[net-next-2.6.git] / fs / notify / dnotify / dnotify.c

/*
 * Directory notifications for Linux.
 *
 * 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
 * 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>
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/dnotify.h>
#include <linux/init.h>
#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 *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;
};

/*
 * 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)
{
	__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 = 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;
	__u32 test_mask = event->mask & ~FS_EVENT_ON_CHILD;

	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 & test_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,
				      int data_type)
{
	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;

	inode = filp->f_path.dentry->d_inode;
	if (!S_ISDIR(inode->i_mode))
		return;

	spin_lock(&inode->i_lock);
	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;
			kmem_cache_free(dnotify_struct_cache, dn);
			dnotify_recalc_inode_mask(entry);
			break;
		}
		prev = &dn->dn_next;
	}

	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 inode *inode;
	fl_owner_t id = current->files;
	struct file *f;
	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);
		error = 0;
		goto out_err;
	}

	/* dnotify only works on directories */
	inode = filp->f_path.dentry->d_inode;
	if (!S_ISDIR(inode->i_mode)) {
		error = -ENOTDIR;
		goto out_err;
	}

	/* 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;
	}

	/* 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;
	}

	/* convert the userspace DN_* "arg" to the internal FS_* defines in fsnotify */
	mask = convert_arg(arg);

	/* 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;

	/* 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);
	entry = fsnotify_find_mark_entry(dnotify_group, inode);
	spin_unlock(&inode->i_lock);
	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, 0);
		spin_lock(&new_entry->lock);
		entry = new_entry;
		dnentry = new_dnentry;
		/* we used new_entry, so don't free it */
		new_entry = NULL;
	}

	rcu_read_lock();
	f = fcheck(fd);
	rcu_read_unlock();

	/* 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;
	}

	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;
}

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);
	if (IS_ERR(dnotify_group))
		panic("unable to allocate fsnotify group for dnotify\n");
	return 0;
}

module_init(dnotify_init)
Commit	Line	Data
1da177e4 LT	1	/*
	2	* Directory notifications for Linux.
	3	*
	4	* Copyright (C) 2000,2001,2002 Stephen Rothwell
	5	*
3c5119c0 EP	6	* Copyright (C) 2009 Eric Paris <Red Hat Inc>
	7	* dnotify was largly rewritten to use the new fsnotify infrastructure
	8	*
1da177e4 LT	9	* This program is free software; you can redistribute it and/or modify it
	10	* under the terms of the GNU General Public License as published by the
	11	* Free Software Foundation; either version 2, or (at your option) any
	12	* later version.
	13	*
	14	* This program is distributed in the hope that it will be useful, but
	15	* WITHOUT ANY WARRANTY; without even the implied warranty of
	16	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	17	* General Public License for more details.
	18	*/
	19	#include <linux/fs.h>
	20	#include <linux/module.h>
	21	#include <linux/sched.h>
	22	#include <linux/dnotify.h>
	23	#include <linux/init.h>
	24	#include <linux/spinlock.h>
	25	#include <linux/slab.h>
9f3acc31	26	#include <linux/fdtable.h>
3c5119c0	27	#include <linux/fsnotify_backend.h>
1da177e4	28
fa3536cc	29	int dir_notify_enable __read_mostly = 1;
1da177e4	30
3c5119c0 EP	31	static struct kmem_cache *dnotify_struct_cache __read_mostly;
	32	static struct kmem_cache *dnotify_mark_entry_cache __read_mostly;
	33	static struct fsnotify_group *dnotify_group __read_mostly;
	34	static DEFINE_MUTEX(dnotify_mark_mutex);
	35
	36	/*
	37	* dnotify will attach one of these to each inode (i_fsnotify_mark_entries) which
	38	* is being watched by dnotify. If multiple userspace applications are watching
	39	* the same directory with dnotify their information is chained in dn
	40	*/
	41	struct dnotify_mark_entry {
	42	struct fsnotify_mark_entry fsn_entry;
	43	struct dnotify_struct *dn;
	44	};
1da177e4	45
3c5119c0 EP	46	/*
	47	* When a process starts or stops watching an inode the set of events which
	48	* dnotify cares about for that inode may change. This function runs the
	49	* list of everything receiving dnotify events about this directory and calculates
	50	* the set of all those events. After it updates what dnotify is interested in
	51	* it calls the fsnotify function so it can update the set of all events relevant
	52	* to this inode.
	53	*/
	54	static void dnotify_recalc_inode_mask(struct fsnotify_mark_entry *entry)
1da177e4	55	{
3c5119c0	56	__u32 new_mask, old_mask;
1da177e4	57	struct dnotify_struct *dn;
3c5119c0 EP	58	struct dnotify_mark_entry *dnentry = container_of(entry,
	59	struct dnotify_mark_entry,
	60	fsn_entry);
	61
	62	assert_spin_locked(&entry->lock);
1da177e4	63
3c5119c0	64	old_mask = entry->mask;
1da177e4	65	new_mask = 0;
3c5119c0 EP	66	for (dn = dnentry->dn; dn != NULL; dn = dn->dn_next)
	67	new_mask \|= (dn->dn_mask & ~FS_DN_MULTISHOT);
	68	entry->mask = new_mask;
	69
	70	if (old_mask == new_mask)
	71	return;
	72
	73	if (entry->inode)
	74	fsnotify_recalc_inode_mask(entry->inode);
1da177e4 LT	75	}
1da177e4 LT	76
3c5119c0 EP	77	/*
	78	* Mains fsnotify call where events are delivered to dnotify.
	79	* Find the dnotify mark on the relevant inode, run the list of dnotify structs
	80	* on that mark and determine which of them has expressed interest in receiving
	81	* events of this type. When found send the correct process and signal and
	82	* destroy the dnotify struct if it was not registered to receive multiple
	83	* events.
	84	*/
	85	static int dnotify_handle_event(struct fsnotify_group *group,
	86	struct fsnotify_event *event)
	87	{
	88	struct fsnotify_mark_entry *entry = NULL;
	89	struct dnotify_mark_entry *dnentry;
	90	struct inode *to_tell;
	91	struct dnotify_struct *dn;
	92	struct dnotify_struct **prev;
	93	struct fown_struct *fown;
94552684	94	__u32 test_mask = event->mask & ~FS_EVENT_ON_CHILD;
3c5119c0 EP	95
	96	to_tell = event->to_tell;
	97
	98	spin_lock(&to_tell->i_lock);
	99	entry = fsnotify_find_mark_entry(group, to_tell);
	100	spin_unlock(&to_tell->i_lock);
	101
	102	/* unlikely since we alreay passed dnotify_should_send_event() */
	103	if (unlikely(!entry))
	104	return 0;
	105	dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry);
	106
	107	spin_lock(&entry->lock);
	108	prev = &dnentry->dn;
	109	while ((dn = *prev) != NULL) {
94552684	110	if ((dn->dn_mask & test_mask) == 0) {
3c5119c0 EP	111	prev = &dn->dn_next;
	112	continue;
	113	}
	114	fown = &dn->dn_filp->f_owner;
	115	send_sigio(fown, dn->dn_fd, POLL_MSG);
	116	if (dn->dn_mask & FS_DN_MULTISHOT)
	117	prev = &dn->dn_next;
	118	else {
	119	*prev = dn->dn_next;
	120	kmem_cache_free(dnotify_struct_cache, dn);
	121	dnotify_recalc_inode_mask(entry);
	122	}
	123	}
	124
	125	spin_unlock(&entry->lock);
	126	fsnotify_put_mark(entry);
	127
	128	return 0;
	129	}
	130
	131	/*
	132	* Given an inode and mask determine if dnotify would be interested in sending
	133	* userspace notification for that pair.
	134	*/
	135	static bool dnotify_should_send_event(struct fsnotify_group *group,
7b0a04fb EP	136	struct inode *inode, __u32 mask,
7b0a04fb EP	137	int data_type)
3c5119c0 EP	138	{
	139	struct fsnotify_mark_entry *entry;
	140	bool send;
	141
	142	/* !dir_notify_enable should never get here, don't waste time checking
	143	if (!dir_notify_enable)
	144	return 0; */
	145
	146	/* not a dir, dnotify doesn't care */
	147	if (!S_ISDIR(inode->i_mode))
	148	return false;
	149
	150	spin_lock(&inode->i_lock);
	151	entry = fsnotify_find_mark_entry(group, inode);
	152	spin_unlock(&inode->i_lock);
	153
	154	/* no mark means no dnotify watch */
	155	if (!entry)
	156	return false;
	157
e42e2773	158	mask = (mask & ~FS_EVENT_ON_CHILD);
5ac697b7	159	send = (mask & entry->mask);
ce61856b	160
3c5119c0 EP	161	fsnotify_put_mark(entry); /* matches fsnotify_find_mark_entry */
	162
	163	return send;
	164	}
	165
3c5119c0 EP	166	static void dnotify_free_mark(struct fsnotify_mark_entry *entry)
	167	{
	168	struct dnotify_mark_entry *dnentry = container_of(entry,
	169	struct dnotify_mark_entry,
	170	fsn_entry);
	171
	172	BUG_ON(dnentry->dn);
	173
	174	kmem_cache_free(dnotify_mark_entry_cache, dnentry);
	175	}
	176
	177	static struct fsnotify_ops dnotify_fsnotify_ops = {
	178	.handle_event = dnotify_handle_event,
	179	.should_send_event = dnotify_should_send_event,
	180	.free_group_priv = NULL,
a092ee20	181	.freeing_mark = NULL,
e4aff117	182	.free_event_priv = NULL,
3c5119c0 EP	183	};
	184
	185	/*
	186	* Called every time a file is closed. Looks first for a dnotify mark on the
	187	* inode. If one is found run all of the ->dn entries attached to that
	188	* mark for one relevant to this process closing the file and remove that
	189	* dnotify_struct. If that was the last dnotify_struct also remove the
	190	* fsnotify_mark_entry.
	191	*/
1da177e4 LT	192	void dnotify_flush(struct file *filp, fl_owner_t id)
1da177e4 LT	193	{
3c5119c0 EP	194	struct fsnotify_mark_entry *entry;
3c5119c0 EP	195	struct dnotify_mark_entry *dnentry;
1da177e4 LT	196	struct dnotify_struct *dn;
	197	struct dnotify_struct **prev;
	198	struct inode *inode;
	199
0f7fc9e4	200	inode = filp->f_path.dentry->d_inode;
1da177e4 LT	201	if (!S_ISDIR(inode->i_mode))
1da177e4 LT	202	return;
3c5119c0	203
1da177e4	204	spin_lock(&inode->i_lock);
3c5119c0 EP	205	entry = fsnotify_find_mark_entry(dnotify_group, inode);
	206	spin_unlock(&inode->i_lock);
	207	if (!entry)
	208	return;
	209	dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry);
	210
	211	mutex_lock(&dnotify_mark_mutex);
	212
	213	spin_lock(&entry->lock);
	214	prev = &dnentry->dn;
1da177e4 LT	215	while ((dn = *prev) != NULL) {
	216	if ((dn->dn_owner == id) && (dn->dn_filp == filp)) {
	217	*prev = dn->dn_next;
3c5119c0 EP	218	kmem_cache_free(dnotify_struct_cache, dn);
3c5119c0 EP	219	dnotify_recalc_inode_mask(entry);
1da177e4 LT	220	break;
	221	}
	222	prev = &dn->dn_next;
	223	}
3c5119c0 EP	224
	225	spin_unlock(&entry->lock);
	226
	227	/* nothing else could have found us thanks to the dnotify_mark_mutex */
	228	if (dnentry->dn == NULL)
	229	fsnotify_destroy_mark_by_entry(entry);
	230
	231	fsnotify_recalc_group_mask(dnotify_group);
	232
	233	mutex_unlock(&dnotify_mark_mutex);
	234
	235	fsnotify_put_mark(entry);
	236	}
	237
	238	/* this conversion is done only at watch creation */
	239	static __u32 convert_arg(unsigned long arg)
	240	{
	241	__u32 new_mask = FS_EVENT_ON_CHILD;
	242
	243	if (arg & DN_MULTISHOT)
	244	new_mask \|= FS_DN_MULTISHOT;
	245	if (arg & DN_DELETE)
	246	new_mask \|= (FS_DELETE \| FS_MOVED_FROM);
	247	if (arg & DN_MODIFY)
	248	new_mask \|= FS_MODIFY;
	249	if (arg & DN_ACCESS)
	250	new_mask \|= FS_ACCESS;
	251	if (arg & DN_ATTRIB)
	252	new_mask \|= FS_ATTRIB;
	253	if (arg & DN_RENAME)
	254	new_mask \|= FS_DN_RENAME;
	255	if (arg & DN_CREATE)
	256	new_mask \|= (FS_CREATE \| FS_MOVED_TO);
	257
	258	return new_mask;
1da177e4 LT	259	}
1da177e4 LT	260
3c5119c0 EP	261	/*
	262	* If multiple processes watch the same inode with dnotify there is only one
	263	* dnotify mark in inode->i_fsnotify_mark_entries but we chain a dnotify_struct
	264	* onto that mark. This function either attaches the new dnotify_struct onto
	265	* that list, or it \|= the mask onto an existing dnofiy_struct.
	266	*/
	267	static int attach_dn(struct dnotify_struct dn, struct dnotify_mark_entry dnentry,
	268	fl_owner_t id, int fd, struct file *filp, __u32 mask)
	269	{
	270	struct dnotify_struct *odn;
	271
	272	odn = dnentry->dn;
	273	while (odn != NULL) {
	274	/* adding more events to existing dnofiy_struct? */
	275	if ((odn->dn_owner == id) && (odn->dn_filp == filp)) {
	276	odn->dn_fd = fd;
	277	odn->dn_mask \|= mask;
	278	return -EEXIST;
	279	}
	280	odn = odn->dn_next;
	281	}
	282
	283	dn->dn_mask = mask;
	284	dn->dn_fd = fd;
	285	dn->dn_filp = filp;
	286	dn->dn_owner = id;
	287	dn->dn_next = dnentry->dn;
	288	dnentry->dn = dn;
	289
	290	return 0;
	291	}
	292
	293	/*
	294	* When a process calls fcntl to attach a dnotify watch to a directory it ends
	295	* up here. Allocate both a mark for fsnotify to add and a dnotify_struct to be
	296	* attached to the fsnotify_mark.
	297	*/
1da177e4 LT	298	int fcntl_dirnotify(int fd, struct file *filp, unsigned long arg)
1da177e4 LT	299	{
3c5119c0 EP	300	struct dnotify_mark_entry new_dnentry, dnentry;
3c5119c0 EP	301	struct fsnotify_mark_entry new_entry, entry;
1da177e4	302	struct dnotify_struct *dn;
1da177e4 LT	303	struct inode *inode;
1da177e4 LT	304	fl_owner_t id = current->files;
214b7049	305	struct file *f;
3c5119c0 EP	306	int destroy = 0, error = 0;
	307	__u32 mask;
	308
	309	/* we use these to tell if we need to kfree */
	310	new_entry = NULL;
	311	dn = NULL;
1da177e4	312
3c5119c0 EP	313	if (!dir_notify_enable) {
	314	error = -EINVAL;
	315	goto out_err;
	316	}
	317
	318	/* a 0 mask means we are explicitly removing the watch */
1da177e4 LT	319	if ((arg & ~DN_MULTISHOT) == 0) {
1da177e4 LT	320	dnotify_flush(filp, id);
3c5119c0 EP	321	error = 0;
3c5119c0 EP	322	goto out_err;
1da177e4	323	}
3c5119c0 EP	324
3c5119c0 EP	325	/* dnotify only works on directories */
0f7fc9e4	326	inode = filp->f_path.dentry->d_inode;
3c5119c0 EP	327	if (!S_ISDIR(inode->i_mode)) {
	328	error = -ENOTDIR;
	329	goto out_err;
1da177e4 LT	330	}
1da177e4 LT	331
3c5119c0 EP	332	/* expect most fcntl to add new rather than augment old */
	333	dn = kmem_cache_alloc(dnotify_struct_cache, GFP_KERNEL);
	334	if (!dn) {
	335	error = -ENOMEM;
	336	goto out_err;
	337	}
214b7049	338
3c5119c0 EP	339	/* new fsnotify mark, we expect most fcntl calls to add a new mark */
	340	new_dnentry = kmem_cache_alloc(dnotify_mark_entry_cache, GFP_KERNEL);
	341	if (!new_dnentry) {
	342	error = -ENOMEM;
	343	goto out_err;
	344	}
1da177e4	345
3c5119c0 EP	346	/* convert the userspace DN_* "arg" to the internal FS_* defines in fsnotify */
3c5119c0 EP	347	mask = convert_arg(arg);
1da177e4	348
3c5119c0 EP	349	/* set up the new_entry and new_dnentry */
	350	new_entry = &new_dnentry->fsn_entry;
	351	fsnotify_init_mark(new_entry, dnotify_free_mark);
	352	new_entry->mask = mask;
	353	new_dnentry->dn = NULL;
1da177e4	354
3c5119c0 EP	355	/* this is needed to prevent the fcntl/close race described below */
3c5119c0 EP	356	mutex_lock(&dnotify_mark_mutex);
1da177e4	357
3c5119c0	358	/* add the new_entry or find an old one. */
1da177e4	359	spin_lock(&inode->i_lock);
3c5119c0	360	entry = fsnotify_find_mark_entry(dnotify_group, inode);
1da177e4	361	spin_unlock(&inode->i_lock);
3c5119c0 EP	362	if (entry) {
	363	dnentry = container_of(entry, struct dnotify_mark_entry, fsn_entry);
	364	spin_lock(&entry->lock);
	365	} else {
40554c3d	366	fsnotify_add_mark(new_entry, dnotify_group, inode, 0);
3c5119c0 EP	367	spin_lock(&new_entry->lock);
	368	entry = new_entry;
	369	dnentry = new_dnentry;
	370	/* we used new_entry, so don't free it */
	371	new_entry = NULL;
	372	}
1da177e4	373
3c5119c0 EP	374	rcu_read_lock();
	375	f = fcheck(fd);
	376	rcu_read_unlock();
1da177e4	377
3c5119c0 EP	378	/* if (f != filp) means that we lost a race and another task/thread
	379	* actually closed the fd we are still playing with before we grabbed
	380	* the dnotify_mark_mutex and entry->lock. Since closing the fd is the
	381	* only time we clean up the mark entries we need to get our mark off
	382	* the list. */
	383	if (f != filp) {
	384	/* if we added ourselves, shoot ourselves, it's possible that
	385	* the flush actually did shoot this entry. That's fine too
	386	* since multiple calls to destroy_mark is perfectly safe, if
	387	* we found a dnentry already attached to the inode, just sod
	388	* off silently as the flush at close time dealt with it.
	389	*/
	390	if (dnentry == new_dnentry)
	391	destroy = 1;
	392	goto out;
	393	}
1da177e4	394
3c5119c0 EP	395	error = __f_setown(filp, task_pid(current), PIDTYPE_PID, 0);
	396	if (error) {
	397	/* if we added, we must shoot */
	398	if (dnentry == new_dnentry)
	399	destroy = 1;
	400	goto out;
1da177e4	401	}
3c5119c0 EP	402
	403	error = attach_dn(dn, dnentry, id, fd, filp, mask);
	404	/* !error means that we attached the dn to the dnentry, so don't free it */
	405	if (!error)
	406	dn = NULL;
	407	/* -EEXIST means that we didn't add this new dn and used an old one.
	408	* that isn't an error (and the unused dn should be freed) */
	409	else if (error == -EEXIST)
	410	error = 0;
	411
	412	dnotify_recalc_inode_mask(entry);
	413	out:
	414	spin_unlock(&entry->lock);
	415
	416	if (destroy)
	417	fsnotify_destroy_mark_by_entry(entry);
	418
	419	fsnotify_recalc_group_mask(dnotify_group);
	420
	421	mutex_unlock(&dnotify_mark_mutex);
	422	fsnotify_put_mark(entry);
	423	out_err:
	424	if (new_entry)
	425	fsnotify_put_mark(new_entry);
	426	if (dn)
	427	kmem_cache_free(dnotify_struct_cache, dn);
	428	return error;
1da177e4	429	}
1da177e4 LT	430
	431	static int __init dnotify_init(void)
	432	{
3c5119c0 EP	433	dnotify_struct_cache = KMEM_CACHE(dnotify_struct, SLAB_PANIC);
	434	dnotify_mark_entry_cache = KMEM_CACHE(dnotify_mark_entry, SLAB_PANIC);
	435
	436	dnotify_group = fsnotify_obtain_group(DNOTIFY_GROUP_NUM,
	437	0, &dnotify_fsnotify_ops);
	438	if (IS_ERR(dnotify_group))
	439	panic("unable to allocate fsnotify group for dnotify\n");
1da177e4 LT	440	return 0;
	441	}
	442
	443	module_init(dnotify_init)