[net-next-2.6.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_fd(struct fsnotify_group *group, struct fsnotify_event *event)
{
	int client_fd;
	struct dentry *dentry;
	struct vfsmount *mnt;
	struct file *new_file;

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

	return client_fd;
}

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);
	metadata->pid = pid_vnr(event->tgid);
	metadata->fd = create_fd(group, event);

	return metadata->fd;
}

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 < 0)
		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 __u32 fanotify_mark_remove_from_mask(struct fsnotify_mark *fsn_mark, __u32 mask)
{
	__u32 oldmask;

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

	if (!(oldmask & ~mask))
		fsnotify_destroy_mark(fsn_mark);

	return mask & oldmask;
}

static int fanotify_remove_mark(struct fsnotify_group *group, struct inode *inode,
				struct vfsmount *mnt, __u32 mask)
{
	struct fsnotify_mark *fsn_mark = NULL;
	__u32 removed;

	BUG_ON(inode && mnt);
	BUG_ON(!inode && !mnt);

	if (inode)
		fsn_mark = fsnotify_find_inode_mark(group, inode);
	else if (mnt)
		fsn_mark = fsnotify_find_vfsmount_mark(group, mnt);
	else
		BUG();

	if (!fsn_mark)
		return -ENOENT;

	removed = fanotify_mark_remove_from_mask(fsn_mark, mask);
	/* matches the fsnotify_find_inode_mark() */
	fsnotify_put_mark(fsn_mark);

	if (removed & group->mask)
		fsnotify_recalc_group_mask(group);
	if (inode) {
		if (removed & inode->i_fsnotify_mask)
			fsnotify_recalc_inode_mask(inode);
	} else if (mnt) {
		if (removed & mnt->mnt_fsnotify_mask)
			fsnotify_recalc_vfsmount_mask(mnt);
	}

	return 0;
}

static __u32 fanotify_mark_add_to_mask(struct fsnotify_mark *fsn_mark, __u32 mask)
{
	__u32 oldmask;

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

	return mask & ~oldmask;
}

static int fanotify_add_vfsmount_mark(struct fsnotify_group *group,
				      struct vfsmount *mnt, __u32 mask)
{
	struct fsnotify_mark *fsn_mark;
	__u32 added;

	fsn_mark = fsnotify_find_vfsmount_mark(group, mnt);
	if (!fsn_mark) {
		int ret;

		fsn_mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
		if (!fsn_mark)
			return -ENOMEM;

		fsnotify_init_mark(fsn_mark, fanotify_free_mark);
		ret = fsnotify_add_mark(fsn_mark, group, NULL, mnt, 0);
		if (ret) {
			fanotify_free_mark(fsn_mark);
			return ret;
		}
	}
	added = fanotify_mark_add_to_mask(fsn_mark, mask);
	fsnotify_put_mark(fsn_mark);
	if (added) {
		if (added & ~group->mask)
			fsnotify_recalc_group_mask(group);
		if (added & ~mnt->mnt_fsnotify_mask)
			fsnotify_recalc_vfsmount_mask(mnt);
	}
	return 0;
}

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

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

	fsn_mark = fsnotify_find_inode_mark(group, inode);
	if (!fsn_mark) {
		int ret;

		fsn_mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
		if (!fsn_mark)
			return -ENOMEM;

		fsnotify_init_mark(fsn_mark, fanotify_free_mark);
		ret = fsnotify_add_mark(fsn_mark, group, inode, NULL, 0);
		if (ret) {
			fanotify_free_mark(fsn_mark);
			return ret;
		}
	}
	added = fanotify_mark_add_to_mask(fsn_mark, mask);
	fsnotify_put_mark(fsn_mark);
	if (added) {
		if (added & ~group->mask)
			fsnotify_recalc_group_mask(group);
		if (added & ~inode->i_fsnotify_mask)
			fsnotify_recalc_inode_mask(inode);
	}
	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_DEFINE(fanotify_mark)(int fanotify_fd, unsigned int flags,
			      __u64 mask, int dfd,
			      const char  __user * pathname)
{
	struct inode *inode = NULL;
	struct vfsmount *mnt = NULL;
	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 */
	if (!(flags & FAN_MARK_MOUNT))
		inode = path.dentry->d_inode;
	else
		mnt = path.mnt;
	group = filp->private_data;

	/* create/update an inode mark */
	switch (flags & (FAN_MARK_ADD | FAN_MARK_REMOVE)) {
	case FAN_MARK_ADD:
		if (flags & FAN_MARK_MOUNT)
			ret = fanotify_add_vfsmount_mark(group, mnt, mask);
		else
			ret = fanotify_add_inode_mark(group, inode, mask);
		break;
	case FAN_MARK_REMOVE:
		ret = fanotify_remove_mark(group, inode, mnt, mask);
		break;
	default:
		ret = -EINVAL;
	}

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

#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
asmlinkage long SyS_fanotify_mark(long fanotify_fd, long flags, __u64 mask,
				  long dfd, long pathname)
{
	return SYSC_fanotify_mark((int) fanotify_fd, (unsigned int) flags,
				  mask, (int) dfd,
				  (const char  __user *) pathname);
}
SYSCALL_ALIAS(sys_fanotify_mark, SyS_fanotify_mark);
#endif

/*
 * 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
22aa425d	46	static int create_fd(struct fsnotify_group group, struct fsnotify_event event)
a1014f10 EP	47	{
	48	int client_fd;
	49	struct dentry *dentry;
	50	struct vfsmount *mnt;
	51	struct file *new_file;
	52
22aa425d	53	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
a1014f10 EP	54
	55	client_fd = get_unused_fd();
	56	if (client_fd < 0)
	57	return client_fd;
	58
	59	if (event->data_type != FSNOTIFY_EVENT_PATH) {
	60	WARN_ON(1);
	61	put_unused_fd(client_fd);
	62	return -EINVAL;
	63	}
	64
	65	/*
	66	* we need a new file handle for the userspace program so it can read even if it was
	67	* originally opened O_WRONLY.
	68	*/
	69	dentry = dget(event->path.dentry);
	70	mnt = mntget(event->path.mnt);
	71	/* it's possible this event was an overflow event. in that case dentry and mnt
	72	* are NULL; That's fine, just don't call dentry open */
	73	if (dentry && mnt)
	74	new_file = dentry_open(dentry, mnt,
	75	O_RDONLY \| O_LARGEFILE \| FMODE_NONOTIFY,
	76	current_cred());
	77	else
	78	new_file = ERR_PTR(-EOVERFLOW);
	79	if (IS_ERR(new_file)) {
	80	/*
	81	* we still send an event even if we can't open the file. this
	82	* can happen when say tasks are gone and we try to open their
	83	* /proc files or we try to open a WRONLY file like in sysfs
	84	* we just send the errno to userspace since there isn't much
	85	* else we can do.
	86	*/
	87	put_unused_fd(client_fd);
	88	client_fd = PTR_ERR(new_file);
	89	} else {
	90	fd_install(client_fd, new_file);
	91	}
	92
22aa425d	93	return client_fd;
a1014f10 EP	94	}
	95
	96	static ssize_t fill_event_metadata(struct fsnotify_group *group,
	97	struct fanotify_event_metadata *metadata,
	98	struct fsnotify_event *event)
	99	{
	100	pr_debug("%s: group=%p metadata=%p event=%p\n", __func__,
	101	group, metadata, event);
	102
	103	metadata->event_len = FAN_EVENT_METADATA_LEN;
	104	metadata->vers = FANOTIFY_METADATA_VERSION;
	105	metadata->mask = fanotify_outgoing_mask(event->mask);
32c32632	106	metadata->pid = pid_vnr(event->tgid);
22aa425d	107	metadata->fd = create_fd(group, event);
a1014f10	108
22aa425d	109	return metadata->fd;
a1014f10 EP	110	}
	111
	112	static ssize_t copy_event_to_user(struct fsnotify_group *group,
	113	struct fsnotify_event *event,
	114	char __user *buf)
	115	{
	116	struct fanotify_event_metadata fanotify_event_metadata;
	117	int ret;
	118
	119	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
	120
	121	ret = fill_event_metadata(group, &fanotify_event_metadata, event);
22aa425d	122	if (ret < 0)
a1014f10 EP	123	return ret;
	124
	125	if (copy_to_user(buf, &fanotify_event_metadata, FAN_EVENT_METADATA_LEN))
	126	return -EFAULT;
	127
	128	return FAN_EVENT_METADATA_LEN;
	129	}
	130
	131	/* intofiy userspace file descriptor functions */
	132	static unsigned int fanotify_poll(struct file file, poll_table wait)
	133	{
	134	struct fsnotify_group *group = file->private_data;
	135	int ret = 0;
	136
	137	poll_wait(file, &group->notification_waitq, wait);
	138	mutex_lock(&group->notification_mutex);
	139	if (!fsnotify_notify_queue_is_empty(group))
	140	ret = POLLIN \| POLLRDNORM;
	141	mutex_unlock(&group->notification_mutex);
	142
	143	return ret;
	144	}
	145
	146	static ssize_t fanotify_read(struct file file, char __user buf,
	147	size_t count, loff_t *pos)
	148	{
	149	struct fsnotify_group *group;
	150	struct fsnotify_event *kevent;
	151	char __user *start;
	152	int ret;
	153	DEFINE_WAIT(wait);
	154
	155	start = buf;
	156	group = file->private_data;
	157
	158	pr_debug("%s: group=%p\n", __func__, group);
	159
	160	while (1) {
	161	prepare_to_wait(&group->notification_waitq, &wait, TASK_INTERRUPTIBLE);
	162
	163	mutex_lock(&group->notification_mutex);
	164	kevent = get_one_event(group, count);
	165	mutex_unlock(&group->notification_mutex);
	166
	167	if (kevent) {
	168	ret = PTR_ERR(kevent);
	169	if (IS_ERR(kevent))
	170	break;
	171	ret = copy_event_to_user(group, kevent, buf);
	172	fsnotify_put_event(kevent);
	173	if (ret < 0)
	174	break;
	175	buf += ret;
	176	count -= ret;
	177	continue;
	178	}
	179
	180	ret = -EAGAIN;
	181	if (file->f_flags & O_NONBLOCK)
	182	break;
	183	ret = -EINTR;
	184	if (signal_pending(current))
	185	break;
	186
187	if (start != buf)
188	break;
189
190	schedule();
191	}
192
193	finish_wait(&group->notification_waitq, &wait);
194	if (start != buf && ret != -EFAULT)
195	ret = buf - start;
196	return ret;
197	}
198
52c923dd EP	199	static int fanotify_release(struct inode ignored, struct file file)
	200	{
	201	struct fsnotify_group *group = file->private_data;
	202
	203	pr_debug("%s: file=%p group=%p\n", __func__, file, group);
	204
	205	/* matches the fanotify_init->fsnotify_alloc_group */
	206	fsnotify_put_group(group);
	207
	208	return 0;
	209	}
	210
a1014f10 EP	211	static long fanotify_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
	212	{
	213	struct fsnotify_group *group;
	214	struct fsnotify_event_holder *holder;
	215	void __user *p;
	216	int ret = -ENOTTY;
	217	size_t send_len = 0;
	218
	219	group = file->private_data;
	220
	221	p = (void __user *) arg;
	222
	223	switch (cmd) {
	224	case FIONREAD:
	225	mutex_lock(&group->notification_mutex);
	226	list_for_each_entry(holder, &group->notification_list, event_list)
	227	send_len += FAN_EVENT_METADATA_LEN;
	228	mutex_unlock(&group->notification_mutex);
	229	ret = put_user(send_len, (int __user *) p);
	230	break;
	231	}
	232
	233	return ret;
	234	}
	235
52c923dd	236	static const struct file_operations fanotify_fops = {
a1014f10 EP	237	.poll = fanotify_poll,
a1014f10 EP	238	.read = fanotify_read,
52c923dd EP	239	.fasync = NULL,
52c923dd EP	240	.release = fanotify_release,
a1014f10 EP	241	.unlocked_ioctl = fanotify_ioctl,
a1014f10 EP	242	.compat_ioctl = fanotify_ioctl,
52c923dd EP	243	};
52c923dd EP	244
2a3edf86 EP	245	static void fanotify_free_mark(struct fsnotify_mark *fsn_mark)
	246	{
	247	kmem_cache_free(fanotify_mark_cache, fsn_mark);
	248	}
	249
	250	static int fanotify_find_path(int dfd, const char __user *filename,
	251	struct path *path, unsigned int flags)
	252	{
	253	int ret;
	254
	255	pr_debug("%s: dfd=%d filename=%p flags=%x\n", __func__,
	256	dfd, filename, flags);
	257
	258	if (filename == NULL) {
	259	struct file *file;
	260	int fput_needed;
	261
	262	ret = -EBADF;
	263	file = fget_light(dfd, &fput_needed);
	264	if (!file)
	265	goto out;
	266
	267	ret = -ENOTDIR;
	268	if ((flags & FAN_MARK_ONLYDIR) &&
	269	!(S_ISDIR(file->f_path.dentry->d_inode->i_mode))) {
	270	fput_light(file, fput_needed);
	271	goto out;
	272	}
	273
	274	*path = file->f_path;
	275	path_get(path);
	276	fput_light(file, fput_needed);
	277	} else {
	278	unsigned int lookup_flags = 0;
	279
	280	if (!(flags & FAN_MARK_DONT_FOLLOW))
	281	lookup_flags \|= LOOKUP_FOLLOW;
	282	if (flags & FAN_MARK_ONLYDIR)
	283	lookup_flags \|= LOOKUP_DIRECTORY;
	284
	285	ret = user_path_at(dfd, filename, lookup_flags, path);
	286	if (ret)
	287	goto out;
	288	}
	289
	290	/* you can only watch an inode if you have read permissions on it */
	291	ret = inode_permission(path->dentry->d_inode, MAY_READ);
	292	if (ret)
	293	path_put(path);
	294	out:
	295	return ret;
	296	}
	297
088b09b0 AG	298	static __u32 fanotify_mark_remove_from_mask(struct fsnotify_mark *fsn_mark, __u32 mask)
	299	{
	300	__u32 oldmask;
	301
	302	spin_lock(&fsn_mark->lock);
	303	oldmask = fsn_mark->mask;
	304	fsn_mark->mask = oldmask & ~mask;
	305	spin_unlock(&fsn_mark->lock);
	306
	307	if (!(oldmask & ~mask))
	308	fsnotify_destroy_mark(fsn_mark);
	309
	310	return mask & oldmask;
	311	}
	312
88826276	313	static int fanotify_remove_mark(struct fsnotify_group group, struct inode inode,
088b09b0	314	struct vfsmount *mnt, __u32 mask)
88826276 EP	315	{
88826276 EP	316	struct fsnotify_mark *fsn_mark = NULL;
088b09b0	317	__u32 removed;
88826276 EP	318
	319	BUG_ON(inode && mnt);
	320	BUG_ON(!inode && !mnt);
	321
	322	if (inode)
	323	fsn_mark = fsnotify_find_inode_mark(group, inode);
	324	else if (mnt)
	325	fsn_mark = fsnotify_find_vfsmount_mark(group, mnt);
2a3edf86	326	else
88826276	327	BUG();
2a3edf86	328
88826276 EP	329	if (!fsn_mark)
	330	return -ENOENT;
	331
088b09b0	332	removed = fanotify_mark_remove_from_mask(fsn_mark, mask);
5444e298	333	/* matches the fsnotify_find_inode_mark() */
2a3edf86 EP	334	fsnotify_put_mark(fsn_mark);
2a3edf86 EP	335
088b09b0 AG	336	if (removed & group->mask)
	337	fsnotify_recalc_group_mask(group);
	338	if (inode) {
	339	if (removed & inode->i_fsnotify_mask)
	340	fsnotify_recalc_inode_mask(inode);
	341	} else if (mnt) {
	342	if (removed & mnt->mnt_fsnotify_mask)
	343	fsnotify_recalc_vfsmount_mask(mnt);
	344	}
	345
2a3edf86 EP	346	return 0;
	347	}
	348
912ee394 AG	349	static __u32 fanotify_mark_add_to_mask(struct fsnotify_mark *fsn_mark, __u32 mask)
	350	{
	351	__u32 oldmask;
	352
	353	spin_lock(&fsn_mark->lock);
	354	oldmask = fsn_mark->mask;
	355	fsn_mark->mask = oldmask \| mask;
	356	spin_unlock(&fsn_mark->lock);
	357
	358	return mask & ~oldmask;
	359	}
	360
52202dfb AG	361	static int fanotify_add_vfsmount_mark(struct fsnotify_group *group,
52202dfb AG	362	struct vfsmount *mnt, __u32 mask)
2a3edf86 EP	363	{
2a3edf86 EP	364	struct fsnotify_mark *fsn_mark;
912ee394	365	__u32 added;
2a3edf86	366
88826276 EP	367	fsn_mark = fsnotify_find_vfsmount_mark(group, mnt);
88826276 EP	368	if (!fsn_mark) {
88826276 EP	369	int ret;
88826276 EP	370
912ee394 AG	371	fsn_mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
912ee394 AG	372	if (!fsn_mark)
52202dfb	373	return -ENOMEM;
88826276	374
912ee394 AG	375	fsnotify_init_mark(fsn_mark, fanotify_free_mark);
912ee394 AG	376	ret = fsnotify_add_mark(fsn_mark, group, NULL, mnt, 0);
88826276	377	if (ret) {
912ee394	378	fanotify_free_mark(fsn_mark);
52202dfb	379	return ret;
88826276	380	}
88826276	381	}
912ee394	382	added = fanotify_mark_add_to_mask(fsn_mark, mask);
52202dfb	383	fsnotify_put_mark(fsn_mark);
912ee394 AG	384	if (added) {
	385	if (added & ~group->mask)
	386	fsnotify_recalc_group_mask(group);
	387	if (added & ~mnt->mnt_fsnotify_mask)
	388	fsnotify_recalc_vfsmount_mask(mnt);
	389	}
52202dfb	390	return 0;
88826276 EP	391	}
88826276 EP	392
52202dfb AG	393	static int fanotify_add_inode_mark(struct fsnotify_group *group,
52202dfb AG	394	struct inode *inode, __u32 mask)
88826276 EP	395	{
88826276 EP	396	struct fsnotify_mark *fsn_mark;
912ee394	397	__u32 added;
88826276 EP	398
88826276 EP	399	pr_debug("%s: group=%p inode=%p\n", __func__, group, inode);
2a3edf86	400
5444e298	401	fsn_mark = fsnotify_find_inode_mark(group, inode);
2a3edf86	402	if (!fsn_mark) {
88826276	403	int ret;
2a3edf86	404
912ee394 AG	405	fsn_mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
912ee394 AG	406	if (!fsn_mark)
52202dfb	407	return -ENOMEM;
2a3edf86	408
912ee394 AG	409	fsnotify_init_mark(fsn_mark, fanotify_free_mark);
912ee394 AG	410	ret = fsnotify_add_mark(fsn_mark, group, inode, NULL, 0);
2a3edf86	411	if (ret) {
912ee394	412	fanotify_free_mark(fsn_mark);
52202dfb	413	return ret;
2a3edf86	414	}
2a3edf86	415	}
912ee394	416	added = fanotify_mark_add_to_mask(fsn_mark, mask);
52202dfb	417	fsnotify_put_mark(fsn_mark);
912ee394 AG	418	if (added) {
	419	if (added & ~group->mask)
	420	fsnotify_recalc_group_mask(group);
	421	if (added & ~inode->i_fsnotify_mask)
	422	fsnotify_recalc_inode_mask(inode);
	423	}
52202dfb	424	return 0;
88826276	425	}
2a3edf86	426
2a3edf86 EP	427	static bool fanotify_mark_validate_input(int flags,
	428	__u32 mask)
	429	{
	430	pr_debug("%s: flags=%x mask=%x\n", __func__, flags, mask);
	431
	432	/* are flags valid of this operation? */
	433	if (!fanotify_mark_flags_valid(flags))
	434	return false;
	435	/* is the mask valid? */
	436	if (!fanotify_mask_valid(mask))
	437	return false;
	438	return true;
	439	}
	440
52c923dd	441	/* fanotify syscalls */
11637e4b EP	442	SYSCALL_DEFINE3(fanotify_init, unsigned int, flags, unsigned int, event_f_flags,
	443	unsigned int, priority)
	444	{
52c923dd EP	445	struct fsnotify_group *group;
	446	int f_flags, fd;
	447
	448	pr_debug("%s: flags=%d event_f_flags=%d priority=%d\n",
	449	__func__, flags, event_f_flags, priority);
	450
	451	if (event_f_flags)
	452	return -EINVAL;
	453	if (priority)
	454	return -EINVAL;
	455
	456	if (!capable(CAP_SYS_ADMIN))
	457	return -EACCES;
	458
	459	if (flags & ~FAN_ALL_INIT_FLAGS)
	460	return -EINVAL;
	461
	462	f_flags = (O_RDONLY \| FMODE_NONOTIFY);
	463	if (flags & FAN_CLOEXEC)
	464	f_flags \|= O_CLOEXEC;
	465	if (flags & FAN_NONBLOCK)
	466	f_flags \|= O_NONBLOCK;
	467
	468	/* fsnotify_alloc_group takes a ref. Dropped in fanotify_release */
	469	group = fsnotify_alloc_group(&fanotify_fsnotify_ops);
	470	if (IS_ERR(group))
	471	return PTR_ERR(group);
	472
	473	fd = anon_inode_getfd("[fanotify]", &fanotify_fops, group, f_flags);
	474	if (fd < 0)
	475	goto out_put_group;
	476
	477	return fd;
	478
	479	out_put_group:
	480	fsnotify_put_group(group);
	481	return fd;
11637e4b	482	}
bbaa4168	483
9bbfc964 HC	484	SYSCALL_DEFINE(fanotify_mark)(int fanotify_fd, unsigned int flags,
	485	__u64 mask, int dfd,
	486	const char __user * pathname)
bbaa4168	487	{
0ff21db9 EP	488	struct inode *inode = NULL;
0ff21db9 EP	489	struct vfsmount *mnt = NULL;
2a3edf86 EP	490	struct fsnotify_group *group;
	491	struct file *filp;
	492	struct path path;
	493	int ret, fput_needed;
	494
	495	pr_debug("%s: fanotify_fd=%d flags=%x dfd=%d pathname=%p mask=%llx\n",
	496	__func__, fanotify_fd, flags, dfd, pathname, mask);
	497
	498	/* we only use the lower 32 bits as of right now. */
	499	if (mask & ((__u64)0xffffffff << 32))
	500	return -EINVAL;
	501
	502	if (!fanotify_mark_validate_input(flags, mask))
	503	return -EINVAL;
	504
	505	filp = fget_light(fanotify_fd, &fput_needed);
	506	if (unlikely(!filp))
	507	return -EBADF;
	508
	509	/* verify that this is indeed an fanotify instance */
	510	ret = -EINVAL;
	511	if (unlikely(filp->f_op != &fanotify_fops))
	512	goto fput_and_out;
	513
	514	ret = fanotify_find_path(dfd, pathname, &path, flags);
	515	if (ret)
	516	goto fput_and_out;
	517
	518	/* inode held in place by reference to path; group by fget on fd */
eac8e9e8	519	if (!(flags & FAN_MARK_MOUNT))
0ff21db9 EP	520	inode = path.dentry->d_inode;
	521	else
	522	mnt = path.mnt;
2a3edf86 EP	523	group = filp->private_data;
	524
	525	/* create/update an inode mark */
c6223f46 AG	526	switch (flags & (FAN_MARK_ADD \| FAN_MARK_REMOVE)) {
c6223f46 AG	527	case FAN_MARK_ADD:
eac8e9e8	528	if (flags & FAN_MARK_MOUNT)
0ff21db9 EP	529	ret = fanotify_add_vfsmount_mark(group, mnt, mask);
	530	else
	531	ret = fanotify_add_inode_mark(group, inode, mask);
c6223f46 AG	532	break;
c6223f46 AG	533	case FAN_MARK_REMOVE:
0ff21db9	534	ret = fanotify_remove_mark(group, inode, mnt, mask);
c6223f46 AG	535	break;
	536	default:
	537	ret = -EINVAL;
	538	}
2a3edf86 EP	539
	540	path_put(&path);
	541	fput_and_out:
	542	fput_light(filp, fput_needed);
	543	return ret;
	544	}
	545
9bbfc964 HC	546	#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
	547	asmlinkage long SyS_fanotify_mark(long fanotify_fd, long flags, __u64 mask,
	548	long dfd, long pathname)
	549	{
	550	return SYSC_fanotify_mark((int) fanotify_fd, (unsigned int) flags,
	551	mask, (int) dfd,
	552	(const char __user *) pathname);
	553	}
	554	SYSCALL_ALIAS(sys_fanotify_mark, SyS_fanotify_mark);
	555	#endif
	556
2a3edf86 EP	557	/*
	558	* fanotify_user_setup - Our initialization function. Note that we cannnot return
	559	* error because we have compiled-in VFS hooks. So an (unlikely) failure here
	560	* must result in panic().
	561	*/
	562	static int __init fanotify_user_setup(void)
	563	{
	564	fanotify_mark_cache = KMEM_CACHE(fsnotify_mark, SLAB_PANIC);
	565
	566	return 0;
bbaa4168	567	}
2a3edf86	568	device_initcall(fanotify_user_setup);