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

#include <linux/fanotify.h>
#include <linux/fdtable.h>
#include <linux/fsnotify_backend.h>
#include <linux/init.h>
#include <linux/jiffies.h>
#include <linux/kernel.h> /* UINT_MAX */
#include <linux/mount.h>
#include <linux/sched.h>
#include <linux/types.h>
#include <linux/wait.h>

static bool should_merge(struct fsnotify_event *old, struct fsnotify_event *new)
{
	pr_debug("%s: old=%p new=%p\n", __func__, old, new);

	if (old->to_tell == new->to_tell &&
	    old->data_type == new->data_type &&
	    old->tgid == new->tgid) {
		switch (old->data_type) {
		case (FSNOTIFY_EVENT_PATH):
			if ((old->path.mnt == new->path.mnt) &&
			    (old->path.dentry == new->path.dentry))
				return true;
		case (FSNOTIFY_EVENT_NONE):
			return true;
		default:
			BUG();
		};
	}
	return false;
}

/* Note, if we return an event in *arg that a reference is being held... */
static int fanotify_merge(struct list_head *list,
			  struct fsnotify_event *event,
			  void **arg)
{
	struct fsnotify_event_holder *test_holder;
	struct fsnotify_event *test_event;
	struct fsnotify_event *new_event;
	struct fsnotify_event **return_event = (struct fsnotify_event **)arg;
	int ret = 0;

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

	*return_event = NULL;

	/* and the list better be locked by something too! */

	list_for_each_entry_reverse(test_holder, list, event_list) {
		test_event = test_holder->event;
		if (should_merge(test_event, event)) {
			fsnotify_get_event(test_event);
			*return_event = test_event;

			ret = -EEXIST;
			/* if they are exactly the same we are done */
			if (test_event->mask == event->mask)
				goto out;

			/*
			 * if the refcnt == 1 this is the only queue
			 * for this event and so we can update the mask
			 * in place.
			 */
			if (atomic_read(&test_event->refcnt) == 1) {
				test_event->mask |= event->mask;
				goto out;
			}

			/* can't allocate memory, merge was no possible */
			new_event = fsnotify_clone_event(test_event);
			if (unlikely(!new_event)) {
				ret = 0;
				goto out;
			}

			/* we didn't return the test_event, so drop that ref */
			fsnotify_put_event(test_event);
			/* the reference we return on new_event is from clone */
			*return_event = new_event;

			/* build new event and replace it on the list */
			new_event->mask = (test_event->mask | event->mask);
			fsnotify_replace_event(test_holder, new_event);

			break;
		}
	}
out:
	return ret;
}

#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
static int fanotify_get_response_from_access(struct fsnotify_group *group,
					     struct fsnotify_event *event)
{
	int ret;

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

	wait_event(group->fanotify_data.access_waitq, event->response);

	/* userspace responded, convert to something usable */
	spin_lock(&event->lock);
	switch (event->response) {
	case FAN_ALLOW:
		ret = 0;
		break;
	case FAN_DENY:
	default:
		ret = -EPERM;
	}
	event->response = 0;
	spin_unlock(&event->lock);

	return ret;
}
#endif

static int fanotify_handle_event(struct fsnotify_group *group, struct fsnotify_event *event)
{
	int ret;
	struct fsnotify_event *notify_event = NULL;

	BUILD_BUG_ON(FAN_ACCESS != FS_ACCESS);
	BUILD_BUG_ON(FAN_MODIFY != FS_MODIFY);
	BUILD_BUG_ON(FAN_CLOSE_NOWRITE != FS_CLOSE_NOWRITE);
	BUILD_BUG_ON(FAN_CLOSE_WRITE != FS_CLOSE_WRITE);
	BUILD_BUG_ON(FAN_OPEN != FS_OPEN);
	BUILD_BUG_ON(FAN_EVENT_ON_CHILD != FS_EVENT_ON_CHILD);
	BUILD_BUG_ON(FAN_Q_OVERFLOW != FS_Q_OVERFLOW);
	BUILD_BUG_ON(FAN_OPEN_PERM != FS_OPEN_PERM);
	BUILD_BUG_ON(FAN_ACCESS_PERM != FS_ACCESS_PERM);

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

	ret = fsnotify_add_notify_event(group, event, NULL, fanotify_merge,
					(void **)&notify_event);
	/* -EEXIST means this event was merged with another, not that it was an error */
	if (ret == -EEXIST)
		ret = 0;
	if (ret)
		goto out;

#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
	if (event->mask & FAN_ALL_PERM_EVENTS) {
		/* if we merged we need to wait on the new event */
		if (notify_event)
			event = notify_event;
		ret = fanotify_get_response_from_access(group, event);
	}
#endif

out:
	if (notify_event)
		fsnotify_put_event(notify_event);
	return ret;
}

static bool should_send_vfsmount_event(struct fsnotify_group *group, struct vfsmount *mnt,
				       struct inode *inode, __u32 mask)
{
	struct fsnotify_mark *mnt_mark;
	struct fsnotify_mark *inode_mark;

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

	mnt_mark = fsnotify_find_vfsmount_mark(group, mnt);
	if (!mnt_mark)
		return false;

	mask &= mnt_mark->mask;
	mask &= ~mnt_mark->ignored_mask;

	if (mask) {
		inode_mark = fsnotify_find_inode_mark(group, inode);
		if (inode_mark) {
			mask &= ~inode_mark->ignored_mask;
			fsnotify_put_mark(inode_mark);
		}
	}

	/* find took a reference */
	fsnotify_put_mark(mnt_mark);

	return mask;
}

static bool should_send_inode_event(struct fsnotify_group *group, struct inode *inode,
				    __u32 mask)
{
	struct fsnotify_mark *fsn_mark;

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

	fsn_mark = fsnotify_find_inode_mark(group, inode);
	if (!fsn_mark)
		return false;

	/* if the event is for a child and this inode doesn't care about
	 * events on the child, don't send it! */
	if ((mask & FS_EVENT_ON_CHILD) &&
	    !(fsn_mark->mask & FS_EVENT_ON_CHILD)) {
		mask = 0;
	} else {
		/*
		 * We care about children, but do we care about this particular
		 * type of event?
		 */
		mask &= ~FS_EVENT_ON_CHILD;
		mask &= fsn_mark->mask;
		mask &= ~fsn_mark->ignored_mask;
	}

	/* find took a reference */
	fsnotify_put_mark(fsn_mark);

	return mask;
}

static bool fanotify_should_send_event(struct fsnotify_group *group, struct inode *to_tell,
				       struct vfsmount *mnt, __u32 mask, void *data,
				       int data_type)
{
	pr_debug("%s: group=%p to_tell=%p mnt=%p mask=%x data=%p data_type=%d\n",
		 __func__, group, to_tell, mnt, mask, data, data_type);

	/* sorry, fanotify only gives a damn about files and dirs */
	if (!S_ISREG(to_tell->i_mode) &&
	    !S_ISDIR(to_tell->i_mode))
		return false;

	/* if we don't have enough info to send an event to userspace say no */
	if (data_type != FSNOTIFY_EVENT_PATH)
		return false;

	if (mnt)
		return should_send_vfsmount_event(group, mnt, to_tell, mask);
	else
		return should_send_inode_event(group, to_tell, mask);
}

const struct fsnotify_ops fanotify_fsnotify_ops = {
	.handle_event = fanotify_handle_event,
	.should_send_event = fanotify_should_send_event,
	.free_group_priv = NULL,
	.free_event_priv = NULL,
	.freeing_mark = NULL,
};
Commit	Line	Data
	1	#include <linux/fanotify.h>
	2	#include <linux/fdtable.h>
	3	#include <linux/fsnotify_backend.h>
	4	#include <linux/init.h>
	5	#include <linux/jiffies.h>
	6	#include <linux/kernel.h> /* UINT_MAX */
	7	#include <linux/mount.h>
	8	#include <linux/sched.h>
	9	#include <linux/types.h>
	10	#include <linux/wait.h>
	11
	12	static bool should_merge(struct fsnotify_event old, struct fsnotify_event new)
	13	{
	14	pr_debug("%s: old=%p new=%p\n", __func__, old, new);
	15
	16	if (old->to_tell == new->to_tell &&
	17	old->data_type == new->data_type &&
	18	old->tgid == new->tgid) {
	19	switch (old->data_type) {
	20	case (FSNOTIFY_EVENT_PATH):
	21	if ((old->path.mnt == new->path.mnt) &&
	22	(old->path.dentry == new->path.dentry))
	23	return true;
	24	case (FSNOTIFY_EVENT_NONE):
	25	return true;
	26	default:
	27	BUG();
	28	};
	29	}
	30	return false;
	31	}
	32
	33	/* Note, if we return an event in arg that a reference is being held... /
	34	static int fanotify_merge(struct list_head *list,
	35	struct fsnotify_event *event,
	36	void **arg)
	37	{
	38	struct fsnotify_event_holder *test_holder;
	39	struct fsnotify_event *test_event;
	40	struct fsnotify_event *new_event;
	41	struct fsnotify_event return_event = (struct fsnotify_event )arg;
	42	int ret = 0;
	43
	44	pr_debug("%s: list=%p event=%p\n", __func__, list, event);
	45
	46	*return_event = NULL;
	47
	48	/* and the list better be locked by something too! */
	49
	50	list_for_each_entry_reverse(test_holder, list, event_list) {
	51	test_event = test_holder->event;
	52	if (should_merge(test_event, event)) {
	53	fsnotify_get_event(test_event);
	54	*return_event = test_event;
	55
	56	ret = -EEXIST;
	57	/* if they are exactly the same we are done */
	58	if (test_event->mask == event->mask)
	59	goto out;
	60
	61	/*
	62	* if the refcnt == 1 this is the only queue
	63	* for this event and so we can update the mask
	64	* in place.
	65	*/
	66	if (atomic_read(&test_event->refcnt) == 1) {
	67	test_event->mask \|= event->mask;
	68	goto out;
	69	}
	70
	71	/* can't allocate memory, merge was no possible */
	72	new_event = fsnotify_clone_event(test_event);
	73	if (unlikely(!new_event)) {
	74	ret = 0;
	75	goto out;
	76	}
	77
	78	/* we didn't return the test_event, so drop that ref */
	79	fsnotify_put_event(test_event);
	80	/* the reference we return on new_event is from clone */
	81	*return_event = new_event;
	82
	83	/* build new event and replace it on the list */
	84	new_event->mask = (test_event->mask \| event->mask);
	85	fsnotify_replace_event(test_holder, new_event);
	86
	87	break;
	88	}
	89	}
	90	out:
	91	return ret;
	92	}
	93
	94	#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
	95	static int fanotify_get_response_from_access(struct fsnotify_group *group,
	96	struct fsnotify_event *event)
	97	{
	98	int ret;
	99
	100	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
	101
	102	wait_event(group->fanotify_data.access_waitq, event->response);
	103
	104	/* userspace responded, convert to something usable */
	105	spin_lock(&event->lock);
	106	switch (event->response) {
	107	case FAN_ALLOW:
	108	ret = 0;
	109	break;
	110	case FAN_DENY:
	111	default:
	112	ret = -EPERM;
	113	}
	114	event->response = 0;
	115	spin_unlock(&event->lock);
	116
	117	return ret;
	118	}
	119	#endif
	120
	121	static int fanotify_handle_event(struct fsnotify_group group, struct fsnotify_event event)
	122	{
	123	int ret;
	124	struct fsnotify_event *notify_event = NULL;
	125
	126	BUILD_BUG_ON(FAN_ACCESS != FS_ACCESS);
	127	BUILD_BUG_ON(FAN_MODIFY != FS_MODIFY);
	128	BUILD_BUG_ON(FAN_CLOSE_NOWRITE != FS_CLOSE_NOWRITE);
	129	BUILD_BUG_ON(FAN_CLOSE_WRITE != FS_CLOSE_WRITE);
	130	BUILD_BUG_ON(FAN_OPEN != FS_OPEN);
	131	BUILD_BUG_ON(FAN_EVENT_ON_CHILD != FS_EVENT_ON_CHILD);
	132	BUILD_BUG_ON(FAN_Q_OVERFLOW != FS_Q_OVERFLOW);
	133	BUILD_BUG_ON(FAN_OPEN_PERM != FS_OPEN_PERM);
	134	BUILD_BUG_ON(FAN_ACCESS_PERM != FS_ACCESS_PERM);
	135
	136	pr_debug("%s: group=%p event=%p\n", __func__, group, event);
	137
	138	ret = fsnotify_add_notify_event(group, event, NULL, fanotify_merge,
	139	(void **)&notify_event);
	140	/* -EEXIST means this event was merged with another, not that it was an error */
	141	if (ret == -EEXIST)
	142	ret = 0;
	143	if (ret)
	144	goto out;
	145
	146	#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
	147	if (event->mask & FAN_ALL_PERM_EVENTS) {
	148	/* if we merged we need to wait on the new event */
	149	if (notify_event)
	150	event = notify_event;
	151	ret = fanotify_get_response_from_access(group, event);
	152	}
	153	#endif
	154
	155	out:
	156	if (notify_event)
	157	fsnotify_put_event(notify_event);
	158	return ret;
	159	}
	160
	161	static bool should_send_vfsmount_event(struct fsnotify_group group, struct vfsmount mnt,
	162	struct inode *inode, __u32 mask)
	163	{
	164	struct fsnotify_mark *mnt_mark;
	165	struct fsnotify_mark *inode_mark;
	166
	167	pr_debug("%s: group=%p vfsmount=%p mask=%x\n",
	168	__func__, group, mnt, mask);
	169
	170	mnt_mark = fsnotify_find_vfsmount_mark(group, mnt);
	171	if (!mnt_mark)
	172	return false;
	173
	174	mask &= mnt_mark->mask;
	175	mask &= ~mnt_mark->ignored_mask;
	176
	177	if (mask) {
	178	inode_mark = fsnotify_find_inode_mark(group, inode);
	179	if (inode_mark) {
	180	mask &= ~inode_mark->ignored_mask;
	181	fsnotify_put_mark(inode_mark);
	182	}
	183	}
	184
	185	/* find took a reference */
	186	fsnotify_put_mark(mnt_mark);
	187
	188	return mask;
	189	}
	190
	191	static bool should_send_inode_event(struct fsnotify_group group, struct inode inode,
	192	__u32 mask)
	193	{
	194	struct fsnotify_mark *fsn_mark;
	195
	196	pr_debug("%s: group=%p inode=%p mask=%x\n",
	197	__func__, group, inode, mask);
	198
	199	fsn_mark = fsnotify_find_inode_mark(group, inode);
	200	if (!fsn_mark)
	201	return false;
	202
	203	/* if the event is for a child and this inode doesn't care about
	204	* events on the child, don't send it! */
	205	if ((mask & FS_EVENT_ON_CHILD) &&
	206	!(fsn_mark->mask & FS_EVENT_ON_CHILD)) {
	207	mask = 0;
	208	} else {
	209	/*
	210	* We care about children, but do we care about this particular
	211	* type of event?
	212	*/
	213	mask &= ~FS_EVENT_ON_CHILD;
	214	mask &= fsn_mark->mask;
	215	mask &= ~fsn_mark->ignored_mask;
	216	}
	217
	218	/* find took a reference */
	219	fsnotify_put_mark(fsn_mark);
	220
	221	return mask;
	222	}
	223
	224	static bool fanotify_should_send_event(struct fsnotify_group group, struct inode to_tell,
	225	struct vfsmount mnt, __u32 mask, void data,
	226	int data_type)
	227	{
	228	pr_debug("%s: group=%p to_tell=%p mnt=%p mask=%x data=%p data_type=%d\n",
	229	__func__, group, to_tell, mnt, mask, data, data_type);
	230
	231	/* sorry, fanotify only gives a damn about files and dirs */
	232	if (!S_ISREG(to_tell->i_mode) &&
	233	!S_ISDIR(to_tell->i_mode))
	234	return false;
	235
	236	/* if we don't have enough info to send an event to userspace say no */
	237	if (data_type != FSNOTIFY_EVENT_PATH)
	238	return false;
	239
	240	if (mnt)
	241	return should_send_vfsmount_event(group, mnt, to_tell, mask);
	242	else
	243	return should_send_inode_event(group, to_tell, mask);
	244	}
	245
	246	const struct fsnotify_ops fanotify_fsnotify_ops = {
	247	.handle_event = fanotify_handle_event,
	248	.should_send_event = fanotify_should_send_event,
	249	.free_group_priv = NULL,
	250	.free_event_priv = NULL,
	251	.freeing_mark = NULL,
	252	};