xref: /openbmc/linux/kernel/bpf/hashtab.c (revision 6d8e62c3)
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;
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 */
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;
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 
93 free_htab:
94 	kfree(htab);
95 	return ERR_PTR(err);
96 }
97 
98 static inline u32 htab_map_hash(const void *key, u32 key_len)
99 {
100 	return jhash(key, key_len, 0);
101 }
102 
103 static inline struct hlist_head *select_bucket(struct bpf_htab *htab, u32 hash)
104 {
105 	return &htab->buckets[hash & (htab->n_buckets - 1)];
106 }
107 
108 static 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 */
121 static 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 */
146 static 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 
184 find_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 */
204 static 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;
272 err:
273 	spin_unlock_irqrestore(&htab->lock, flags);
274 	kfree(l_new);
275 	return ret;
276 }
277 
278 /* Called from syscall or from eBPF program */
279 static 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 
311 static 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 */
329 static 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 
348 static 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 
357 static struct bpf_map_type_list tl = {
358 	.ops = &htab_ops,
359 	.type = BPF_MAP_TYPE_HASH,
360 };
361 
362 static int __init register_htab_map(void)
363 {
364 	bpf_register_map_type(&tl);
365 	return 0;
366 }
367 late_initcall(register_htab_map);
368