder.event = NULL;
 	INIT_LIST_HEAD(&event->holder.event_list);
 	atomic_set(&event->refcnt, 1);
 
 	spin_lock_init(&event->lock);
 
-	event->path.dentry = NULL;
-	event->path.mnt = NULL;
-	event->inode = NULL;
-	event->data_type = FSNOTIFY_EVENT_NONE;
-
 	INIT_LIST_HEAD(&event->private_data_list);
+}
+
+/*
+ * Caller damn well better be holding whatever mutex is protecting the
+ * old_holder->event_list and the new_event must be a clean event which
+ * cannot be found anywhere else in the kernel.
+ */
+int fsnotify_replace_event(struct fsnotify_event_holder *old_holder,
+			   struct fsnotify_event *new_event)
+{
+	struct fsnotify_event *old_event = old_holder->event;
+	struct fsnotify_event_holder *new_holder = &new_event->holder;
 
-	event->to_tell = NULL;
+	enum event_spinlock_class {
+		SPINLOCK_OLD,
+		SPINLOCK_NEW,
+	};
 
-	event->file_name = NULL;
-	event->name_len = 0;
+	/*
+	 * if the new_event's embedded holder is in use someone
+	 * screwed up and didn't give us a clean new event.
+	 */
+	BUG_ON(!list_empty(&new_holder->event_list));
+
+	spin_lock_nested(&old_event->lock, SPINLOCK_OLD);
+	spin_lock_nested(&new_event->lock, SPINLOCK_NEW);
+
+	new_holder->event = new_event;
+	list_replace_init(&old_holder->event_list, &new_holder->event_list);
+
+	spin_unlock(&new_event->lock);
+	spin_unlock(&old_event->lock);
+
+	/* event == holder means we are referenced through the in event holder */
+	if (old_holder != &old_event->holder)
+		fsnotify_destroy_event_holder(old_holder);
+
+	fsnotify_get_event(new_event); /* on the list take reference */
+	fsnotify_put_event(old_event); /* off the list, drop reference */
+
+	return 0;
+}
+
+struct fsnotify_event *fsnotify_clone_event(struct fsnotify_event *old_event)
+{
+	struct fsnotify_event *event;
 
-	event->sync_cookie = 0;
+	event = kmem_cache_alloc(fsnotify_event_cachep, GFP_KERNEL);
+	if (!event)
+		return NULL;
+
+	memcpy(event, old_event, sizeof(*event));
+	initialize_event(event);
+
+	if (event->name_len) {
+		event->file_name = kstrdup(old_event->file_name, GFP_KERNEL);
+		if (!event->file_name) {
+			kmem_cache_free(fsnotify_event_cachep, event);
+			return NULL;
+		}
+	}
+	if (event->data_type == FSNOTIFY_EVENT_PATH)
+		path_get(&event->path);
+
+	return event;
 }
 
 /*
@@ -353,7 +370,7 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask,
 {
 	struct fsnotify_event *event;
 
-	event = kmem_cache_alloc(fsnotify_event_cachep, gfp);
+	event = kmem_cache_zalloc(fsnotify_event_cachep, gfp);
 	if (!event)
 		return NULL;
 
@@ -370,6 +387,7 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask,
 
 	event->sync_cookie = cookie;
 	event->to_tell = to_tell;
+	event->data_type = data_type;
 
 	switch (data_type) {
 	case FSNOTIFY_EVENT_FILE: {
@@ -386,12 +404,10 @@ struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32 mask,
 		event->path.dentry = path->dentry;
 		event->path.mnt = path->mnt;
 		path_get(&event->path);
-		event->data_type = FSNOTIFY_EVENT_PATH;
 		break;
 	}
 	case FSNOTIFY_EVENT_INODE:
 		event->inode = data;
-		event->data_type = FSNOTIFY_EVENT_INODE;
 		break;
 	case FSNOTIFY_EVENT_NONE:
 		event->inode = NULL;
@@ -412,8 +428,11 @@ __init int fsnotify_notification_init(void)
 	fsnotify_event_cachep = KMEM_CACHE(fsnotify_event, SLAB_PANIC);
 	fsnotify_event_holder_cachep = KMEM_CACHE(fsnotify_event_holder, SLAB_PANIC);
 
-	initialize_event(&q_overflow_event);
-	q_overflow_event.mask = FS_Q_OVERFLOW;
+	q_overflow_event = fsnotify_create_event(NULL, FS_Q_OVERFLOW, NULL,
+						 FSNOTIFY_EVENT_NONE, NULL, 0,
+						 GFP_KERNEL);
+	if (!q_overflow_event)
+		panic("unable to allocate fsnotify q_overflow_event\n");
 
 	return 0;
 }

fs/open.c

@@ -1042,7 +1042,7 @@ long do_sys_open(int dfd, const char __user *filename, int flags, int mode)
 				put_unused_fd(fd);
 				fd = PTR_ERR(f);
 			} else {
-				fsnotify_open(f->f_path.dentry);
+				fsnotify_open(f);
 				fd_install(fd, f);
 			}
 		}

fs/read_write.c

@@ -293,7 +293,7 @@ ssize_t vfs_read(struct file *file, char __user *buf, size_t count, loff_t *pos)
 		else
 			ret = do_sync_read(file, buf, count, pos);
 		if (ret > 0) {
-			fsnotify_access(file->f_path.dentry);
+			fsnotify_access(file);
 			add_rchar(current, ret);
 		}
 		inc_syscr(current);
@@ -348,7 +348,7 @@ ssize_t vfs_write(struct file *file, const char __user *buf, size_t count, loff_
 		else
 			ret = do_sync_write(file, buf, count, pos);
 		if (ret > 0) {
-			fsnotify_modify(file->f_path.dentry);
+			fsnotify_modify(file);
 			add_wchar(current, ret);
 		}
 		inc_syscw(current);
@@ -656,9 +656,9 @@ out:
 		kfree(iov);
 	if ((ret + (type == READ)) > 0) {
 		if (type == READ)
-			fsnotify_access(file->f_path.dentry);
+			fsnotify_access(file);
 		else
-			fsnotify_modify(file->f_path.dentry);
+			fsnotify_modify(file);
 	}
 	return ret;
 }

include/linux/fsnotify.h

@@ -29,9 +29,14 @@ static inline void fsnotify_d_instantiate(struct dentry *entry,
 }
 
 /* Notify this dentry's parent about a child's events. */
-static inline void fsnotify_parent(struct dentry *dentry, __u32 mask)
+static inline void fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask)
 {
-	__fsnotify_parent(dentry, mask);
+	BUG_ON(file && dentry);
+
+	if (file)
+		dentry = file->f_path.dentry;
+
+	__fsnotify_parent(file, dentry, mask);
 
 	inotify_dentry_parent_queue_event(dentry, mask, 0, dentry->d_name.name);
 }
@@ -124,7 +129,7 @@ static inline void fsnotify_nameremove(struct dentry *dentry, int isdir)
 	if (isdir)
 		mask |= FS_IN_ISDIR;
 
-	fsnotify_parent(dentry, mask);
+	fsnotify_parent(NULL, dentry, mask);
 }
 
 /*
@@ -183,9 +188,9 @@ static inline void fsnotify_mkdir(struct inode *inode, struct dentry *dentry)
 /*
  * fsnotify_access - file was read
  */
-static inline void fsnotify_access(struct dentry *dentry)
+static inline void fsnotify_access(struct file *file)
 {
-	struct inode *inode = dentry->d_inode;
+	struct inode *inode = file->f_path.dentry->d_inode;
 	__u32 mask = FS_ACCESS;
 
 	if (S_ISDIR(inode->i_mode))
@@ -193,16 +198,16 @@ static inline void fsnotify_access(struct dentry *dentry)
 
 	inotify_inode_queue_event(inode, mask, 0, NULL, NULL);
 
-	fsnotify_parent(dentry, mask);
-	fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0);
+	fsnotify_parent(file, NULL, mask);
+	fsnotify(inode, mask, file, FSNOTIFY_EVENT_FILE, NULL, 0);
 }
 
 /*
  * fsnotify_modify - file was modified
  */
-static inline void fsnotify_modify(struct dentry *dentry)
+static inline void fsnotify_modify(struct file *file)
 {
-	struct inode *inode = dentry->d_inode;
+	struct inode *inode = file->f_path.dentry->d_inode;
 	__u32 mask = FS_MODIFY;
 
 	if (S_ISDIR(inode->i_mode))
@@ -210,16 +215,16 @@ static inline void fsnotify_modify(struct dentry *dentry)
 
 	inotify_inode_queue_event(inode, mask, 0, NULL, NULL);
 
-	fsnotify_parent(dentry, mask);
-	fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0);
+	fsnotify_parent(file, NULL, mask);
+	fsnotify(inode, mask, file, FSNOTIFY_EVENT_FILE, NULL, 0);
 }
 
 /*
  * fsnotify_open - file was opened
  */
-static inline void fsnotify_open(struct dentry *dentry)
+static inline void fsnotify_open(struct file *file)
 {
-	struct inode *inode = dentry->d_inode;
+	struct inode *inode = file->f_path.dentry->d_inode;
 	__u32 mask = FS_OPEN;
 
 	if (S_ISDIR(inode->i_mode))
@@ -227,8 +232,8 @@ static inline void fsnotify_open(struct dentry *dentry)
 
 	inotify_inode_queue_event(inode, mask, 0, NULL, NULL);
 
-	fsnotify_parent(dentry, mask);
-	fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0);
+	fsnotify_parent(file, NULL, mask);
+	fsnotify(inode, mask, file, FSNOTIFY_EVENT_FILE, NULL, 0);
 }
 
 /*
@@ -236,8 +241,7 @@ static inline void fsnotify_open(struct dentry *dentry)
  */
 static inline void fsnotify_close(struct file *file)
 {
-	struct dentry *dentry = file->f_path.dentry;
-	struct inode *inode = dentry->d_inode;
+	struct inode *inode = file->f_path.dentry->d_inode;
 	fmode_t mode = file->f_mode;
 	__u32 mask = (mode & FMODE_WRITE) ? FS_CLOSE_WRITE : FS_CLOSE_NOWRITE;
 
@@ -246,7 +250,7 @@ static inline void fsnotify_close(struct file *file)
 
 	inotify_inode_queue_event(inode, mask, 0, NULL, NULL);
 
-	fsnotify_parent(dentry, mask);
+	fsnotify_parent(file, NULL, mask);
 	fsnotify(inode, mask, file, FSNOTIFY_EVENT_FILE, NULL, 0);
 }
 
@@ -263,7 +267,7 @@ static inline void fsnotify_xattr(struct dentry *dentry)
 
 	inotify_inode_queue_event(inode, mask, 0, NULL, NULL);
 
-	fsnotify_parent(dentry, mask);
+	fsnotify_parent(NULL, dentry, mask);
 	fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0);
 }
 
@@ -299,7 +303,7 @@ static inline void fsnotify_change(struct dentry *dentry, unsigned int ia_valid)
 			mask |= FS_IN_ISDIR;
 		inotify_inode_queue_event(inode, mask, 0, NULL, NULL);
 
-		fsnotify_parent(dentry, mask);
+		fsnotify_parent(NULL, dentry, mask);
 		fsnotify(inode, mask, inode, FSNOTIFY_EVENT_INODE, NULL, 0);
 	}
 }

include/linux/fsnotify_backend.h

@@ -58,9 +58,7 @@
 				   FS_MOVED_FROM | FS_MOVED_TO | FS_CREATE |\
 				   FS_DELETE)
 
-/* listeners that hard code group numbers near the top */
-#define DNOTIFY_GROUP_NUM	UINT_MAX
-#define INOTIFY_GROUP_NUM	(DNOTIFY_GROUP_NUM-1)
+#define FS_MOVE			(FS_MOVED_FROM | FS_MOVED_TO)
 
 struct fsnotify_group;
 struct fsnotify_event;
@@ -80,7 +78,8 @@ struct fsnotify_event_private_data;
  *		valid group and inode to use to clean up.
  */
 struct fsnotify_ops {
-	bool (*should_send_event)(struct fsnotify_group *group, struct inode *inode, __u32 mask);
+	bool (*should_send_event)(struct fsnotify_group *group, struct inode *inode,
+				  __u32 mask, void *data, int data_type);
 	int (*handle_event)(struct fsnotify_group *group, struct fsnotify_event *event);
 	void (*free_group_priv)(struct fsnotify_group *group);
 	void (*freeing_mark)(struct fsnotify_mark_entry *entry, struct fsnotify_group *group);
@@ -119,7 +118,6 @@ struct fsnotify_group {
 	 * closed.
 	 */
 	atomic_t refcnt;		/* things with interest in this group */
-	unsigned int group_num;		/* simply prevents accidental group collision */
 
 	const struct fsnotify_ops *ops;	/* how this group handles things */
 
@@ -254,7 +252,7 @@ struct fsnotify_mark_entry {
 /* main fsnotify call to send events */
 extern void fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is,
 		     const char *name, u32 cookie);
-extern void __fsnotify_parent(struct dentry *dentry, __u32 mask);
+extern void __fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask);
 extern void __fsnotify_inode_delete(struct inode *inode);
 extern u32 fsnotify_get_cookie(void);
 
@@ -307,8 +305,7 @@ static inline void __fsnotify_d_instantiate(struct dentry *dentry, struct inode
 /* must call when a group changes its ->mask */
 extern void fsnotify_recalc_global_mask(void);
 /* get a reference to an existing or create a new group */
-extern struct fsnotify_group *fsnotify_obtain_group(unsigned int group_num,
-						    __u32 mask,
+extern struct fsnotify_group *fsnotify_obtain_group(__u32 mask,
 						    const struct fsnotify_ops *ops);
 /* run all marks associated with this group and update group->mask */
 extern void fsnotify_recalc_group_mask(struct fsnotify_group *group);
@@ -323,8 +320,10 @@ extern struct fsnotify_event_private_data *fsnotify_remove_priv_from_event(struc
 									   struct fsnotify_event *event);
 
 /* attach the event to the group notification queue */
-extern int fsnotify_add_notify_event(struct fsnotify_group *group, struct fsnotify_event *event,
-				     struct fsnotify_event_private_data *priv);
+extern int fsnotify_add_notify_event(struct fsnotify_group *group,
+				     struct fsnotify_event *event,
+				     struct fsnotify_event_private_data *priv,
+				     int (*merge)(struct list_head *, struct fsnotify_event *));
 /* true if the group notification queue is empty */
 extern bool fsnotify_notify_queue_is_empty(struct fsnotify_group *group);
 /* return, but do not dequeue the first event on the notification queue */
@@ -339,8 +338,10 @@ extern void fsnotify_recalc_inode_mask(struct inode *inode);
 extern void fsnotify_init_mark(struct fsnotify_mark_entry *entry, void (*free_mark)(struct fsnotify_mark_entry *entry));
 /* find (and take a reference) to a mark associated with group and inode */
 extern struct fsnotify_mark_entry *fsnotify_find_mark_entry(struct fsnotify_group *group, struct inode *inode);
+/* copy the values from old into new */
+extern void fsnotify_duplicate_mark(struct fsnotify_mark_entry *new, struct fsnotify_mark_entry *old);
 /* attach the mark to both the group and the inode */
-extern int fsnotify_add_mark(struct fsnotify_mark_entry *entry, struct fsnotify_group *group, struct inode *inode);
+extern int fsnotify_add_mark(struct fsnotify_mark_entry *entry, struct fsnotify_group *group, struct inode *inode, int allow_dups);
 /* given a mark, flag it to be freed when all references are dropped */
 extern void fsnotify_destroy_mark_by_entry(struct fsnotify_mark_entry *entry);
 /* run all the marks in a group, and flag them to be freed */
@@ -354,13 +355,18 @@ extern struct fsnotify_event *fsnotify_create_event(struct inode *to_tell, __u32
 						    void *data, int data_is, const char *name,
 						    u32 cookie, gfp_t gfp);
 
+/* fanotify likes to change events after they are on lists... */
+extern struct fsnotify_event *fsnotify_clone_event(struct fsnotify_event *old_event);
+extern int fsnotify_replace_event(struct fsnotify_event_holder *old_holder,
+				  struct fsnotify_event *new_event);
+
 #else
 
 static inline void fsnotify(struct inode *to_tell, __u32 mask, void *data, int data_is,
 			    const char *name, u32 cookie)
 {}
 
-static inline void __fsnotify_parent(struct dentry *dentry, __u32 mask)
+static inline void __fsnotify_parent(struct file *file, struct dentry *dentry, __u32 mask)
 {}
 
 static inline void __fsnotify_inode_delete(struct inode *inode)

init/Kconfig

@@ -302,13 +302,17 @@ config AUDITSYSCALL
 	help
 	  Enable low-overhead system-call auditing infrastructure that
 	  can be used independently or with another kernel subsystem,
-	  such as SELinux.  To use audit's filesystem watch feature, please
-	  ensure that INOTIFY is configured.
+	  such as SELinux.
+
+config AUDIT_WATCH
+	def_bool y
+	depends on AUDITSYSCALL
+	select FSNOTIFY
 
 config AUDIT_TREE
 	def_bool y
 	depends on AUDITSYSCALL
-	select INOTIFY
+	select FSNOTIFY
 
 menu "RCU Subsystem"
 

kernel/Makefile

@@ -69,10 +69,11 @@ obj-$(CONFIG_IKCONFIG) += configs.o
 obj-$(CONFIG_RESOURCE_COUNTERS) += res_counter.o
 obj-$(CONFIG_STOP_MACHINE) += stop_machine.o
 obj-$(CONFIG_KPROBES_SANITY_TEST) += test_kprobes.o
-obj-$(CONFIG_AUDIT) += audit.o auditfilter.o audit_watch.o
+obj-$(CONFIG_AUDIT) += audit.o auditfilter.o
 obj-$(CONFIG_AUDITSYSCALL) += auditsc.o
-obj-$(CONFIG_GCOV_KERNEL) += gcov/
+obj-$(CONFIG_AUDIT_WATCH) += audit_watch.o
 obj-$(CONFIG_AUDIT_TREE) += audit_tree.o
+obj-$(CONFIG_GCOV_KERNEL) += gcov/
 obj-$(CONFIG_KPROBES) += kprobes.o
 obj-$(CONFIG_KGDB) += kgdb.o
 obj-$(CONFIG_DETECT_SOFTLOCKUP) += softlockup.o

kernel/audit.c

@@ -55,7 +55,6 @@
 #include <net/netlink.h>
 #include <linux/skbuff.h>
 #include <linux/netlink.h>
-#include <linux/inotify.h>
 #include <linux/freezer.h>
 #include <linux/tty.h>
 

kernel/audit.h

@@ -103,21 +103,27 @@ extern struct mutex audit_filter_mutex;
 extern void audit_free_rule_rcu(struct rcu_head *);
 extern struct list_head audit_filter_list[];
 
+extern struct audit_entry *audit_dupe_rule(struct audit_krule *old);
+
 /* audit watch functions */
-extern unsigned long audit_watch_inode(struct audit_watch *watch);
-extern dev_t audit_watch_dev(struct audit_watch *watch);
+#ifdef CONFIG_AUDIT_WATCH
 extern void audit_put_watch(struct audit_watch *watch);
 extern void audit_get_watch(struct audit_watch *watch);
 extern int audit_to_watch(struct audit_krule *krule, char *path, int len, u32 op);
-extern int audit_add_watch(struct audit_krule *krule);
-extern void audit_remove_watch(struct audit_watch *watch);
-extern void audit_remove_watch_rule(struct audit_krule *krule, struct list_head *list);
-extern void audit_inotify_unregister(struct list_head *in_list);
+extern int audit_add_watch(struct audit_krule *krule, struct list_head **list);
+extern void audit_remove_watch_rule(struct audit_krule *krule);
 extern char *audit_watch_path(struct audit_watch *watch);
-extern struct list_head *audit_watch_rules(struct audit_watch *watch);
-
-extern struct audit_entry *audit_dupe_rule(struct audit_krule *old,
-					   struct audit_watch *watch);
+extern int audit_watch_compare(struct audit_watch *watch, unsigned long ino, dev_t dev);
+#else
+#define audit_put_watch(w) {}
+#define audit_get_watch(w) {}
+#define audit_to_watch(k, p, l, o) (-EINVAL)
+#define audit_add_watch(k, l) (-EINVAL)
+#define audit_remove_watch_rule(k) BUG()
+#define audit_watch_path(w) ""
+#define audit_watch_compare(w, i, d) 0
+
+#endif /* CONFIG_AUDIT_WATCH */
 
 #ifdef CONFIG_AUDIT_TREE
 extern struct audit_chunk *audit_tree_lookup(const struct inode *);

kernel/auditfilter.c

@@ -70,6 +70,7 @@ static inline void audit_free_rule(struct audit_entry *e)
 {
 	int i;
 	struct audit_krule *erule = &e->rule;
+
 	/* some rules don't have associated watches */
 	if (erule->watch)
 		audit_put_watch(erule->watch);
@@ -745,8 +746,7 @@ static inline int audit_dupe_lsm_field(struct audit_field *df,
  * rule with the new rule in the filterlist, then free the old rule.
  * The rlist element is undefined; list manipulations are handled apart from
  * the initial copy. */
-struct audit_entry *audit_dupe_rule(struct audit_krule *old,
-				    struct audit_watch *watch)
+struct audit_entry *audit_dupe_rule(struct audit_krule *old)
 {
 	u32 fcount = old->field_count;
 	struct audit_entry *entry;
@@ -768,8 +768,8 @@ struct audit_entry *audit_dupe_rule(struct audit_krule *old,
 	new->prio = old->prio;
 	new->buflen = old->buflen;
 	new->inode_f = old->inode_f;
-	new->watch = NULL;
 	new->field_count = old->field_count;
+
 	/*
 	 * note that we are OK with not refcounting here; audit_match_tree()
 	 * never dereferences tree and we can't get false positives there
@@ -810,9 +810,9 @@ struct audit_entry *audit_dupe_rule(struct audit_krule *old,
 		}
 	}
 
-	if (watch) {
-		audit_get_watch(watch);
-		new->watch = watch;
+	if (old->watch) {
+		audit_get_watch(old->watch);
+		new->watch = old->watch;
 	}
 
 	return entry;
@@ -865,7 +865,7 @@ static inline int audit_add_rule(struct audit_entry *entry)
 	struct audit_watch *watch = entry->rule.watch;
 	struct audit_tree *tree = entry->rule.tree;
 	struct list_head *list;
-	int h, err;
+	int err;
 #ifdef CONFIG_AUDITSYSCALL
 	int dont_count = 0;
 
@@ -888,15 +888,11 @@ static inline int audit_add_rule(struct audit_entry *entry)
 
 	if (watch) {
 		/* audit_filter_mutex is dropped and re-taken during this call */
-		err = audit_add_watch(&entry->rule);
+		err = audit_add_watch(&entry->rule, &list);
 		if (err) {
 			mutex_unlock(&audit_filter_mutex);
 			goto error;
 		}
-		/* entry->rule.watch may have changed during audit_add_watch() */
-		watch = entry->rule.watch;
-		h = audit_hash_ino((u32)audit_watch_inode(watch));
-		list = &audit_inode_hash[h];
 	}
 	if (tree) {
 		err = audit_add_tree_rule(&entry->rule);
@@ -948,7 +944,6 @@ static inline int audit_del_rule(struct audit_entry *entry)
 	struct audit_watch *watch = entry->rule.watch;
 	struct audit_tree *tree = entry->rule.tree;
 	struct list_head *list;
-	LIST_HEAD(inotify_list);
 	int ret = 0;
 #ifdef CONFIG_AUDITSYSCALL
 	int dont_count = 0;
@@ -968,7 +963,7 @@ static inline int audit_del_rule(struct audit_entry *entry)
 	}
 
 	if (e->rule.watch)
-		audit_remove_watch_rule(&e->rule, &inotify_list);
+		audit_remove_watch_rule(&e->rule);
 
 	if (e->rule.tree)
 		audit_remove_tree_rule(&e->rule);
@@ -986,9 +981,6 @@ static inline int audit_del_rule(struct audit_entry *entry)
 #endif
 	mutex_unlock(&audit_filter_mutex);
 
-	if (!list_empty(&inotify_list))
-		audit_inotify_unregister(&inotify_list);
-
 out:
 	if (watch)
 		audit_put_watch(watch); /* match initial get */
@@ -1322,30 +1314,23 @@ static int update_lsm_rule(struct audit_krule *r)
 {
 	struct audit_entry *entry = container_of(r, struct audit_entry, rule);
 	struct audit_entry *nentry;
-	struct audit_watch *watch;
-	struct audit_tree *tree;
 	int err = 0;
 
 	if (!security_audit_rule_known(r))
 		return 0;
 
-	watch = r->watch;
-	tree = r->tree;
-	nentry = audit_dupe_rule(r, watch);
+	nentry = audit_dupe_rule(r);
 	if (IS_ERR(nentry)) {
 		/* save the first error encountered for the
 		 * return value */
 		err = PTR_ERR(nentry);
 		audit_panic("error updating LSM filters");
-		if (watch)
+		if (r->watch)
 			list_del(&r->rlist);
 		list_del_rcu(&entry->list);
 		list_del(&r->list);
 	} else {
-		if (watch) {
-			list_add(&nentry->rule.rlist, audit_watch_rules(watch));
-			list_del(&r->rlist);
-		} else if (tree)
+		if (r->watch || r->tree)
 			list_replace_init(&r->rlist, &nentry->rule.rlist);
 		list_replace_rcu(&entry->list, &nentry->list);
 		list_replace(&r->list, &nentry->rule.list);

kernel/auditsc.c

@@ -168,12 +168,12 @@ struct audit_context {
 	int		    in_syscall;	/* 1 if task is in a syscall */
 	enum audit_state    state, current_state;
 	unsigned int	    serial;     /* serial number for record */
-	struct timespec	    ctime;      /* time of syscall entry */
 	int		    major;      /* syscall number */
+	struct timespec	    ctime;      /* time of syscall entry */
 	unsigned long	    argv[4];    /* syscall arguments */
-	int		    return_valid; /* return code is valid */
 	long		    return_code;/* syscall return code */
 	u64		    prio;
+	int		    return_valid; /* return code is valid */
 	int		    name_count;
 	struct audit_names  names[AUDIT_NAMES];
 	char *		    filterkey;	/* key for rule that triggered record */
@@ -198,8 +198,8 @@ struct audit_context {
 	char		    target_comm[TASK_COMM_LEN];
 
 	struct audit_tree_refs *trees, *first_trees;
-	int tree_count;
 	struct list_head killed_trees;
+	int tree_count;
 
 	int type;
 	union {
@@ -549,9 +549,8 @@ static int audit_filter_rules(struct task_struct *tsk,
 			}
 			break;
 		case AUDIT_WATCH:
-			if (name && audit_watch_inode(rule->watch) != (unsigned long)-1)
-				result = (name->dev == audit_watch_dev(rule->watch) &&
-					  name->ino == audit_watch_inode(rule->watch));
+			if (name)
+				result = audit_watch_compare(rule->watch, name->ino, name->dev);
 			break;
 		case AUDIT_DIR:
 			if (ctx)
@@ -1726,7 +1725,7 @@ static inline void handle_one(const struct inode *inode)
 	struct audit_tree_refs *p;
 	struct audit_chunk *chunk;
 	int count;
-	if (likely(list_empty(&inode->inotify_watches)))
+	if (likely(hlist_empty(&inode->i_fsnotify_mark_entries)))
 		return;
 	context = current->audit_context;
 	p = context->trees;
@@ -1769,7 +1768,7 @@ retry:
 	seq = read_seqbegin(&rename_lock);
 	for(;;) {
 		struct inode *inode = d->d_inode;
-		if (inode && unlikely(!list_empty(&inode->inotify_watches))) {
+		if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_mark_entries))) {
 			struct audit_chunk *chunk;
 			chunk = audit_tree_lookup(inode);
 			if (chunk) {