xref: /openbmc/linux/net/core/sock_map.c (revision cf9441ad)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2017 - 2018 Covalent IO, Inc. http://covalent.io */
3 
4 #include <linux/bpf.h>
5 #include <linux/filter.h>
6 #include <linux/errno.h>
7 #include <linux/file.h>
8 #include <linux/net.h>
9 #include <linux/workqueue.h>
10 #include <linux/skmsg.h>
11 #include <linux/list.h>
12 #include <linux/jhash.h>
13 
14 struct bpf_stab {
15 	struct bpf_map map;
16 	struct sock **sks;
17 	struct sk_psock_progs progs;
18 	raw_spinlock_t lock;
19 };
20 
21 #define SOCK_CREATE_FLAG_MASK				\
22 	(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY)
23 
24 static struct bpf_map *sock_map_alloc(union bpf_attr *attr)
25 {
26 	struct bpf_stab *stab;
27 	u64 cost;
28 	int err;
29 
30 	if (!capable(CAP_NET_ADMIN))
31 		return ERR_PTR(-EPERM);
32 	if (attr->max_entries == 0 ||
33 	    attr->key_size    != 4 ||
34 	    attr->value_size  != 4 ||
35 	    attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
36 		return ERR_PTR(-EINVAL);
37 
38 	stab = kzalloc(sizeof(*stab), GFP_USER);
39 	if (!stab)
40 		return ERR_PTR(-ENOMEM);
41 
42 	bpf_map_init_from_attr(&stab->map, attr);
43 	raw_spin_lock_init(&stab->lock);
44 
45 	/* Make sure page count doesn't overflow. */
46 	cost = (u64) stab->map.max_entries * sizeof(struct sock *);
47 	err = bpf_map_charge_init(&stab->map.memory, cost);
48 	if (err)
49 		goto free_stab;
50 
51 	stab->sks = bpf_map_area_alloc(stab->map.max_entries *
52 				       sizeof(struct sock *),
53 				       stab->map.numa_node);
54 	if (stab->sks)
55 		return &stab->map;
56 	err = -ENOMEM;
57 	bpf_map_charge_finish(&stab->map.memory);
58 free_stab:
59 	kfree(stab);
60 	return ERR_PTR(err);
61 }
62 
63 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog)
64 {
65 	u32 ufd = attr->target_fd;
66 	struct bpf_map *map;
67 	struct fd f;
68 	int ret;
69 
70 	f = fdget(ufd);
71 	map = __bpf_map_get(f);
72 	if (IS_ERR(map))
73 		return PTR_ERR(map);
74 	ret = sock_map_prog_update(map, prog, attr->attach_type);
75 	fdput(f);
76 	return ret;
77 }
78 
79 static void sock_map_sk_acquire(struct sock *sk)
80 	__acquires(&sk->sk_lock.slock)
81 {
82 	lock_sock(sk);
83 	preempt_disable();
84 	rcu_read_lock();
85 }
86 
87 static void sock_map_sk_release(struct sock *sk)
88 	__releases(&sk->sk_lock.slock)
89 {
90 	rcu_read_unlock();
91 	preempt_enable();
92 	release_sock(sk);
93 }
94 
95 static void sock_map_add_link(struct sk_psock *psock,
96 			      struct sk_psock_link *link,
97 			      struct bpf_map *map, void *link_raw)
98 {
99 	link->link_raw = link_raw;
100 	link->map = map;
101 	spin_lock_bh(&psock->link_lock);
102 	list_add_tail(&link->list, &psock->link);
103 	spin_unlock_bh(&psock->link_lock);
104 }
105 
106 static void sock_map_del_link(struct sock *sk,
107 			      struct sk_psock *psock, void *link_raw)
108 {
109 	struct sk_psock_link *link, *tmp;
110 	bool strp_stop = false;
111 
112 	spin_lock_bh(&psock->link_lock);
113 	list_for_each_entry_safe(link, tmp, &psock->link, list) {
114 		if (link->link_raw == link_raw) {
115 			struct bpf_map *map = link->map;
116 			struct bpf_stab *stab = container_of(map, struct bpf_stab,
117 							     map);
118 			if (psock->parser.enabled && stab->progs.skb_parser)
119 				strp_stop = true;
120 			list_del(&link->list);
121 			sk_psock_free_link(link);
122 		}
123 	}
124 	spin_unlock_bh(&psock->link_lock);
125 	if (strp_stop) {
126 		write_lock_bh(&sk->sk_callback_lock);
127 		sk_psock_stop_strp(sk, psock);
128 		write_unlock_bh(&sk->sk_callback_lock);
129 	}
130 }
131 
132 static void sock_map_unref(struct sock *sk, void *link_raw)
133 {
134 	struct sk_psock *psock = sk_psock(sk);
135 
136 	if (likely(psock)) {
137 		sock_map_del_link(sk, psock, link_raw);
138 		sk_psock_put(sk, psock);
139 	}
140 }
141 
142 static int sock_map_link(struct bpf_map *map, struct sk_psock_progs *progs,
143 			 struct sock *sk)
144 {
145 	struct bpf_prog *msg_parser, *skb_parser, *skb_verdict;
146 	bool skb_progs, sk_psock_is_new = false;
147 	struct sk_psock *psock;
148 	int ret;
149 
150 	skb_verdict = READ_ONCE(progs->skb_verdict);
151 	skb_parser = READ_ONCE(progs->skb_parser);
152 	skb_progs = skb_parser && skb_verdict;
153 	if (skb_progs) {
154 		skb_verdict = bpf_prog_inc_not_zero(skb_verdict);
155 		if (IS_ERR(skb_verdict))
156 			return PTR_ERR(skb_verdict);
157 		skb_parser = bpf_prog_inc_not_zero(skb_parser);
158 		if (IS_ERR(skb_parser)) {
159 			bpf_prog_put(skb_verdict);
160 			return PTR_ERR(skb_parser);
161 		}
162 	}
163 
164 	msg_parser = READ_ONCE(progs->msg_parser);
165 	if (msg_parser) {
166 		msg_parser = bpf_prog_inc_not_zero(msg_parser);
167 		if (IS_ERR(msg_parser)) {
168 			ret = PTR_ERR(msg_parser);
169 			goto out;
170 		}
171 	}
172 
173 	psock = sk_psock_get_checked(sk);
174 	if (IS_ERR(psock)) {
175 		ret = PTR_ERR(psock);
176 		goto out_progs;
177 	}
178 
179 	if (psock) {
180 		if ((msg_parser && READ_ONCE(psock->progs.msg_parser)) ||
181 		    (skb_progs  && READ_ONCE(psock->progs.skb_parser))) {
182 			sk_psock_put(sk, psock);
183 			ret = -EBUSY;
184 			goto out_progs;
185 		}
186 	} else {
187 		psock = sk_psock_init(sk, map->numa_node);
188 		if (!psock) {
189 			ret = -ENOMEM;
190 			goto out_progs;
191 		}
192 		sk_psock_is_new = true;
193 	}
194 
195 	if (msg_parser)
196 		psock_set_prog(&psock->progs.msg_parser, msg_parser);
197 	if (sk_psock_is_new) {
198 		ret = tcp_bpf_init(sk);
199 		if (ret < 0)
200 			goto out_drop;
201 	} else {
202 		tcp_bpf_reinit(sk);
203 	}
204 
205 	write_lock_bh(&sk->sk_callback_lock);
206 	if (skb_progs && !psock->parser.enabled) {
207 		ret = sk_psock_init_strp(sk, psock);
208 		if (ret) {
209 			write_unlock_bh(&sk->sk_callback_lock);
210 			goto out_drop;
211 		}
212 		psock_set_prog(&psock->progs.skb_verdict, skb_verdict);
213 		psock_set_prog(&psock->progs.skb_parser, skb_parser);
214 		sk_psock_start_strp(sk, psock);
215 	}
216 	write_unlock_bh(&sk->sk_callback_lock);
217 	return 0;
218 out_drop:
219 	sk_psock_put(sk, psock);
220 out_progs:
221 	if (msg_parser)
222 		bpf_prog_put(msg_parser);
223 out:
224 	if (skb_progs) {
225 		bpf_prog_put(skb_verdict);
226 		bpf_prog_put(skb_parser);
227 	}
228 	return ret;
229 }
230 
231 static void sock_map_free(struct bpf_map *map)
232 {
233 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
234 	int i;
235 
236 	synchronize_rcu();
237 	rcu_read_lock();
238 	raw_spin_lock_bh(&stab->lock);
239 	for (i = 0; i < stab->map.max_entries; i++) {
240 		struct sock **psk = &stab->sks[i];
241 		struct sock *sk;
242 
243 		sk = xchg(psk, NULL);
244 		if (sk)
245 			sock_map_unref(sk, psk);
246 	}
247 	raw_spin_unlock_bh(&stab->lock);
248 	rcu_read_unlock();
249 
250 	bpf_map_area_free(stab->sks);
251 	kfree(stab);
252 }
253 
254 static void sock_map_release_progs(struct bpf_map *map)
255 {
256 	psock_progs_drop(&container_of(map, struct bpf_stab, map)->progs);
257 }
258 
259 static struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
260 {
261 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
262 
263 	WARN_ON_ONCE(!rcu_read_lock_held());
264 
265 	if (unlikely(key >= map->max_entries))
266 		return NULL;
267 	return READ_ONCE(stab->sks[key]);
268 }
269 
270 static void *sock_map_lookup(struct bpf_map *map, void *key)
271 {
272 	return ERR_PTR(-EOPNOTSUPP);
273 }
274 
275 static int __sock_map_delete(struct bpf_stab *stab, struct sock *sk_test,
276 			     struct sock **psk)
277 {
278 	struct sock *sk;
279 
280 	raw_spin_lock_bh(&stab->lock);
281 	sk = *psk;
282 	if (!sk_test || sk_test == sk)
283 		*psk = NULL;
284 	raw_spin_unlock_bh(&stab->lock);
285 	if (unlikely(!sk))
286 		return -EINVAL;
287 	sock_map_unref(sk, psk);
288 	return 0;
289 }
290 
291 static void sock_map_delete_from_link(struct bpf_map *map, struct sock *sk,
292 				      void *link_raw)
293 {
294 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
295 
296 	__sock_map_delete(stab, sk, link_raw);
297 }
298 
299 static int sock_map_delete_elem(struct bpf_map *map, void *key)
300 {
301 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
302 	u32 i = *(u32 *)key;
303 	struct sock **psk;
304 
305 	if (unlikely(i >= map->max_entries))
306 		return -EINVAL;
307 
308 	psk = &stab->sks[i];
309 	return __sock_map_delete(stab, NULL, psk);
310 }
311 
312 static int sock_map_get_next_key(struct bpf_map *map, void *key, void *next)
313 {
314 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
315 	u32 i = key ? *(u32 *)key : U32_MAX;
316 	u32 *key_next = next;
317 
318 	if (i == stab->map.max_entries - 1)
319 		return -ENOENT;
320 	if (i >= stab->map.max_entries)
321 		*key_next = 0;
322 	else
323 		*key_next = i + 1;
324 	return 0;
325 }
326 
327 static int sock_map_update_common(struct bpf_map *map, u32 idx,
328 				  struct sock *sk, u64 flags)
329 {
330 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
331 	struct sk_psock_link *link;
332 	struct sk_psock *psock;
333 	struct sock *osk;
334 	int ret;
335 
336 	WARN_ON_ONCE(!rcu_read_lock_held());
337 	if (unlikely(flags > BPF_EXIST))
338 		return -EINVAL;
339 	if (unlikely(idx >= map->max_entries))
340 		return -E2BIG;
341 
342 	link = sk_psock_init_link();
343 	if (!link)
344 		return -ENOMEM;
345 
346 	ret = sock_map_link(map, &stab->progs, sk);
347 	if (ret < 0)
348 		goto out_free;
349 
350 	psock = sk_psock(sk);
351 	WARN_ON_ONCE(!psock);
352 
353 	raw_spin_lock_bh(&stab->lock);
354 	osk = stab->sks[idx];
355 	if (osk && flags == BPF_NOEXIST) {
356 		ret = -EEXIST;
357 		goto out_unlock;
358 	} else if (!osk && flags == BPF_EXIST) {
359 		ret = -ENOENT;
360 		goto out_unlock;
361 	}
362 
363 	sock_map_add_link(psock, link, map, &stab->sks[idx]);
364 	stab->sks[idx] = sk;
365 	if (osk)
366 		sock_map_unref(osk, &stab->sks[idx]);
367 	raw_spin_unlock_bh(&stab->lock);
368 	return 0;
369 out_unlock:
370 	raw_spin_unlock_bh(&stab->lock);
371 	if (psock)
372 		sk_psock_put(sk, psock);
373 out_free:
374 	sk_psock_free_link(link);
375 	return ret;
376 }
377 
378 static bool sock_map_op_okay(const struct bpf_sock_ops_kern *ops)
379 {
380 	return ops->op == BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB ||
381 	       ops->op == BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB;
382 }
383 
384 static bool sock_map_sk_is_suitable(const struct sock *sk)
385 {
386 	return sk->sk_type == SOCK_STREAM &&
387 	       sk->sk_protocol == IPPROTO_TCP;
388 }
389 
390 static int sock_map_update_elem(struct bpf_map *map, void *key,
391 				void *value, u64 flags)
392 {
393 	u32 ufd = *(u32 *)value;
394 	u32 idx = *(u32 *)key;
395 	struct socket *sock;
396 	struct sock *sk;
397 	int ret;
398 
399 	sock = sockfd_lookup(ufd, &ret);
400 	if (!sock)
401 		return ret;
402 	sk = sock->sk;
403 	if (!sk) {
404 		ret = -EINVAL;
405 		goto out;
406 	}
407 	if (!sock_map_sk_is_suitable(sk) ||
408 	    sk->sk_state != TCP_ESTABLISHED) {
409 		ret = -EOPNOTSUPP;
410 		goto out;
411 	}
412 
413 	sock_map_sk_acquire(sk);
414 	ret = sock_map_update_common(map, idx, sk, flags);
415 	sock_map_sk_release(sk);
416 out:
417 	fput(sock->file);
418 	return ret;
419 }
420 
421 BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, sops,
422 	   struct bpf_map *, map, void *, key, u64, flags)
423 {
424 	WARN_ON_ONCE(!rcu_read_lock_held());
425 
426 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
427 		   sock_map_op_okay(sops)))
428 		return sock_map_update_common(map, *(u32 *)key, sops->sk,
429 					      flags);
430 	return -EOPNOTSUPP;
431 }
432 
433 const struct bpf_func_proto bpf_sock_map_update_proto = {
434 	.func		= bpf_sock_map_update,
435 	.gpl_only	= false,
436 	.pkt_access	= true,
437 	.ret_type	= RET_INTEGER,
438 	.arg1_type	= ARG_PTR_TO_CTX,
439 	.arg2_type	= ARG_CONST_MAP_PTR,
440 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
441 	.arg4_type	= ARG_ANYTHING,
442 };
443 
444 BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb,
445 	   struct bpf_map *, map, u32, key, u64, flags)
446 {
447 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
448 
449 	if (unlikely(flags & ~(BPF_F_INGRESS)))
450 		return SK_DROP;
451 	tcb->bpf.flags = flags;
452 	tcb->bpf.sk_redir = __sock_map_lookup_elem(map, key);
453 	if (!tcb->bpf.sk_redir)
454 		return SK_DROP;
455 	return SK_PASS;
456 }
457 
458 const struct bpf_func_proto bpf_sk_redirect_map_proto = {
459 	.func           = bpf_sk_redirect_map,
460 	.gpl_only       = false,
461 	.ret_type       = RET_INTEGER,
462 	.arg1_type	= ARG_PTR_TO_CTX,
463 	.arg2_type      = ARG_CONST_MAP_PTR,
464 	.arg3_type      = ARG_ANYTHING,
465 	.arg4_type      = ARG_ANYTHING,
466 };
467 
468 BPF_CALL_4(bpf_msg_redirect_map, struct sk_msg *, msg,
469 	   struct bpf_map *, map, u32, key, u64, flags)
470 {
471 	if (unlikely(flags & ~(BPF_F_INGRESS)))
472 		return SK_DROP;
473 	msg->flags = flags;
474 	msg->sk_redir = __sock_map_lookup_elem(map, key);
475 	if (!msg->sk_redir)
476 		return SK_DROP;
477 	return SK_PASS;
478 }
479 
480 const struct bpf_func_proto bpf_msg_redirect_map_proto = {
481 	.func           = bpf_msg_redirect_map,
482 	.gpl_only       = false,
483 	.ret_type       = RET_INTEGER,
484 	.arg1_type	= ARG_PTR_TO_CTX,
485 	.arg2_type      = ARG_CONST_MAP_PTR,
486 	.arg3_type      = ARG_ANYTHING,
487 	.arg4_type      = ARG_ANYTHING,
488 };
489 
490 const struct bpf_map_ops sock_map_ops = {
491 	.map_alloc		= sock_map_alloc,
492 	.map_free		= sock_map_free,
493 	.map_get_next_key	= sock_map_get_next_key,
494 	.map_update_elem	= sock_map_update_elem,
495 	.map_delete_elem	= sock_map_delete_elem,
496 	.map_lookup_elem	= sock_map_lookup,
497 	.map_release_uref	= sock_map_release_progs,
498 	.map_check_btf		= map_check_no_btf,
499 };
500 
501 struct bpf_htab_elem {
502 	struct rcu_head rcu;
503 	u32 hash;
504 	struct sock *sk;
505 	struct hlist_node node;
506 	u8 key[0];
507 };
508 
509 struct bpf_htab_bucket {
510 	struct hlist_head head;
511 	raw_spinlock_t lock;
512 };
513 
514 struct bpf_htab {
515 	struct bpf_map map;
516 	struct bpf_htab_bucket *buckets;
517 	u32 buckets_num;
518 	u32 elem_size;
519 	struct sk_psock_progs progs;
520 	atomic_t count;
521 };
522 
523 static inline u32 sock_hash_bucket_hash(const void *key, u32 len)
524 {
525 	return jhash(key, len, 0);
526 }
527 
528 static struct bpf_htab_bucket *sock_hash_select_bucket(struct bpf_htab *htab,
529 						       u32 hash)
530 {
531 	return &htab->buckets[hash & (htab->buckets_num - 1)];
532 }
533 
534 static struct bpf_htab_elem *
535 sock_hash_lookup_elem_raw(struct hlist_head *head, u32 hash, void *key,
536 			  u32 key_size)
537 {
538 	struct bpf_htab_elem *elem;
539 
540 	hlist_for_each_entry_rcu(elem, head, node) {
541 		if (elem->hash == hash &&
542 		    !memcmp(&elem->key, key, key_size))
543 			return elem;
544 	}
545 
546 	return NULL;
547 }
548 
549 static struct sock *__sock_hash_lookup_elem(struct bpf_map *map, void *key)
550 {
551 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
552 	u32 key_size = map->key_size, hash;
553 	struct bpf_htab_bucket *bucket;
554 	struct bpf_htab_elem *elem;
555 
556 	WARN_ON_ONCE(!rcu_read_lock_held());
557 
558 	hash = sock_hash_bucket_hash(key, key_size);
559 	bucket = sock_hash_select_bucket(htab, hash);
560 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
561 
562 	return elem ? elem->sk : NULL;
563 }
564 
565 static void sock_hash_free_elem(struct bpf_htab *htab,
566 				struct bpf_htab_elem *elem)
567 {
568 	atomic_dec(&htab->count);
569 	kfree_rcu(elem, rcu);
570 }
571 
572 static void sock_hash_delete_from_link(struct bpf_map *map, struct sock *sk,
573 				       void *link_raw)
574 {
575 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
576 	struct bpf_htab_elem *elem_probe, *elem = link_raw;
577 	struct bpf_htab_bucket *bucket;
578 
579 	WARN_ON_ONCE(!rcu_read_lock_held());
580 	bucket = sock_hash_select_bucket(htab, elem->hash);
581 
582 	/* elem may be deleted in parallel from the map, but access here
583 	 * is okay since it's going away only after RCU grace period.
584 	 * However, we need to check whether it's still present.
585 	 */
586 	raw_spin_lock_bh(&bucket->lock);
587 	elem_probe = sock_hash_lookup_elem_raw(&bucket->head, elem->hash,
588 					       elem->key, map->key_size);
589 	if (elem_probe && elem_probe == elem) {
590 		hlist_del_rcu(&elem->node);
591 		sock_map_unref(elem->sk, elem);
592 		sock_hash_free_elem(htab, elem);
593 	}
594 	raw_spin_unlock_bh(&bucket->lock);
595 }
596 
597 static int sock_hash_delete_elem(struct bpf_map *map, void *key)
598 {
599 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
600 	u32 hash, key_size = map->key_size;
601 	struct bpf_htab_bucket *bucket;
602 	struct bpf_htab_elem *elem;
603 	int ret = -ENOENT;
604 
605 	hash = sock_hash_bucket_hash(key, key_size);
606 	bucket = sock_hash_select_bucket(htab, hash);
607 
608 	raw_spin_lock_bh(&bucket->lock);
609 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
610 	if (elem) {
611 		hlist_del_rcu(&elem->node);
612 		sock_map_unref(elem->sk, elem);
613 		sock_hash_free_elem(htab, elem);
614 		ret = 0;
615 	}
616 	raw_spin_unlock_bh(&bucket->lock);
617 	return ret;
618 }
619 
620 static struct bpf_htab_elem *sock_hash_alloc_elem(struct bpf_htab *htab,
621 						  void *key, u32 key_size,
622 						  u32 hash, struct sock *sk,
623 						  struct bpf_htab_elem *old)
624 {
625 	struct bpf_htab_elem *new;
626 
627 	if (atomic_inc_return(&htab->count) > htab->map.max_entries) {
628 		if (!old) {
629 			atomic_dec(&htab->count);
630 			return ERR_PTR(-E2BIG);
631 		}
632 	}
633 
634 	new = kmalloc_node(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN,
635 			   htab->map.numa_node);
636 	if (!new) {
637 		atomic_dec(&htab->count);
638 		return ERR_PTR(-ENOMEM);
639 	}
640 	memcpy(new->key, key, key_size);
641 	new->sk = sk;
642 	new->hash = hash;
643 	return new;
644 }
645 
646 static int sock_hash_update_common(struct bpf_map *map, void *key,
647 				   struct sock *sk, u64 flags)
648 {
649 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
650 	u32 key_size = map->key_size, hash;
651 	struct bpf_htab_elem *elem, *elem_new;
652 	struct bpf_htab_bucket *bucket;
653 	struct sk_psock_link *link;
654 	struct sk_psock *psock;
655 	int ret;
656 
657 	WARN_ON_ONCE(!rcu_read_lock_held());
658 	if (unlikely(flags > BPF_EXIST))
659 		return -EINVAL;
660 
661 	link = sk_psock_init_link();
662 	if (!link)
663 		return -ENOMEM;
664 
665 	ret = sock_map_link(map, &htab->progs, sk);
666 	if (ret < 0)
667 		goto out_free;
668 
669 	psock = sk_psock(sk);
670 	WARN_ON_ONCE(!psock);
671 
672 	hash = sock_hash_bucket_hash(key, key_size);
673 	bucket = sock_hash_select_bucket(htab, hash);
674 
675 	raw_spin_lock_bh(&bucket->lock);
676 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
677 	if (elem && flags == BPF_NOEXIST) {
678 		ret = -EEXIST;
679 		goto out_unlock;
680 	} else if (!elem && flags == BPF_EXIST) {
681 		ret = -ENOENT;
682 		goto out_unlock;
683 	}
684 
685 	elem_new = sock_hash_alloc_elem(htab, key, key_size, hash, sk, elem);
686 	if (IS_ERR(elem_new)) {
687 		ret = PTR_ERR(elem_new);
688 		goto out_unlock;
689 	}
690 
691 	sock_map_add_link(psock, link, map, elem_new);
692 	/* Add new element to the head of the list, so that
693 	 * concurrent search will find it before old elem.
694 	 */
695 	hlist_add_head_rcu(&elem_new->node, &bucket->head);
696 	if (elem) {
697 		hlist_del_rcu(&elem->node);
698 		sock_map_unref(elem->sk, elem);
699 		sock_hash_free_elem(htab, elem);
700 	}
701 	raw_spin_unlock_bh(&bucket->lock);
702 	return 0;
703 out_unlock:
704 	raw_spin_unlock_bh(&bucket->lock);
705 	sk_psock_put(sk, psock);
706 out_free:
707 	sk_psock_free_link(link);
708 	return ret;
709 }
710 
711 static int sock_hash_update_elem(struct bpf_map *map, void *key,
712 				 void *value, u64 flags)
713 {
714 	u32 ufd = *(u32 *)value;
715 	struct socket *sock;
716 	struct sock *sk;
717 	int ret;
718 
719 	sock = sockfd_lookup(ufd, &ret);
720 	if (!sock)
721 		return ret;
722 	sk = sock->sk;
723 	if (!sk) {
724 		ret = -EINVAL;
725 		goto out;
726 	}
727 	if (!sock_map_sk_is_suitable(sk) ||
728 	    sk->sk_state != TCP_ESTABLISHED) {
729 		ret = -EOPNOTSUPP;
730 		goto out;
731 	}
732 
733 	sock_map_sk_acquire(sk);
734 	ret = sock_hash_update_common(map, key, sk, flags);
735 	sock_map_sk_release(sk);
736 out:
737 	fput(sock->file);
738 	return ret;
739 }
740 
741 static int sock_hash_get_next_key(struct bpf_map *map, void *key,
742 				  void *key_next)
743 {
744 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
745 	struct bpf_htab_elem *elem, *elem_next;
746 	u32 hash, key_size = map->key_size;
747 	struct hlist_head *head;
748 	int i = 0;
749 
750 	if (!key)
751 		goto find_first_elem;
752 	hash = sock_hash_bucket_hash(key, key_size);
753 	head = &sock_hash_select_bucket(htab, hash)->head;
754 	elem = sock_hash_lookup_elem_raw(head, hash, key, key_size);
755 	if (!elem)
756 		goto find_first_elem;
757 
758 	elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&elem->node)),
759 				     struct bpf_htab_elem, node);
760 	if (elem_next) {
761 		memcpy(key_next, elem_next->key, key_size);
762 		return 0;
763 	}
764 
765 	i = hash & (htab->buckets_num - 1);
766 	i++;
767 find_first_elem:
768 	for (; i < htab->buckets_num; i++) {
769 		head = &sock_hash_select_bucket(htab, i)->head;
770 		elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
771 					     struct bpf_htab_elem, node);
772 		if (elem_next) {
773 			memcpy(key_next, elem_next->key, key_size);
774 			return 0;
775 		}
776 	}
777 
778 	return -ENOENT;
779 }
780 
781 static struct bpf_map *sock_hash_alloc(union bpf_attr *attr)
782 {
783 	struct bpf_htab *htab;
784 	int i, err;
785 	u64 cost;
786 
787 	if (!capable(CAP_NET_ADMIN))
788 		return ERR_PTR(-EPERM);
789 	if (attr->max_entries == 0 ||
790 	    attr->key_size    == 0 ||
791 	    attr->value_size  != 4 ||
792 	    attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
793 		return ERR_PTR(-EINVAL);
794 	if (attr->key_size > MAX_BPF_STACK)
795 		return ERR_PTR(-E2BIG);
796 
797 	htab = kzalloc(sizeof(*htab), GFP_USER);
798 	if (!htab)
799 		return ERR_PTR(-ENOMEM);
800 
801 	bpf_map_init_from_attr(&htab->map, attr);
802 
803 	htab->buckets_num = roundup_pow_of_two(htab->map.max_entries);
804 	htab->elem_size = sizeof(struct bpf_htab_elem) +
805 			  round_up(htab->map.key_size, 8);
806 	if (htab->buckets_num == 0 ||
807 	    htab->buckets_num > U32_MAX / sizeof(struct bpf_htab_bucket)) {
808 		err = -EINVAL;
809 		goto free_htab;
810 	}
811 
812 	cost = (u64) htab->buckets_num * sizeof(struct bpf_htab_bucket) +
813 	       (u64) htab->elem_size * htab->map.max_entries;
814 	if (cost >= U32_MAX - PAGE_SIZE) {
815 		err = -EINVAL;
816 		goto free_htab;
817 	}
818 
819 	htab->buckets = bpf_map_area_alloc(htab->buckets_num *
820 					   sizeof(struct bpf_htab_bucket),
821 					   htab->map.numa_node);
822 	if (!htab->buckets) {
823 		err = -ENOMEM;
824 		goto free_htab;
825 	}
826 
827 	for (i = 0; i < htab->buckets_num; i++) {
828 		INIT_HLIST_HEAD(&htab->buckets[i].head);
829 		raw_spin_lock_init(&htab->buckets[i].lock);
830 	}
831 
832 	return &htab->map;
833 free_htab:
834 	kfree(htab);
835 	return ERR_PTR(err);
836 }
837 
838 static void sock_hash_free(struct bpf_map *map)
839 {
840 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
841 	struct bpf_htab_bucket *bucket;
842 	struct bpf_htab_elem *elem;
843 	struct hlist_node *node;
844 	int i;
845 
846 	synchronize_rcu();
847 	rcu_read_lock();
848 	for (i = 0; i < htab->buckets_num; i++) {
849 		bucket = sock_hash_select_bucket(htab, i);
850 		raw_spin_lock_bh(&bucket->lock);
851 		hlist_for_each_entry_safe(elem, node, &bucket->head, node) {
852 			hlist_del_rcu(&elem->node);
853 			sock_map_unref(elem->sk, elem);
854 		}
855 		raw_spin_unlock_bh(&bucket->lock);
856 	}
857 	rcu_read_unlock();
858 
859 	bpf_map_area_free(htab->buckets);
860 	kfree(htab);
861 }
862 
863 static void sock_hash_release_progs(struct bpf_map *map)
864 {
865 	psock_progs_drop(&container_of(map, struct bpf_htab, map)->progs);
866 }
867 
868 BPF_CALL_4(bpf_sock_hash_update, struct bpf_sock_ops_kern *, sops,
869 	   struct bpf_map *, map, void *, key, u64, flags)
870 {
871 	WARN_ON_ONCE(!rcu_read_lock_held());
872 
873 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
874 		   sock_map_op_okay(sops)))
875 		return sock_hash_update_common(map, key, sops->sk, flags);
876 	return -EOPNOTSUPP;
877 }
878 
879 const struct bpf_func_proto bpf_sock_hash_update_proto = {
880 	.func		= bpf_sock_hash_update,
881 	.gpl_only	= false,
882 	.pkt_access	= true,
883 	.ret_type	= RET_INTEGER,
884 	.arg1_type	= ARG_PTR_TO_CTX,
885 	.arg2_type	= ARG_CONST_MAP_PTR,
886 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
887 	.arg4_type	= ARG_ANYTHING,
888 };
889 
890 BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb,
891 	   struct bpf_map *, map, void *, key, u64, flags)
892 {
893 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
894 
895 	if (unlikely(flags & ~(BPF_F_INGRESS)))
896 		return SK_DROP;
897 	tcb->bpf.flags = flags;
898 	tcb->bpf.sk_redir = __sock_hash_lookup_elem(map, key);
899 	if (!tcb->bpf.sk_redir)
900 		return SK_DROP;
901 	return SK_PASS;
902 }
903 
904 const struct bpf_func_proto bpf_sk_redirect_hash_proto = {
905 	.func           = bpf_sk_redirect_hash,
906 	.gpl_only       = false,
907 	.ret_type       = RET_INTEGER,
908 	.arg1_type	= ARG_PTR_TO_CTX,
909 	.arg2_type      = ARG_CONST_MAP_PTR,
910 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
911 	.arg4_type      = ARG_ANYTHING,
912 };
913 
914 BPF_CALL_4(bpf_msg_redirect_hash, struct sk_msg *, msg,
915 	   struct bpf_map *, map, void *, key, u64, flags)
916 {
917 	if (unlikely(flags & ~(BPF_F_INGRESS)))
918 		return SK_DROP;
919 	msg->flags = flags;
920 	msg->sk_redir = __sock_hash_lookup_elem(map, key);
921 	if (!msg->sk_redir)
922 		return SK_DROP;
923 	return SK_PASS;
924 }
925 
926 const struct bpf_func_proto bpf_msg_redirect_hash_proto = {
927 	.func           = bpf_msg_redirect_hash,
928 	.gpl_only       = false,
929 	.ret_type       = RET_INTEGER,
930 	.arg1_type	= ARG_PTR_TO_CTX,
931 	.arg2_type      = ARG_CONST_MAP_PTR,
932 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
933 	.arg4_type      = ARG_ANYTHING,
934 };
935 
936 const struct bpf_map_ops sock_hash_ops = {
937 	.map_alloc		= sock_hash_alloc,
938 	.map_free		= sock_hash_free,
939 	.map_get_next_key	= sock_hash_get_next_key,
940 	.map_update_elem	= sock_hash_update_elem,
941 	.map_delete_elem	= sock_hash_delete_elem,
942 	.map_lookup_elem	= sock_map_lookup,
943 	.map_release_uref	= sock_hash_release_progs,
944 	.map_check_btf		= map_check_no_btf,
945 };
946 
947 static struct sk_psock_progs *sock_map_progs(struct bpf_map *map)
948 {
949 	switch (map->map_type) {
950 	case BPF_MAP_TYPE_SOCKMAP:
951 		return &container_of(map, struct bpf_stab, map)->progs;
952 	case BPF_MAP_TYPE_SOCKHASH:
953 		return &container_of(map, struct bpf_htab, map)->progs;
954 	default:
955 		break;
956 	}
957 
958 	return NULL;
959 }
960 
961 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
962 			 u32 which)
963 {
964 	struct sk_psock_progs *progs = sock_map_progs(map);
965 
966 	if (!progs)
967 		return -EOPNOTSUPP;
968 
969 	switch (which) {
970 	case BPF_SK_MSG_VERDICT:
971 		psock_set_prog(&progs->msg_parser, prog);
972 		break;
973 	case BPF_SK_SKB_STREAM_PARSER:
974 		psock_set_prog(&progs->skb_parser, prog);
975 		break;
976 	case BPF_SK_SKB_STREAM_VERDICT:
977 		psock_set_prog(&progs->skb_verdict, prog);
978 		break;
979 	default:
980 		return -EOPNOTSUPP;
981 	}
982 
983 	return 0;
984 }
985 
986 void sk_psock_unlink(struct sock *sk, struct sk_psock_link *link)
987 {
988 	switch (link->map->map_type) {
989 	case BPF_MAP_TYPE_SOCKMAP:
990 		return sock_map_delete_from_link(link->map, sk,
991 						 link->link_raw);
992 	case BPF_MAP_TYPE_SOCKHASH:
993 		return sock_hash_delete_from_link(link->map, sk,
994 						  link->link_raw);
995 	default:
996 		break;
997 	}
998 }
999