[deliverable/linux.git] / kernel / bpf / hashtab.c

/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of version 2 of the GNU General Public
 * License as published by the Free Software Foundation.
 *
 * 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/bpf.h>
#include <linux/jhash.h>
#include <linux/filter.h>
#include <linux/vmalloc.h>

struct bpf_htab {
	struct bpf_map map;
	struct hlist_head *buckets;
	raw_spinlock_t lock;
	u32 count;	/* number of elements in this hashtable */
	u32 n_buckets;	/* number of hash buckets */
	u32 elem_size;	/* size of each element in bytes */
};

/* each htab element is struct htab_elem + key + value */
struct htab_elem {
	struct hlist_node hash_node;
	struct rcu_head rcu;
	u32 hash;
	char key[0] __aligned(8);
};

/* Called from syscall */
static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
{
	struct bpf_htab *htab;
	int err, i;

	htab = kzalloc(sizeof(*htab), GFP_USER);
	if (!htab)
		return ERR_PTR(-ENOMEM);

	/* mandatory map attributes */
	htab->map.key_size = attr->key_size;
	htab->map.value_size = attr->value_size;
	htab->map.max_entries = attr->max_entries;

	/* check sanity of attributes.
	 * value_size == 0 may be allowed in the future to use map as a set
	 */
	err = -EINVAL;
	if (htab->map.max_entries == 0 || htab->map.key_size == 0 ||
	    htab->map.value_size == 0)
		goto free_htab;

	/* hash table size must be power of 2 */
	htab->n_buckets = roundup_pow_of_two(htab->map.max_entries);

	err = -E2BIG;
	if (htab->map.key_size > MAX_BPF_STACK)
		/* eBPF programs initialize keys on stack, so they cannot be
		 * larger than max stack size
		 */
		goto free_htab;

	err = -ENOMEM;
	/* prevent zero size kmalloc and check for u32 overflow */
	if (htab->n_buckets == 0 ||
	    htab->n_buckets > U32_MAX / sizeof(struct hlist_head))
		goto free_htab;

	htab->buckets = kmalloc_array(htab->n_buckets, sizeof(struct hlist_head),
				      GFP_USER | __GFP_NOWARN);

	if (!htab->buckets) {
		htab->buckets = vmalloc(htab->n_buckets * sizeof(struct hlist_head));
		if (!htab->buckets)
			goto free_htab;
	}

	for (i = 0; i < htab->n_buckets; i++)
		INIT_HLIST_HEAD(&htab->buckets[i]);

	raw_spin_lock_init(&htab->lock);
	htab->count = 0;

	htab->elem_size = sizeof(struct htab_elem) +
			  round_up(htab->map.key_size, 8) +
			  htab->map.value_size;

	htab->map.pages = round_up(htab->n_buckets * sizeof(struct hlist_head) +
				   htab->elem_size * htab->map.max_entries,
				   PAGE_SIZE) >> PAGE_SHIFT;
	return &htab->map;

free_htab:
	kfree(htab);
	return ERR_PTR(err);
}

static inline u32 htab_map_hash(const void *key, u32 key_len)
{
	return jhash(key, key_len, 0);
}

static inline struct hlist_head *select_bucket(struct bpf_htab *htab, u32 hash)
{
	return &htab->buckets[hash & (htab->n_buckets - 1)];
}

static struct htab_elem *lookup_elem_raw(struct hlist_head *head, u32 hash,
					 void *key, u32 key_size)
{
	struct htab_elem *l;

	hlist_for_each_entry_rcu(l, head, hash_node)
		if (l->hash == hash && !memcmp(&l->key, key, key_size))
			return l;

	return NULL;
}

/* Called from syscall or from eBPF program */
static void *htab_map_lookup_elem(struct bpf_map *map, void *key)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	struct hlist_head *head;
	struct htab_elem *l;
	u32 hash, key_size;

	/* Must be called with rcu_read_lock. */
	WARN_ON_ONCE(!rcu_read_lock_held());

	key_size = map->key_size;

	hash = htab_map_hash(key, key_size);

	head = select_bucket(htab, hash);

	l = lookup_elem_raw(head, hash, key, key_size);

	if (l)
		return l->key + round_up(map->key_size, 8);

	return NULL;
}

/* Called from syscall */
static int htab_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	struct hlist_head *head;
	struct htab_elem *l, *next_l;
	u32 hash, key_size;
	int i;

	WARN_ON_ONCE(!rcu_read_lock_held());

	key_size = map->key_size;

	hash = htab_map_hash(key, key_size);

	head = select_bucket(htab, hash);

	/* lookup the key */
	l = lookup_elem_raw(head, hash, key, key_size);

	if (!l) {
		i = 0;
		goto find_first_elem;
	}

	/* key was found, get next key in the same bucket */
	next_l = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&l->hash_node)),
				  struct htab_elem, hash_node);

	if (next_l) {
		/* if next elem in this hash list is non-zero, just return it */
		memcpy(next_key, next_l->key, key_size);
		return 0;
	}

	/* no more elements in this hash list, go to the next bucket */
	i = hash & (htab->n_buckets - 1);
	i++;

find_first_elem:
	/* iterate over buckets */
	for (; i < htab->n_buckets; i++) {
		head = select_bucket(htab, i);

		/* pick first element in the bucket */
		next_l = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
					  struct htab_elem, hash_node);
		if (next_l) {
			/* if it's not empty, just return it */
			memcpy(next_key, next_l->key, key_size);
			return 0;
		}
	}

	/* itereated over all buckets and all elements */
	return -ENOENT;
}

/* Called from syscall or from eBPF program */
static int htab_map_update_elem(struct bpf_map *map, void *key, void *value,
				u64 map_flags)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	struct htab_elem *l_new, *l_old;
	struct hlist_head *head;
	unsigned long flags;
	u32 key_size;
	int ret;

	if (map_flags > BPF_EXIST)
		/* unknown flags */
		return -EINVAL;

	WARN_ON_ONCE(!rcu_read_lock_held());

	/* allocate new element outside of lock */
	l_new = kmalloc(htab->elem_size, GFP_ATOMIC);
	if (!l_new)
		return -ENOMEM;

	key_size = map->key_size;

	memcpy(l_new->key, key, key_size);
	memcpy(l_new->key + round_up(key_size, 8), value, map->value_size);

	l_new->hash = htab_map_hash(l_new->key, key_size);

	/* bpf_map_update_elem() can be called in_irq() */
	raw_spin_lock_irqsave(&htab->lock, flags);

	head = select_bucket(htab, l_new->hash);

	l_old = lookup_elem_raw(head, l_new->hash, key, key_size);

	if (!l_old && unlikely(htab->count >= map->max_entries)) {
		/* if elem with this 'key' doesn't exist and we've reached
		 * max_entries limit, fail insertion of new elem
		 */
		ret = -E2BIG;
		goto err;
	}

	if (l_old && map_flags == BPF_NOEXIST) {
		/* elem already exists */
		ret = -EEXIST;
		goto err;
	}

	if (!l_old && map_flags == BPF_EXIST) {
		/* elem doesn't exist, cannot update it */
		ret = -ENOENT;
		goto err;
	}

	/* add new element to the head of the list, so that concurrent
	 * search will find it before old elem
	 */
	hlist_add_head_rcu(&l_new->hash_node, head);
	if (l_old) {
		hlist_del_rcu(&l_old->hash_node);
		kfree_rcu(l_old, rcu);
	} else {
		htab->count++;
	}
	raw_spin_unlock_irqrestore(&htab->lock, flags);

	return 0;
err:
	raw_spin_unlock_irqrestore(&htab->lock, flags);
	kfree(l_new);
	return ret;
}

/* Called from syscall or from eBPF program */
static int htab_map_delete_elem(struct bpf_map *map, void *key)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	struct hlist_head *head;
	struct htab_elem *l;
	unsigned long flags;
	u32 hash, key_size;
	int ret = -ENOENT;

	WARN_ON_ONCE(!rcu_read_lock_held());

	key_size = map->key_size;

	hash = htab_map_hash(key, key_size);

	raw_spin_lock_irqsave(&htab->lock, flags);

	head = select_bucket(htab, hash);

	l = lookup_elem_raw(head, hash, key, key_size);

	if (l) {
		hlist_del_rcu(&l->hash_node);
		htab->count--;
		kfree_rcu(l, rcu);
		ret = 0;
	}

	raw_spin_unlock_irqrestore(&htab->lock, flags);
	return ret;
}

static void delete_all_elements(struct bpf_htab *htab)
{
	int i;

	for (i = 0; i < htab->n_buckets; i++) {
		struct hlist_head *head = select_bucket(htab, i);
		struct hlist_node *n;
		struct htab_elem *l;

		hlist_for_each_entry_safe(l, n, head, hash_node) {
			hlist_del_rcu(&l->hash_node);
			htab->count--;
			kfree(l);
		}
	}
}

/* Called when map->refcnt goes to zero, either from workqueue or from syscall */
static void htab_map_free(struct bpf_map *map)
{
	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);

	/* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
	 * so the programs (can be more than one that used this map) were
	 * disconnected from events. Wait for outstanding critical sections in
	 * these programs to complete
	 */
	synchronize_rcu();

	/* some of kfree_rcu() callbacks for elements of this map may not have
	 * executed. It's ok. Proceed to free residual elements and map itself
	 */
	delete_all_elements(htab);
	kvfree(htab->buckets);
	kfree(htab);
}

static const struct bpf_map_ops htab_ops = {
	.map_alloc = htab_map_alloc,
	.map_free = htab_map_free,
	.map_get_next_key = htab_map_get_next_key,
	.map_lookup_elem = htab_map_lookup_elem,
	.map_update_elem = htab_map_update_elem,
	.map_delete_elem = htab_map_delete_elem,
};

static struct bpf_map_type_list htab_type __read_mostly = {
	.ops = &htab_ops,
	.type = BPF_MAP_TYPE_HASH,
};

static int __init register_htab_map(void)
{
	bpf_register_map_type(&htab_type);
	return 0;
}
late_initcall(register_htab_map);
Commit	Line	Data
0f8e4bd8 AS	1	/* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
	2	*
	3	* This program is free software; you can redistribute it and/or
	4	* modify it under the terms of version 2 of the GNU General Public
	5	* License as published by the Free Software Foundation.
	6	*
	7	* This program is distributed in the hope that it will be useful, but
	8	* WITHOUT ANY WARRANTY; without even the implied warranty of
	9	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	10	* General Public License for more details.
	11	*/
	12	#include <linux/bpf.h>
	13	#include <linux/jhash.h>
	14	#include <linux/filter.h>
	15	#include <linux/vmalloc.h>
	16
	17	struct bpf_htab {
	18	struct bpf_map map;
	19	struct hlist_head *buckets;
ac00881f	20	raw_spinlock_t lock;
0f8e4bd8 AS	21	u32 count; /* number of elements in this hashtable */
	22	u32 n_buckets; /* number of hash buckets */
	23	u32 elem_size; /* size of each element in bytes */
	24	};
	25
	26	/* each htab element is struct htab_elem + key + value */
	27	struct htab_elem {
	28	struct hlist_node hash_node;
	29	struct rcu_head rcu;
	30	u32 hash;
	31	char key[0] __aligned(8);
	32	};
	33
	34	/* Called from syscall */
	35	static struct bpf_map htab_map_alloc(union bpf_attr attr)
	36	{
	37	struct bpf_htab *htab;
	38	int err, i;
	39
	40	htab = kzalloc(sizeof(*htab), GFP_USER);
	41	if (!htab)
	42	return ERR_PTR(-ENOMEM);
	43
	44	/* mandatory map attributes */
	45	htab->map.key_size = attr->key_size;
	46	htab->map.value_size = attr->value_size;
	47	htab->map.max_entries = attr->max_entries;
	48
	49	/* check sanity of attributes.
	50	* value_size == 0 may be allowed in the future to use map as a set
	51	*/
	52	err = -EINVAL;
	53	if (htab->map.max_entries == 0 \|\| htab->map.key_size == 0 \|\|
	54	htab->map.value_size == 0)
	55	goto free_htab;
	56
	57	/* hash table size must be power of 2 */
	58	htab->n_buckets = roundup_pow_of_two(htab->map.max_entries);
	59
	60	err = -E2BIG;
	61	if (htab->map.key_size > MAX_BPF_STACK)
	62	/* eBPF programs initialize keys on stack, so they cannot be
	63	* larger than max stack size
	64	*/
	65	goto free_htab;
	66
	67	err = -ENOMEM;
daaf427c AS	68	/* prevent zero size kmalloc and check for u32 overflow */
	69	if (htab->n_buckets == 0 \|\|
	70	htab->n_buckets > U32_MAX / sizeof(struct hlist_head))
	71	goto free_htab;
	72
0f8e4bd8 AS	73	htab->buckets = kmalloc_array(htab->n_buckets, sizeof(struct hlist_head),
	74	GFP_USER \| __GFP_NOWARN);
	75
	76	if (!htab->buckets) {
	77	htab->buckets = vmalloc(htab->n_buckets * sizeof(struct hlist_head));
	78	if (!htab->buckets)
	79	goto free_htab;
	80	}
	81
	82	for (i = 0; i < htab->n_buckets; i++)
	83	INIT_HLIST_HEAD(&htab->buckets[i]);
	84
ac00881f	85	raw_spin_lock_init(&htab->lock);
0f8e4bd8 AS	86	htab->count = 0;
	87
	88	htab->elem_size = sizeof(struct htab_elem) +
	89	round_up(htab->map.key_size, 8) +
	90	htab->map.value_size;
aaac3ba9 AS	91
	92	htab->map.pages = round_up(htab->n_buckets * sizeof(struct hlist_head) +
	93	htab->elem_size * htab->map.max_entries,
	94	PAGE_SIZE) >> PAGE_SHIFT;
0f8e4bd8 AS	95	return &htab->map;
	96
	97	free_htab:
	98	kfree(htab);
	99	return ERR_PTR(err);
	100	}
	101
	102	static inline u32 htab_map_hash(const void *key, u32 key_len)
	103	{
	104	return jhash(key, key_len, 0);
	105	}
	106
	107	static inline struct hlist_head select_bucket(struct bpf_htab htab, u32 hash)
	108	{
	109	return &htab->buckets[hash & (htab->n_buckets - 1)];
	110	}
	111
	112	static struct htab_elem lookup_elem_raw(struct hlist_head head, u32 hash,
	113	void *key, u32 key_size)
	114	{
	115	struct htab_elem *l;
	116
	117	hlist_for_each_entry_rcu(l, head, hash_node)
	118	if (l->hash == hash && !memcmp(&l->key, key, key_size))
	119	return l;
	120
	121	return NULL;
	122	}
	123
	124	/* Called from syscall or from eBPF program */
	125	static void htab_map_lookup_elem(struct bpf_map map, void *key)
	126	{
	127	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	128	struct hlist_head *head;
	129	struct htab_elem *l;
	130	u32 hash, key_size;
	131
	132	/* Must be called with rcu_read_lock. */
	133	WARN_ON_ONCE(!rcu_read_lock_held());
	134
	135	key_size = map->key_size;
	136
	137	hash = htab_map_hash(key, key_size);
	138
	139	head = select_bucket(htab, hash);
	140
	141	l = lookup_elem_raw(head, hash, key, key_size);
	142
	143	if (l)
	144	return l->key + round_up(map->key_size, 8);
	145
	146	return NULL;
	147	}
	148
	149	/* Called from syscall */
	150	static int htab_map_get_next_key(struct bpf_map map, void key, void *next_key)
	151	{
	152	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	153	struct hlist_head *head;
	154	struct htab_elem l, next_l;
	155	u32 hash, key_size;
	156	int i;
	157
	158	WARN_ON_ONCE(!rcu_read_lock_held());
159
160	key_size = map->key_size;
161
162	hash = htab_map_hash(key, key_size);
163
164	head = select_bucket(htab, hash);
165
166	/* lookup the key */
167	l = lookup_elem_raw(head, hash, key, key_size);
168
169	if (!l) {
170	i = 0;
171	goto find_first_elem;
172	}
173
174	/* key was found, get next key in the same bucket */
175	next_l = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&l->hash_node)),
176	struct htab_elem, hash_node);
177
178	if (next_l) {
179	/* if next elem in this hash list is non-zero, just return it */
180	memcpy(next_key, next_l->key, key_size);
181	return 0;
182	}
183
184	/* no more elements in this hash list, go to the next bucket */
185	i = hash & (htab->n_buckets - 1);
186	i++;
187
188	find_first_elem:
189	/* iterate over buckets */
190	for (; i < htab->n_buckets; i++) {
191	head = select_bucket(htab, i);
192
193	/* pick first element in the bucket */
194	next_l = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
195	struct htab_elem, hash_node);
196	if (next_l) {
197	/* if it's not empty, just return it */
198	memcpy(next_key, next_l->key, key_size);
199	return 0;
200	}
201	}
202
203	/* itereated over all buckets and all elements */
204	return -ENOENT;
205	}
206
207	/* Called from syscall or from eBPF program */
208	static int htab_map_update_elem(struct bpf_map map, void key, void *value,
209	u64 map_flags)
210	{
211	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
212	struct htab_elem l_new, l_old;
213	struct hlist_head *head;
214	unsigned long flags;
215	u32 key_size;
216	int ret;
217
218	if (map_flags > BPF_EXIST)
219	/* unknown flags */
220	return -EINVAL;
221
222	WARN_ON_ONCE(!rcu_read_lock_held());
223
224	/* allocate new element outside of lock */
225	l_new = kmalloc(htab->elem_size, GFP_ATOMIC);
226	if (!l_new)
227	return -ENOMEM;
228
229	key_size = map->key_size;
230
231	memcpy(l_new->key, key, key_size);
232	memcpy(l_new->key + round_up(key_size, 8), value, map->value_size);
233
234	l_new->hash = htab_map_hash(l_new->key, key_size);
235
236	/* bpf_map_update_elem() can be called in_irq() */
ac00881f	237	raw_spin_lock_irqsave(&htab->lock, flags);
0f8e4bd8 AS	238
	239	head = select_bucket(htab, l_new->hash);
	240
	241	l_old = lookup_elem_raw(head, l_new->hash, key, key_size);
	242
	243	if (!l_old && unlikely(htab->count >= map->max_entries)) {
	244	/* if elem with this 'key' doesn't exist and we've reached
	245	* max_entries limit, fail insertion of new elem
	246	*/
	247	ret = -E2BIG;
	248	goto err;
	249	}
	250
	251	if (l_old && map_flags == BPF_NOEXIST) {
	252	/* elem already exists */
	253	ret = -EEXIST;
	254	goto err;
	255	}
	256
	257	if (!l_old && map_flags == BPF_EXIST) {
	258	/* elem doesn't exist, cannot update it */
	259	ret = -ENOENT;
	260	goto err;
	261	}
	262
	263	/* add new element to the head of the list, so that concurrent
	264	* search will find it before old elem
	265	*/
	266	hlist_add_head_rcu(&l_new->hash_node, head);
	267	if (l_old) {
	268	hlist_del_rcu(&l_old->hash_node);
	269	kfree_rcu(l_old, rcu);
	270	} else {
	271	htab->count++;
	272	}
ac00881f	273	raw_spin_unlock_irqrestore(&htab->lock, flags);
0f8e4bd8 AS	274
	275	return 0;
	276	err:
ac00881f	277	raw_spin_unlock_irqrestore(&htab->lock, flags);
0f8e4bd8 AS	278	kfree(l_new);
	279	return ret;
	280	}
	281
	282	/* Called from syscall or from eBPF program */
	283	static int htab_map_delete_elem(struct bpf_map map, void key)
	284	{
	285	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	286	struct hlist_head *head;
	287	struct htab_elem *l;
	288	unsigned long flags;
	289	u32 hash, key_size;
	290	int ret = -ENOENT;
	291
	292	WARN_ON_ONCE(!rcu_read_lock_held());
	293
	294	key_size = map->key_size;
	295
	296	hash = htab_map_hash(key, key_size);
	297
ac00881f	298	raw_spin_lock_irqsave(&htab->lock, flags);
0f8e4bd8 AS	299
	300	head = select_bucket(htab, hash);
	301
	302	l = lookup_elem_raw(head, hash, key, key_size);
	303
	304	if (l) {
	305	hlist_del_rcu(&l->hash_node);
	306	htab->count--;
	307	kfree_rcu(l, rcu);
	308	ret = 0;
	309	}
	310
ac00881f	311	raw_spin_unlock_irqrestore(&htab->lock, flags);
0f8e4bd8 AS	312	return ret;
	313	}
	314
	315	static void delete_all_elements(struct bpf_htab *htab)
	316	{
	317	int i;
	318
	319	for (i = 0; i < htab->n_buckets; i++) {
	320	struct hlist_head *head = select_bucket(htab, i);
	321	struct hlist_node *n;
	322	struct htab_elem *l;
	323
	324	hlist_for_each_entry_safe(l, n, head, hash_node) {
	325	hlist_del_rcu(&l->hash_node);
	326	htab->count--;
	327	kfree(l);
	328	}
	329	}
	330	}
	331
	332	/* Called when map->refcnt goes to zero, either from workqueue or from syscall */
	333	static void htab_map_free(struct bpf_map *map)
	334	{
	335	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
	336
	337	/* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
	338	* so the programs (can be more than one that used this map) were
	339	* disconnected from events. Wait for outstanding critical sections in
	340	* these programs to complete
	341	*/
	342	synchronize_rcu();
	343
	344	/* some of kfree_rcu() callbacks for elements of this map may not have
	345	* executed. It's ok. Proceed to free residual elements and map itself
	346	*/
	347	delete_all_elements(htab);
	348	kvfree(htab->buckets);
	349	kfree(htab);
	350	}
	351
a2c83fff	352	static const struct bpf_map_ops htab_ops = {
0f8e4bd8 AS	353	.map_alloc = htab_map_alloc,
	354	.map_free = htab_map_free,
	355	.map_get_next_key = htab_map_get_next_key,
	356	.map_lookup_elem = htab_map_lookup_elem,
	357	.map_update_elem = htab_map_update_elem,
	358	.map_delete_elem = htab_map_delete_elem,
	359	};
	360
a2c83fff	361	static struct bpf_map_type_list htab_type __read_mostly = {
0f8e4bd8 AS	362	.ops = &htab_ops,
	363	.type = BPF_MAP_TYPE_HASH,
	364	};
	365
	366	static int __init register_htab_map(void)
	367	{
a2c83fff	368	bpf_register_map_type(&htab_type);
0f8e4bd8 AS	369	return 0;
	370	}
	371	late_initcall(register_htab_map);