[deliverable/linux.git] / fs / notify / fanotify / fanotify_user.c

#include <linux/fcntl.h>
#include <linux/file.h>
#include <linux/fs.h>
#include <linux/anon_inodes.h>
#include <linux/fsnotify_backend.h>
#include <linux/init.h>
#include <linux/mount.h>
#include <linux/namei.h>
#include <linux/poll.h>
#include <linux/security.h>
#include <linux/syscalls.h>
#include <linux/types.h>
#include <linux/uaccess.h>

#include <asm/ioctls.h>

#include "fanotify.h"

static struct kmem_cache *fanotify_mark_cache __read_mostly;

/*
 * Get an fsnotify notification event if one exists and is small
 * enough to fit in "count". Return an error pointer if the count
 * is not large enough.
 *
 * Called with the group->notification_mutex held.
 */
static struct fsnotify_event *get_one_event(struct fsnotify_group *group,
					    size_t count)
{
	BUG_ON(!mutex_is_locked(&group->notification_mutex));

	pr_debug("%s: group=%p count=%zd\n", __func__, group, count);

	if (fsnotify_notify_queue_is_empty(group))
		return NULL;

	if (FAN_EVENT_METADATA_LEN > count)
		return ERR_PTR(-EINVAL);

	/* held the notification_mutex the whole time, so this is the
	 * same event we peeked above */
	return fsnotify_remove_notify_event(group);
}

static int create_and_fill_fd(struct fsnotify_group *group,
			      struct fanotify_event_metadata *metadata,
			      struct fsnotify_event *event)
{
	int client_fd;
	struct dentry *dentry;
	struct vfsmount *mnt;
	struct file *new_file;

	pr_debug("%s: group=%p metadata=%p event=%p\n", __func__, group,
		 metadata, event);

	client_fd = get_unused_fd();
	if (client_fd < 0)
		return client_fd;

	if (event->data_type != FSNOTIFY_EVENT_PATH) {
		WARN_ON(1);
		put_unused_fd(client_fd);
		return -EINVAL;
	}

	/*
	 * we need a new file handle for the userspace program so it can read even if it was
	 * originally opened O_WRONLY.
	 */
	dentry = dget(event->path.dentry);
	mnt = mntget(event->path.mnt);
	/* it's possible this event was an overflow event.  in that case dentry and mnt
	 * are NULL;  That's fine, just don't call dentry open */
	if (dentry && mnt)
		new_file = dentry_open(dentry, mnt,
				       O_RDONLY | O_LARGEFILE | FMODE_NONOTIFY,
				       current_cred());
	else
		new_file = ERR_PTR(-EOVERFLOW);
	if (IS_ERR(new_file)) {
		/*
		 * we still send an event even if we can't open the file.  this
		 * can happen when say tasks are gone and we try to open their
		 * /proc files or we try to open a WRONLY file like in sysfs
		 * we just send the errno to userspace since there isn't much
		 * else we can do.
		 */
		put_unused_fd(client_fd);
		client_fd = PTR_ERR(new_file);
	} else {
		fd_install(client_fd, new_file);
	}

	metadata->fd = client_fd;

	return 0;
}

static ssize_t fill_event_metadata(struct fsnotify_group *group,
				   struct fanotify_event_metadata *metadata,
				   struct fsnotify_event *event)
{
	pr_debug("%s: group=%p metadata=%p event=%p\n", __func__,
		 group, metadata, event);

	metadata->event_len = FAN_EVENT_METADATA_LEN;
	metadata->vers = FANOTIFY_METADATA_VERSION;
	metadata->mask = fanotify_outgoing_mask(event->mask);

	return create_and_fill_fd(group, metadata, event);

}

static ssize_t copy_event_to_user(struct fsnotify_group *group,
				  struct fsnotify_event *event,
				  char __user *buf)
{
	struct fanotify_event_metadata fanotify_event_metadata;
	int ret;

	pr_debug("%s: group=%p event=%p\n", __func__, group, event);

	ret = fill_event_metadata(group, &fanotify_event_metadata, event);
	if (ret)
		return ret;

	if (copy_to_user(buf, &fanotify_event_metadata, FAN_EVENT_METADATA_LEN))
		return -EFAULT;

	return FAN_EVENT_METADATA_LEN;
}

/* intofiy userspace file descriptor functions */
static unsigned int fanotify_poll(struct file *file, poll_table *wait)
{
	struct fsnotify_group *group = file->private_data;
	int ret = 0;

	poll_wait(file, &group->notification_waitq, wait);
	mutex_lock(&group->notification_mutex);
	if (!fsnotify_notify_queue_is_empty(group))
		ret = POLLIN | POLLRDNORM;
	mutex_unlock(&group->notification_mutex);

	return ret;
}

static ssize_t fanotify_read(struct file *file, char __user *buf,
			     size_t count, loff_t *pos)
{
	struct fsnotify_group *group;
	struct fsnotify_event *kevent;
	char __user *start;
	int ret;
	DEFINE_WAIT(wait);

	start = buf;
	group = file->private_data;

	pr_debug("%s: group=%p\n", __func__, group);

	while (1) {
		prepare_to_wait(&group->notification_waitq, &wait, TASK_INTERRUPTIBLE);

		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(group, kevent, buf);
			fsnotify_put_event(kevent);
			if (ret < 0)
				break;
			buf += ret;
			count -= ret;
			continue;
		}

		ret = -EAGAIN;
		if (file->f_flags & O_NONBLOCK)
			break;
		ret = -EINTR;
		if (signal_pending(current))
			break;

		if (start != buf)
			break;

		schedule();
	}

	finish_wait(&group->notification_waitq, &wait);
	if (start != buf && ret != -EFAULT)
		ret = buf - start;
	return ret;
}

static int fanotify_release(struct inode *ignored, struct file *file)
{
	struct fsnotify_group *group = file->private_data;

	pr_debug("%s: file=%p group=%p\n", __func__, file, group);

	/* matches the fanotify_init->fsnotify_alloc_group */
	fsnotify_put_group(group);

	return 0;
}

static long fanotify_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
{
	struct fsnotify_group *group;
	struct fsnotify_event_holder *holder;
	void __user *p;
	int ret = -ENOTTY;
	size_t send_len = 0;

	group = file->private_data;

	p = (void __user *) arg;

	switch (cmd) {
	case FIONREAD:
		mutex_lock(&group->notification_mutex);
		list_for_each_entry(holder, &group->notification_list, event_list)
			send_len += FAN_EVENT_METADATA_LEN;
		mutex_unlock(&group->notification_mutex);
		ret = put_user(send_len, (int __user *) p);
		break;
	}

	return ret;
}

static const struct file_operations fanotify_fops = {
	.poll		= fanotify_poll,
	.read		= fanotify_read,
	.fasync		= NULL,
	.release	= fanotify_release,
	.unlocked_ioctl	= fanotify_ioctl,
	.compat_ioctl	= fanotify_ioctl,
};

static void fanotify_free_mark(struct fsnotify_mark *fsn_mark)
{
	kmem_cache_free(fanotify_mark_cache, fsn_mark);
}

static int fanotify_find_path(int dfd, const char __user *filename,
			      struct path *path, unsigned int flags)
{
	int ret;

	pr_debug("%s: dfd=%d filename=%p flags=%x\n", __func__,
		 dfd, filename, flags);

	if (filename == NULL) {
		struct file *file;
		int fput_needed;

		ret = -EBADF;
		file = fget_light(dfd, &fput_needed);
		if (!file)
			goto out;

		ret = -ENOTDIR;
		if ((flags & FAN_MARK_ONLYDIR) &&
		    !(S_ISDIR(file->f_path.dentry->d_inode->i_mode))) {
			fput_light(file, fput_needed);
			goto out;
		}

		*path = file->f_path;
		path_get(path);
		fput_light(file, fput_needed);
	} else {
		unsigned int lookup_flags = 0;

		if (!(flags & FAN_MARK_DONT_FOLLOW))
			lookup_flags |= LOOKUP_FOLLOW;
		if (flags & FAN_MARK_ONLYDIR)
			lookup_flags |= LOOKUP_DIRECTORY;

		ret = user_path_at(dfd, filename, lookup_flags, path);
		if (ret)
			goto out;
	}

	/* you can only watch an inode if you have read permissions on it */
	ret = inode_permission(path->dentry->d_inode, MAY_READ);
	if (ret)
		path_put(path);
out:
	return ret;
}

static int fanotify_remove_mark(struct fsnotify_group *group,
				struct inode *inode,
				__u32 mask)
{
	struct fsnotify_mark *fsn_mark;
	__u32 new_mask;

	pr_debug("%s: group=%p inode=%p mask=%x\n", __func__,
		 group, inode, mask);

	fsn_mark = fsnotify_find_mark(group, inode);
	if (!fsn_mark)
		return -ENOENT;

	spin_lock(&fsn_mark->lock);
	fsn_mark->mask &= ~mask;
	new_mask = fsn_mark->mask;
	spin_unlock(&fsn_mark->lock);

	if (!new_mask)
		fsnotify_destroy_mark(fsn_mark);
	else
		fsnotify_recalc_inode_mask(inode);

	fsnotify_recalc_group_mask(group);

	/* matches the fsnotify_find_mark() */
	fsnotify_put_mark(fsn_mark);

	return 0;
}

static int fanotify_add_mark(struct fsnotify_group *group,
			     struct inode *inode,
			     __u32 mask)
{
	struct fsnotify_mark *fsn_mark;
	__u32 old_mask, new_mask;
	int ret;

	pr_debug("%s: group=%p inode=%p mask=%x\n", __func__,
		 group, inode, mask);

	fsn_mark = fsnotify_find_mark(group, inode);
	if (!fsn_mark) {
		struct fsnotify_mark *new_fsn_mark;

		ret = -ENOMEM;
		new_fsn_mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
		if (!new_fsn_mark)
			goto out;

		fsnotify_init_mark(new_fsn_mark, fanotify_free_mark);
		ret = fsnotify_add_mark(new_fsn_mark, group, inode, 0);
		if (ret) {
			fanotify_free_mark(new_fsn_mark);
			goto out;
		}

		fsn_mark = new_fsn_mark;
	}

	ret = 0;

	spin_lock(&fsn_mark->lock);
	old_mask = fsn_mark->mask;
	fsn_mark->mask |= mask;
	new_mask = fsn_mark->mask;
	spin_unlock(&fsn_mark->lock);

	/* we made changes to a mask, update the group mask and the inode mask
	 * so things happen quickly. */
	if (old_mask != new_mask) {
		/* more bits in old than in new? */
		int dropped = (old_mask & ~new_mask);
		/* more bits in this mark than the inode's mask? */
		int do_inode = (new_mask & ~inode->i_fsnotify_mask);
		/* more bits in this mark than the group? */
		int do_group = (new_mask & ~group->mask);

		/* update the inode with this new mark */
		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);
	}

	/* match the init or the find.... */
	fsnotify_put_mark(fsn_mark);
out:
	return ret;
}

static int fanotify_update_mark(struct fsnotify_group *group,
				struct inode *inode, int flags,
				__u32 mask)
{
	pr_debug("%s: group=%p inode=%p flags=%x mask=%x\n", __func__,
		 group, inode, flags, mask);

	if (flags & FAN_MARK_ADD)
		fanotify_add_mark(group, inode, mask);
	else if (flags & FAN_MARK_REMOVE)
		fanotify_remove_mark(group, inode, mask);
	else
		BUG();

	return 0;
}

static bool fanotify_mark_validate_input(int flags,
					 __u32 mask)
{
	pr_debug("%s: flags=%x mask=%x\n", __func__, flags, mask);

	/* are flags valid of this operation? */
	if (!fanotify_mark_flags_valid(flags))
		return false;
	/* is the mask valid? */
	if (!fanotify_mask_valid(mask))
		return false;
	return true;
}

/* fanotify syscalls */
SYSCALL_DEFINE3(fanotify_init, unsigned int, flags, unsigned int, event_f_flags,
		unsigned int, priority)
{
	struct fsnotify_group *group;
	int f_flags, fd;

	pr_debug("%s: flags=%d event_f_flags=%d priority=%d\n",
		__func__, flags, event_f_flags, priority);

	if (event_f_flags)
		return -EINVAL;
	if (priority)
		return -EINVAL;

	if (!capable(CAP_SYS_ADMIN))
		return -EACCES;

	if (flags & ~FAN_ALL_INIT_FLAGS)
		return -EINVAL;

	f_flags = (O_RDONLY | FMODE_NONOTIFY);
	if (flags & FAN_CLOEXEC)
		f_flags |= O_CLOEXEC;
	if (flags & FAN_NONBLOCK)
		f_flags |= O_NONBLOCK;

	/* fsnotify_alloc_group takes a ref.  Dropped in fanotify_release */
	group = fsnotify_alloc_group(&fanotify_fsnotify_ops);
	if (IS_ERR(group))
		return PTR_ERR(group);

	fd = anon_inode_getfd("[fanotify]", &fanotify_fops, group, f_flags);
	if (fd < 0)
		goto out_put_group;

	return fd;

out_put_group:
	fsnotify_put_group(group);
	return fd;
}

SYSCALL_DEFINE5(fanotify_mark, int, fanotify_fd, unsigned int, flags,
		__u64, mask, int, dfd, const char  __user *, pathname)
{
	struct inode *inode;
	struct fsnotify_group *group;
	struct file *filp;
	struct path path;
	int ret, fput_needed;

	pr_debug("%s: fanotify_fd=%d flags=%x dfd=%d pathname=%p mask=%llx\n",
		 __func__, fanotify_fd, flags, dfd, pathname, mask);

	/* we only use the lower 32 bits as of right now. */
	if (mask & ((__u64)0xffffffff << 32))
		return -EINVAL;

	if (!fanotify_mark_validate_input(flags, mask))
		return -EINVAL;

	filp = fget_light(fanotify_fd, &fput_needed);
	if (unlikely(!filp))
		return -EBADF;

	/* verify that this is indeed an fanotify instance */
	ret = -EINVAL;
	if (unlikely(filp->f_op != &fanotify_fops))
		goto fput_and_out;

	ret = fanotify_find_path(dfd, pathname, &path, flags);
	if (ret)
		goto fput_and_out;

	/* inode held in place by reference to path; group by fget on fd */
	inode = path.dentry->d_inode;
	group = filp->private_data;

	/* create/update an inode mark */
	ret = fanotify_update_mark(group, inode, flags, mask);

	path_put(&path);
fput_and_out:
	fput_light(filp, fput_needed);
	return ret;
}

/*
 * fanotify_user_setup - Our initialization function.  Note that we cannnot return
 * error because we have compiled-in VFS hooks.  So an (unlikely) failure here
 * must result in panic().
 */
static int __init fanotify_user_setup(void)
{
	fanotify_mark_cache = KMEM_CACHE(fsnotify_mark, SLAB_PANIC);

	return 0;
}
device_initcall(fanotify_user_setup);
Commit	Line	Data
11637e4b	1	#include <linux/fcntl.h>
2a3edf86	2	#include <linux/file.h>
11637e4b	3	#include <linux/fs.h>
52c923dd	4	#include <linux/anon_inodes.h>
11637e4b	5	#include <linux/fsnotify_backend.h>
2a3edf86	6	#include <linux/init.h>
a1014f10	7	#include <linux/mount.h>
2a3edf86	8	#include <linux/namei.h>
a1014f10	9	#include <linux/poll.h>
11637e4b EP	10	#include <linux/security.h>
11637e4b EP	11	#include <linux/syscalls.h>
2a3edf86	12	#include <linux/types.h>
a1014f10 EP	13	#include <linux/uaccess.h>
	14
	15	#include <asm/ioctls.h>
11637e4b EP	16
	17	#include "fanotify.h"
	18
2a3edf86 EP	19	static struct kmem_cache *fanotify_mark_cache __read_mostly;
2a3edf86 EP	20
a1014f10 EP	21	/*
	22	* Get an fsnotify notification event if one exists and is small
	23	* enough to fit in "count". Return an error pointer if the count
	24	* is not large enough.
	25	*
	26	* Called with the group->notification_mutex held.
	27	*/
	28	static struct fsnotify_event get_one_event(struct fsnotify_group group,
	29	size_t count)
	30	{
	31	BUG_ON(!mutex_is_locked(&group->notification_mutex));
	32
	33	pr_debug("%s: group=%p count=%zd\n", __func__, group, count);
	34
	35	if (fsnotify_notify_queue_is_empty(group))
	36	return NULL;
	37
	38	if (FAN_EVENT_METADATA_LEN > count)
	39	return ERR_PTR(-EINVAL);
	40
	41	/* held the notification_mutex the whole time, so this is the
	42	* same event we peeked above */
	43	return fsnotify_remove_notify_event(group);
	44	}
	45
	46	static int create_and_fill_fd(struct fsnotify_group *group,
	47	struct fanotify_event_metadata *metadata,
	48	struct fsnotify_event *event)
	49	{
	50	int client_fd;
	51	struct dentry *dentry;
	52	struct vfsmount *mnt;
	53	struct file *new_file;
	54
	55	pr_debug("%s: group=%p metadata=%p event=%p\n", __func__, group,
	56	metadata, event);
	57
	58	client_fd = get_unused_fd();
	59	if (client_fd < 0)
	60	return client_fd;
	61
	62	if (event->data_type != FSNOTIFY_EVENT_PATH) {
	63	WARN_ON(1);
	64	put_unused_fd(client_fd);
	65	return -EINVAL;
	66	}
	67
	68	/*
	69	* we need a new file handle for the userspace program so it can read even if it was
	70	* originally opened O_WRONLY.
	71	*/
	72	dentry = dget(event->path.dentry);
	73	mnt = mntget(event->path.mnt);
	74	/* it's possible this event was an overflow event. in that case dentry and mnt
	75	* are NULL; That's fine, just don't call dentry open */
	76	if (dentry && mnt)
	77	new_file = dentry_open(dentry, mnt,
	78	O_RDONLY \| O_LARGEFILE \| FMODE_NONOTIFY,
	79	current_cred());
	80	else
	81	new_file = ERR_PTR(-EOVERFLOW);
	82	if (IS_ERR(new_file)) {
	83	/*
	84	* we still send an event even if we can't open the file. this
85	* can happen when say tasks are gone and we try to open their
86	* /proc files or we try to open a WRONLY file like in sysfs
87	* we just send the errno to userspace since there isn't much
88	* else we can do.
89	*/
90	put_unused_fd(client_fd);
91	client_fd = PTR_ERR(new_file);
92	} else {
93	fd_install(client_fd, new_file);
94	}
95
96	metadata->fd = client_fd;
97
98	return 0;
99	}
100
101	static ssize_t fill_event_metadata(struct fsnotify_group *group,
102	struct fanotify_event_metadata *metadata,
103	struct fsnotify_event *event)
104	{
105	pr_debug("%s: group=%p metadata=%p event=%p\n", __func__,
106	group, metadata, event);
107
108	metadata->event_len = FAN_EVENT_METADATA_LEN;
109	metadata->vers = FANOTIFY_METADATA_VERSION;
110	metadata->mask = fanotify_outgoing_mask(event->mask);
111
112	return create_and_fill_fd(group, metadata, event);
113
114	}
115
116	static ssize_t copy_event_to_user(struct fsnotify_group *group,
117	struct fsnotify_event *event,
118	char __user *buf)
119	{
120	struct fanotify_event_metadata fanotify_event_metadata;
121	int ret;
122
123	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
124
125	ret = fill_event_metadata(group, &fanotify_event_metadata, event);
126	if (ret)
127	return ret;
128
129	if (copy_to_user(buf, &fanotify_event_metadata, FAN_EVENT_METADATA_LEN))
130	return -EFAULT;
131
132	return FAN_EVENT_METADATA_LEN;
133	}
134
135	/* intofiy userspace file descriptor functions */
136	static unsigned int fanotify_poll(struct file file, poll_table wait)
137	{
138	struct fsnotify_group *group = file->private_data;
139	int ret = 0;
140
141	poll_wait(file, &group->notification_waitq, wait);
142	mutex_lock(&group->notification_mutex);
143	if (!fsnotify_notify_queue_is_empty(group))
144	ret = POLLIN \| POLLRDNORM;
145	mutex_unlock(&group->notification_mutex);
146
147	return ret;
148	}
149
150	static ssize_t fanotify_read(struct file file, char __user buf,
151	size_t count, loff_t *pos)
152	{
153	struct fsnotify_group *group;
154	struct fsnotify_event *kevent;
155	char __user *start;
156	int ret;
157	DEFINE_WAIT(wait);
158
159	start = buf;
160	group = file->private_data;
161
162	pr_debug("%s: group=%p\n", __func__, group);
163
164	while (1) {
165	prepare_to_wait(&group->notification_waitq, &wait, TASK_INTERRUPTIBLE);
166
167	mutex_lock(&group->notification_mutex);
168	kevent = get_one_event(group, count);
169	mutex_unlock(&group->notification_mutex);
170
171	if (kevent) {
172	ret = PTR_ERR(kevent);
173	if (IS_ERR(kevent))
174	break;
175	ret = copy_event_to_user(group, kevent, buf);
176	fsnotify_put_event(kevent);
177	if (ret < 0)
178	break;
179	buf += ret;
180	count -= ret;
181	continue;
182	}
183
184	ret = -EAGAIN;
185	if (file->f_flags & O_NONBLOCK)
186	break;
187	ret = -EINTR;
188	if (signal_pending(current))
189	break;
190
191	if (start != buf)
192	break;
193
194	schedule();
195	}
196
197	finish_wait(&group->notification_waitq, &wait);
198	if (start != buf && ret != -EFAULT)
199	ret = buf - start;
200	return ret;
201	}
202
52c923dd EP	203	static int fanotify_release(struct inode ignored, struct file file)
	204	{
	205	struct fsnotify_group *group = file->private_data;
	206
	207	pr_debug("%s: file=%p group=%p\n", __func__, file, group);
	208
	209	/* matches the fanotify_init->fsnotify_alloc_group */
	210	fsnotify_put_group(group);
	211
	212	return 0;
	213	}
	214
a1014f10 EP	215	static long fanotify_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
	216	{
	217	struct fsnotify_group *group;
	218	struct fsnotify_event_holder *holder;
	219	void __user *p;
	220	int ret = -ENOTTY;
	221	size_t send_len = 0;
	222
	223	group = file->private_data;
	224
	225	p = (void __user *) arg;
	226
	227	switch (cmd) {
	228	case FIONREAD:
	229	mutex_lock(&group->notification_mutex);
	230	list_for_each_entry(holder, &group->notification_list, event_list)
	231	send_len += FAN_EVENT_METADATA_LEN;
	232	mutex_unlock(&group->notification_mutex);
	233	ret = put_user(send_len, (int __user *) p);
	234	break;
	235	}
	236
	237	return ret;
	238	}
	239
52c923dd	240	static const struct file_operations fanotify_fops = {
a1014f10 EP	241	.poll = fanotify_poll,
a1014f10 EP	242	.read = fanotify_read,
52c923dd EP	243	.fasync = NULL,
52c923dd EP	244	.release = fanotify_release,
a1014f10 EP	245	.unlocked_ioctl = fanotify_ioctl,
a1014f10 EP	246	.compat_ioctl = fanotify_ioctl,
52c923dd EP	247	};
52c923dd EP	248
2a3edf86 EP	249	static void fanotify_free_mark(struct fsnotify_mark *fsn_mark)
	250	{
	251	kmem_cache_free(fanotify_mark_cache, fsn_mark);
	252	}
	253
	254	static int fanotify_find_path(int dfd, const char __user *filename,
	255	struct path *path, unsigned int flags)
	256	{
	257	int ret;
	258
	259	pr_debug("%s: dfd=%d filename=%p flags=%x\n", __func__,
	260	dfd, filename, flags);
	261
	262	if (filename == NULL) {
	263	struct file *file;
	264	int fput_needed;
	265
	266	ret = -EBADF;
	267	file = fget_light(dfd, &fput_needed);
	268	if (!file)
	269	goto out;
	270
	271	ret = -ENOTDIR;
	272	if ((flags & FAN_MARK_ONLYDIR) &&
	273	!(S_ISDIR(file->f_path.dentry->d_inode->i_mode))) {
	274	fput_light(file, fput_needed);
	275	goto out;
	276	}
	277
	278	*path = file->f_path;
	279	path_get(path);
	280	fput_light(file, fput_needed);
	281	} else {
	282	unsigned int lookup_flags = 0;
	283
	284	if (!(flags & FAN_MARK_DONT_FOLLOW))
	285	lookup_flags \|= LOOKUP_FOLLOW;
	286	if (flags & FAN_MARK_ONLYDIR)
	287	lookup_flags \|= LOOKUP_DIRECTORY;
	288
	289	ret = user_path_at(dfd, filename, lookup_flags, path);
	290	if (ret)
	291	goto out;
	292	}
	293
	294	/* you can only watch an inode if you have read permissions on it */
	295	ret = inode_permission(path->dentry->d_inode, MAY_READ);
	296	if (ret)
	297	path_put(path);
	298	out:
	299	return ret;
	300	}
	301
	302	static int fanotify_remove_mark(struct fsnotify_group *group,
	303	struct inode *inode,
	304	__u32 mask)
	305	{
	306	struct fsnotify_mark *fsn_mark;
	307	__u32 new_mask;
	308
	309	pr_debug("%s: group=%p inode=%p mask=%x\n", __func__,
	310	group, inode, mask);
	311
	312	fsn_mark = fsnotify_find_mark(group, inode);
313	if (!fsn_mark)
314	return -ENOENT;
315
316	spin_lock(&fsn_mark->lock);
317	fsn_mark->mask &= ~mask;
318	new_mask = fsn_mark->mask;
319	spin_unlock(&fsn_mark->lock);
320
321	if (!new_mask)
322	fsnotify_destroy_mark(fsn_mark);
323	else
324	fsnotify_recalc_inode_mask(inode);
325
326	fsnotify_recalc_group_mask(group);
327
328	/* matches the fsnotify_find_mark() */
329	fsnotify_put_mark(fsn_mark);
330
331	return 0;
332	}
333
334	static int fanotify_add_mark(struct fsnotify_group *group,
335	struct inode *inode,
336	__u32 mask)
337	{
338	struct fsnotify_mark *fsn_mark;
339	__u32 old_mask, new_mask;
340	int ret;
341
342	pr_debug("%s: group=%p inode=%p mask=%x\n", __func__,
343	group, inode, mask);
344
345	fsn_mark = fsnotify_find_mark(group, inode);
346	if (!fsn_mark) {
347	struct fsnotify_mark *new_fsn_mark;
348
349	ret = -ENOMEM;
350	new_fsn_mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
351	if (!new_fsn_mark)
352	goto out;
353
354	fsnotify_init_mark(new_fsn_mark, fanotify_free_mark);
355	ret = fsnotify_add_mark(new_fsn_mark, group, inode, 0);
356	if (ret) {
357	fanotify_free_mark(new_fsn_mark);
358	goto out;
359	}
360
361	fsn_mark = new_fsn_mark;
362	}
363
364	ret = 0;
365
366	spin_lock(&fsn_mark->lock);
367	old_mask = fsn_mark->mask;
368	fsn_mark->mask \|= mask;
369	new_mask = fsn_mark->mask;
370	spin_unlock(&fsn_mark->lock);
371
372	/* we made changes to a mask, update the group mask and the inode mask
373	* so things happen quickly. */
374	if (old_mask != new_mask) {
375	/* more bits in old than in new? */
376	int dropped = (old_mask & ~new_mask);
377	/* more bits in this mark than the inode's mask? */
378	int do_inode = (new_mask & ~inode->i_fsnotify_mask);
379	/* more bits in this mark than the group? */
380	int do_group = (new_mask & ~group->mask);
381
382	/* update the inode with this new mark */
383	if (dropped \|\| do_inode)
384	fsnotify_recalc_inode_mask(inode);
385
386	/* update the group mask with the new mask */
387	if (dropped \|\| do_group)
388	fsnotify_recalc_group_mask(group);
389	}
390
391	/* match the init or the find.... */
392	fsnotify_put_mark(fsn_mark);
393	out:
394	return ret;
395	}
396
397	static int fanotify_update_mark(struct fsnotify_group *group,
398	struct inode *inode, int flags,
399	__u32 mask)
400	{
401	pr_debug("%s: group=%p inode=%p flags=%x mask=%x\n", __func__,
402	group, inode, flags, mask);
403
404	if (flags & FAN_MARK_ADD)
405	fanotify_add_mark(group, inode, mask);
406	else if (flags & FAN_MARK_REMOVE)
407	fanotify_remove_mark(group, inode, mask);
408	else
409	BUG();
410
411	return 0;
412	}
413
414	static bool fanotify_mark_validate_input(int flags,
415	__u32 mask)
416	{
417	pr_debug("%s: flags=%x mask=%x\n", __func__, flags, mask);
418
419	/* are flags valid of this operation? */
420	if (!fanotify_mark_flags_valid(flags))
421	return false;
422	/* is the mask valid? */
423	if (!fanotify_mask_valid(mask))
424	return false;
425	return true;
426	}
427
52c923dd	428	/* fanotify syscalls */
11637e4b EP	429	SYSCALL_DEFINE3(fanotify_init, unsigned int, flags, unsigned int, event_f_flags,
	430	unsigned int, priority)
	431	{
52c923dd EP	432	struct fsnotify_group *group;
	433	int f_flags, fd;
	434
	435	pr_debug("%s: flags=%d event_f_flags=%d priority=%d\n",
	436	__func__, flags, event_f_flags, priority);
	437
	438	if (event_f_flags)
	439	return -EINVAL;
	440	if (priority)
	441	return -EINVAL;
	442
	443	if (!capable(CAP_SYS_ADMIN))
	444	return -EACCES;
	445
	446	if (flags & ~FAN_ALL_INIT_FLAGS)
	447	return -EINVAL;
	448
	449	f_flags = (O_RDONLY \| FMODE_NONOTIFY);
	450	if (flags & FAN_CLOEXEC)
	451	f_flags \|= O_CLOEXEC;
	452	if (flags & FAN_NONBLOCK)
	453	f_flags \|= O_NONBLOCK;
	454
	455	/* fsnotify_alloc_group takes a ref. Dropped in fanotify_release */
	456	group = fsnotify_alloc_group(&fanotify_fsnotify_ops);
	457	if (IS_ERR(group))
	458	return PTR_ERR(group);
	459
	460	fd = anon_inode_getfd("[fanotify]", &fanotify_fops, group, f_flags);
	461	if (fd < 0)
	462	goto out_put_group;
	463
	464	return fd;
	465
	466	out_put_group:
	467	fsnotify_put_group(group);
	468	return fd;
11637e4b	469	}
bbaa4168 EP	470
	471	SYSCALL_DEFINE5(fanotify_mark, int, fanotify_fd, unsigned int, flags,
	472	__u64, mask, int, dfd, const char __user *, pathname)
	473	{
2a3edf86 EP	474	struct inode *inode;
	475	struct fsnotify_group *group;
	476	struct file *filp;
	477	struct path path;
	478	int ret, fput_needed;
	479
	480	pr_debug("%s: fanotify_fd=%d flags=%x dfd=%d pathname=%p mask=%llx\n",
	481	__func__, fanotify_fd, flags, dfd, pathname, mask);
	482
	483	/* we only use the lower 32 bits as of right now. */
	484	if (mask & ((__u64)0xffffffff << 32))
	485	return -EINVAL;
	486
	487	if (!fanotify_mark_validate_input(flags, mask))
	488	return -EINVAL;
	489
	490	filp = fget_light(fanotify_fd, &fput_needed);
	491	if (unlikely(!filp))
	492	return -EBADF;
	493
	494	/* verify that this is indeed an fanotify instance */
	495	ret = -EINVAL;
	496	if (unlikely(filp->f_op != &fanotify_fops))
	497	goto fput_and_out;
	498
	499	ret = fanotify_find_path(dfd, pathname, &path, flags);
	500	if (ret)
	501	goto fput_and_out;
	502
	503	/* inode held in place by reference to path; group by fget on fd */
	504	inode = path.dentry->d_inode;
	505	group = filp->private_data;
	506
	507	/* create/update an inode mark */
	508	ret = fanotify_update_mark(group, inode, flags, mask);
	509
	510	path_put(&path);
	511	fput_and_out:
	512	fput_light(filp, fput_needed);
	513	return ret;
	514	}
	515
	516	/*
	517	* fanotify_user_setup - Our initialization function. Note that we cannnot return
	518	* error because we have compiled-in VFS hooks. So an (unlikely) failure here
	519	* must result in panic().
	520	*/
	521	static int __init fanotify_user_setup(void)
	522	{
	523	fanotify_mark_cache = KMEM_CACHE(fsnotify_mark, SLAB_PANIC);
	524
	525	return 0;
bbaa4168	526	}
2a3edf86	527	device_initcall(fanotify_user_setup);