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
17struct bpf_htab {
18	struct bpf_map map;
19	struct hlist_head *buckets;
20	spinlock_t lock;
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 */
27struct 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 */
35static 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;
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
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
85	spin_lock_init(&htab->lock);
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;
91	return &htab->map;
92
93free_htab:
94	kfree(htab);
95	return ERR_PTR(err);
96}
97
98static inline u32 htab_map_hash(const void *key, u32 key_len)
99{
100	return jhash(key, key_len, 0);
101}
102
103static inline struct hlist_head *select_bucket(struct bpf_htab *htab, u32 hash)
104{
105	return &htab->buckets[hash & (htab->n_buckets - 1)];
106}
107
108static struct htab_elem *lookup_elem_raw(struct hlist_head *head, u32 hash,
109					 void *key, u32 key_size)
110{
111	struct htab_elem *l;
112
113	hlist_for_each_entry_rcu(l, head, hash_node)
114		if (l->hash == hash && !memcmp(&l->key, key, key_size))
115			return l;
116
117	return NULL;
118}
119
120/* Called from syscall or from eBPF program */
121static void *htab_map_lookup_elem(struct bpf_map *map, void *key)
122{
123	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
124	struct hlist_head *head;
125	struct htab_elem *l;
126	u32 hash, key_size;
127
128	/* Must be called with rcu_read_lock. */
129	WARN_ON_ONCE(!rcu_read_lock_held());
130
131	key_size = map->key_size;
132
133	hash = htab_map_hash(key, key_size);
134
135	head = select_bucket(htab, hash);
136
137	l = lookup_elem_raw(head, hash, key, key_size);
138
139	if (l)
140		return l->key + round_up(map->key_size, 8);
141
142	return NULL;
143}
144
145/* Called from syscall */
146static int htab_map_get_next_key(struct bpf_map *map, void *key, void *next_key)
147{
148	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
149	struct hlist_head *head;
150	struct htab_elem *l, *next_l;
151	u32 hash, key_size;
152	int i;
153
154	WARN_ON_ONCE(!rcu_read_lock_held());
155
156	key_size = map->key_size;
157
158	hash = htab_map_hash(key, key_size);
159
160	head = select_bucket(htab, hash);
161
162	/* lookup the key */
163	l = lookup_elem_raw(head, hash, key, key_size);
164
165	if (!l) {
166		i = 0;
167		goto find_first_elem;
168	}
169
170	/* key was found, get next key in the same bucket */
171	next_l = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&l->hash_node)),
172				  struct htab_elem, hash_node);
173
174	if (next_l) {
175		/* if next elem in this hash list is non-zero, just return it */
176		memcpy(next_key, next_l->key, key_size);
177		return 0;
178	}
179
180	/* no more elements in this hash list, go to the next bucket */
181	i = hash & (htab->n_buckets - 1);
182	i++;
183
184find_first_elem:
185	/* iterate over buckets */
186	for (; i < htab->n_buckets; i++) {
187		head = select_bucket(htab, i);
188
189		/* pick first element in the bucket */
190		next_l = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
191					  struct htab_elem, hash_node);
192		if (next_l) {
193			/* if it's not empty, just return it */
194			memcpy(next_key, next_l->key, key_size);
195			return 0;
196		}
197	}
198
199	/* itereated over all buckets and all elements */
200	return -ENOENT;
201}
202
203/* Called from syscall or from eBPF program */
204static int htab_map_update_elem(struct bpf_map *map, void *key, void *value,
205				u64 map_flags)
206{
207	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
208	struct htab_elem *l_new, *l_old;
209	struct hlist_head *head;
210	unsigned long flags;
211	u32 key_size;
212	int ret;
213
214	if (map_flags > BPF_EXIST)
215		/* unknown flags */
216		return -EINVAL;
217
218	WARN_ON_ONCE(!rcu_read_lock_held());
219
220	/* allocate new element outside of lock */
221	l_new = kmalloc(htab->elem_size, GFP_ATOMIC);
222	if (!l_new)
223		return -ENOMEM;
224
225	key_size = map->key_size;
226
227	memcpy(l_new->key, key, key_size);
228	memcpy(l_new->key + round_up(key_size, 8), value, map->value_size);
229
230	l_new->hash = htab_map_hash(l_new->key, key_size);
231
232	/* bpf_map_update_elem() can be called in_irq() */
233	spin_lock_irqsave(&htab->lock, flags);
234
235	head = select_bucket(htab, l_new->hash);
236
237	l_old = lookup_elem_raw(head, l_new->hash, key, key_size);
238
239	if (!l_old && unlikely(htab->count >= map->max_entries)) {
240		/* if elem with this 'key' doesn't exist and we've reached
241		 * max_entries limit, fail insertion of new elem
242		 */
243		ret = -E2BIG;
244		goto err;
245	}
246
247	if (l_old && map_flags == BPF_NOEXIST) {
248		/* elem already exists */
249		ret = -EEXIST;
250		goto err;
251	}
252
253	if (!l_old && map_flags == BPF_EXIST) {
254		/* elem doesn't exist, cannot update it */
255		ret = -ENOENT;
256		goto err;
257	}
258
259	/* add new element to the head of the list, so that concurrent
260	 * search will find it before old elem
261	 */
262	hlist_add_head_rcu(&l_new->hash_node, head);
263	if (l_old) {
264		hlist_del_rcu(&l_old->hash_node);
265		kfree_rcu(l_old, rcu);
266	} else {
267		htab->count++;
268	}
269	spin_unlock_irqrestore(&htab->lock, flags);
270
271	return 0;
272err:
273	spin_unlock_irqrestore(&htab->lock, flags);
274	kfree(l_new);
275	return ret;
276}
277
278/* Called from syscall or from eBPF program */
279static int htab_map_delete_elem(struct bpf_map *map, void *key)
280{
281	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
282	struct hlist_head *head;
283	struct htab_elem *l;
284	unsigned long flags;
285	u32 hash, key_size;
286	int ret = -ENOENT;
287
288	WARN_ON_ONCE(!rcu_read_lock_held());
289
290	key_size = map->key_size;
291
292	hash = htab_map_hash(key, key_size);
293
294	spin_lock_irqsave(&htab->lock, flags);
295
296	head = select_bucket(htab, hash);
297
298	l = lookup_elem_raw(head, hash, key, key_size);
299
300	if (l) {
301		hlist_del_rcu(&l->hash_node);
302		htab->count--;
303		kfree_rcu(l, rcu);
304		ret = 0;
305	}
306
307	spin_unlock_irqrestore(&htab->lock, flags);
308	return ret;
309}
310
311static void delete_all_elements(struct bpf_htab *htab)
312{
313	int i;
314
315	for (i = 0; i < htab->n_buckets; i++) {
316		struct hlist_head *head = select_bucket(htab, i);
317		struct hlist_node *n;
318		struct htab_elem *l;
319
320		hlist_for_each_entry_safe(l, n, head, hash_node) {
321			hlist_del_rcu(&l->hash_node);
322			htab->count--;
323			kfree(l);
324		}
325	}
326}
327
328/* Called when map->refcnt goes to zero, either from workqueue or from syscall */
329static void htab_map_free(struct bpf_map *map)
330{
331	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
332
333	/* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
334	 * so the programs (can be more than one that used this map) were
335	 * disconnected from events. Wait for outstanding critical sections in
336	 * these programs to complete
337	 */
338	synchronize_rcu();
339
340	/* some of kfree_rcu() callbacks for elements of this map may not have
341	 * executed. It's ok. Proceed to free residual elements and map itself
342	 */
343	delete_all_elements(htab);
344	kvfree(htab->buckets);
345	kfree(htab);
346}
347
348static const struct bpf_map_ops htab_ops = {
349	.map_alloc = htab_map_alloc,
350	.map_free = htab_map_free,
351	.map_get_next_key = htab_map_get_next_key,
352	.map_lookup_elem = htab_map_lookup_elem,
353	.map_update_elem = htab_map_update_elem,
354	.map_delete_elem = htab_map_delete_elem,
355};
356
357static struct bpf_map_type_list htab_type __read_mostly = {
358	.ops = &htab_ops,
359	.type = BPF_MAP_TYPE_HASH,
360};
361
362static int __init register_htab_map(void)
363{
364	bpf_register_map_type(&htab_type);
365	return 0;
366}
367late_initcall(register_htab_map);
368