xref: /openbmc/linux/net/core/sock_map.c (revision f519cd13)
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 			lock_sock(sk);
246 			sock_map_unref(sk, psk);
247 			release_sock(sk);
248 		}
249 	}
250 	raw_spin_unlock_bh(&stab->lock);
251 	rcu_read_unlock();
252 
253 	synchronize_rcu();
254 
255 	bpf_map_area_free(stab->sks);
256 	kfree(stab);
257 }
258 
259 static void sock_map_release_progs(struct bpf_map *map)
260 {
261 	psock_progs_drop(&container_of(map, struct bpf_stab, map)->progs);
262 }
263 
264 static struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
265 {
266 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
267 
268 	WARN_ON_ONCE(!rcu_read_lock_held());
269 
270 	if (unlikely(key >= map->max_entries))
271 		return NULL;
272 	return READ_ONCE(stab->sks[key]);
273 }
274 
275 static void *sock_map_lookup(struct bpf_map *map, void *key)
276 {
277 	return ERR_PTR(-EOPNOTSUPP);
278 }
279 
280 static int __sock_map_delete(struct bpf_stab *stab, struct sock *sk_test,
281 			     struct sock **psk)
282 {
283 	struct sock *sk;
284 	int err = 0;
285 
286 	raw_spin_lock_bh(&stab->lock);
287 	sk = *psk;
288 	if (!sk_test || sk_test == sk)
289 		sk = xchg(psk, NULL);
290 
291 	if (likely(sk))
292 		sock_map_unref(sk, psk);
293 	else
294 		err = -EINVAL;
295 
296 	raw_spin_unlock_bh(&stab->lock);
297 	return err;
298 }
299 
300 static void sock_map_delete_from_link(struct bpf_map *map, struct sock *sk,
301 				      void *link_raw)
302 {
303 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
304 
305 	__sock_map_delete(stab, sk, link_raw);
306 }
307 
308 static int sock_map_delete_elem(struct bpf_map *map, void *key)
309 {
310 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
311 	u32 i = *(u32 *)key;
312 	struct sock **psk;
313 
314 	if (unlikely(i >= map->max_entries))
315 		return -EINVAL;
316 
317 	psk = &stab->sks[i];
318 	return __sock_map_delete(stab, NULL, psk);
319 }
320 
321 static int sock_map_get_next_key(struct bpf_map *map, void *key, void *next)
322 {
323 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
324 	u32 i = key ? *(u32 *)key : U32_MAX;
325 	u32 *key_next = next;
326 
327 	if (i == stab->map.max_entries - 1)
328 		return -ENOENT;
329 	if (i >= stab->map.max_entries)
330 		*key_next = 0;
331 	else
332 		*key_next = i + 1;
333 	return 0;
334 }
335 
336 static int sock_map_update_common(struct bpf_map *map, u32 idx,
337 				  struct sock *sk, u64 flags)
338 {
339 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
340 	struct inet_connection_sock *icsk = inet_csk(sk);
341 	struct sk_psock_link *link;
342 	struct sk_psock *psock;
343 	struct sock *osk;
344 	int ret;
345 
346 	WARN_ON_ONCE(!rcu_read_lock_held());
347 	if (unlikely(flags > BPF_EXIST))
348 		return -EINVAL;
349 	if (unlikely(idx >= map->max_entries))
350 		return -E2BIG;
351 	if (unlikely(rcu_access_pointer(icsk->icsk_ulp_data)))
352 		return -EINVAL;
353 
354 	link = sk_psock_init_link();
355 	if (!link)
356 		return -ENOMEM;
357 
358 	ret = sock_map_link(map, &stab->progs, sk);
359 	if (ret < 0)
360 		goto out_free;
361 
362 	psock = sk_psock(sk);
363 	WARN_ON_ONCE(!psock);
364 
365 	raw_spin_lock_bh(&stab->lock);
366 	osk = stab->sks[idx];
367 	if (osk && flags == BPF_NOEXIST) {
368 		ret = -EEXIST;
369 		goto out_unlock;
370 	} else if (!osk && flags == BPF_EXIST) {
371 		ret = -ENOENT;
372 		goto out_unlock;
373 	}
374 
375 	sock_map_add_link(psock, link, map, &stab->sks[idx]);
376 	stab->sks[idx] = sk;
377 	if (osk)
378 		sock_map_unref(osk, &stab->sks[idx]);
379 	raw_spin_unlock_bh(&stab->lock);
380 	return 0;
381 out_unlock:
382 	raw_spin_unlock_bh(&stab->lock);
383 	if (psock)
384 		sk_psock_put(sk, psock);
385 out_free:
386 	sk_psock_free_link(link);
387 	return ret;
388 }
389 
390 static bool sock_map_op_okay(const struct bpf_sock_ops_kern *ops)
391 {
392 	return ops->op == BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB ||
393 	       ops->op == BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB;
394 }
395 
396 static bool sock_map_sk_is_suitable(const struct sock *sk)
397 {
398 	return sk->sk_type == SOCK_STREAM &&
399 	       sk->sk_protocol == IPPROTO_TCP;
400 }
401 
402 static int sock_map_update_elem(struct bpf_map *map, void *key,
403 				void *value, u64 flags)
404 {
405 	u32 ufd = *(u32 *)value;
406 	u32 idx = *(u32 *)key;
407 	struct socket *sock;
408 	struct sock *sk;
409 	int ret;
410 
411 	sock = sockfd_lookup(ufd, &ret);
412 	if (!sock)
413 		return ret;
414 	sk = sock->sk;
415 	if (!sk) {
416 		ret = -EINVAL;
417 		goto out;
418 	}
419 	if (!sock_map_sk_is_suitable(sk) ||
420 	    sk->sk_state != TCP_ESTABLISHED) {
421 		ret = -EOPNOTSUPP;
422 		goto out;
423 	}
424 
425 	sock_map_sk_acquire(sk);
426 	ret = sock_map_update_common(map, idx, sk, flags);
427 	sock_map_sk_release(sk);
428 out:
429 	fput(sock->file);
430 	return ret;
431 }
432 
433 BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, sops,
434 	   struct bpf_map *, map, void *, key, u64, flags)
435 {
436 	WARN_ON_ONCE(!rcu_read_lock_held());
437 
438 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
439 		   sock_map_op_okay(sops)))
440 		return sock_map_update_common(map, *(u32 *)key, sops->sk,
441 					      flags);
442 	return -EOPNOTSUPP;
443 }
444 
445 const struct bpf_func_proto bpf_sock_map_update_proto = {
446 	.func		= bpf_sock_map_update,
447 	.gpl_only	= false,
448 	.pkt_access	= true,
449 	.ret_type	= RET_INTEGER,
450 	.arg1_type	= ARG_PTR_TO_CTX,
451 	.arg2_type	= ARG_CONST_MAP_PTR,
452 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
453 	.arg4_type	= ARG_ANYTHING,
454 };
455 
456 BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb,
457 	   struct bpf_map *, map, u32, key, u64, flags)
458 {
459 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
460 
461 	if (unlikely(flags & ~(BPF_F_INGRESS)))
462 		return SK_DROP;
463 	tcb->bpf.flags = flags;
464 	tcb->bpf.sk_redir = __sock_map_lookup_elem(map, key);
465 	if (!tcb->bpf.sk_redir)
466 		return SK_DROP;
467 	return SK_PASS;
468 }
469 
470 const struct bpf_func_proto bpf_sk_redirect_map_proto = {
471 	.func           = bpf_sk_redirect_map,
472 	.gpl_only       = false,
473 	.ret_type       = RET_INTEGER,
474 	.arg1_type	= ARG_PTR_TO_CTX,
475 	.arg2_type      = ARG_CONST_MAP_PTR,
476 	.arg3_type      = ARG_ANYTHING,
477 	.arg4_type      = ARG_ANYTHING,
478 };
479 
480 BPF_CALL_4(bpf_msg_redirect_map, struct sk_msg *, msg,
481 	   struct bpf_map *, map, u32, key, u64, flags)
482 {
483 	if (unlikely(flags & ~(BPF_F_INGRESS)))
484 		return SK_DROP;
485 	msg->flags = flags;
486 	msg->sk_redir = __sock_map_lookup_elem(map, key);
487 	if (!msg->sk_redir)
488 		return SK_DROP;
489 	return SK_PASS;
490 }
491 
492 const struct bpf_func_proto bpf_msg_redirect_map_proto = {
493 	.func           = bpf_msg_redirect_map,
494 	.gpl_only       = false,
495 	.ret_type       = RET_INTEGER,
496 	.arg1_type	= ARG_PTR_TO_CTX,
497 	.arg2_type      = ARG_CONST_MAP_PTR,
498 	.arg3_type      = ARG_ANYTHING,
499 	.arg4_type      = ARG_ANYTHING,
500 };
501 
502 const struct bpf_map_ops sock_map_ops = {
503 	.map_alloc		= sock_map_alloc,
504 	.map_free		= sock_map_free,
505 	.map_get_next_key	= sock_map_get_next_key,
506 	.map_update_elem	= sock_map_update_elem,
507 	.map_delete_elem	= sock_map_delete_elem,
508 	.map_lookup_elem	= sock_map_lookup,
509 	.map_release_uref	= sock_map_release_progs,
510 	.map_check_btf		= map_check_no_btf,
511 };
512 
513 struct bpf_htab_elem {
514 	struct rcu_head rcu;
515 	u32 hash;
516 	struct sock *sk;
517 	struct hlist_node node;
518 	u8 key[0];
519 };
520 
521 struct bpf_htab_bucket {
522 	struct hlist_head head;
523 	raw_spinlock_t lock;
524 };
525 
526 struct bpf_htab {
527 	struct bpf_map map;
528 	struct bpf_htab_bucket *buckets;
529 	u32 buckets_num;
530 	u32 elem_size;
531 	struct sk_psock_progs progs;
532 	atomic_t count;
533 };
534 
535 static inline u32 sock_hash_bucket_hash(const void *key, u32 len)
536 {
537 	return jhash(key, len, 0);
538 }
539 
540 static struct bpf_htab_bucket *sock_hash_select_bucket(struct bpf_htab *htab,
541 						       u32 hash)
542 {
543 	return &htab->buckets[hash & (htab->buckets_num - 1)];
544 }
545 
546 static struct bpf_htab_elem *
547 sock_hash_lookup_elem_raw(struct hlist_head *head, u32 hash, void *key,
548 			  u32 key_size)
549 {
550 	struct bpf_htab_elem *elem;
551 
552 	hlist_for_each_entry_rcu(elem, head, node) {
553 		if (elem->hash == hash &&
554 		    !memcmp(&elem->key, key, key_size))
555 			return elem;
556 	}
557 
558 	return NULL;
559 }
560 
561 static struct sock *__sock_hash_lookup_elem(struct bpf_map *map, void *key)
562 {
563 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
564 	u32 key_size = map->key_size, hash;
565 	struct bpf_htab_bucket *bucket;
566 	struct bpf_htab_elem *elem;
567 
568 	WARN_ON_ONCE(!rcu_read_lock_held());
569 
570 	hash = sock_hash_bucket_hash(key, key_size);
571 	bucket = sock_hash_select_bucket(htab, hash);
572 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
573 
574 	return elem ? elem->sk : NULL;
575 }
576 
577 static void sock_hash_free_elem(struct bpf_htab *htab,
578 				struct bpf_htab_elem *elem)
579 {
580 	atomic_dec(&htab->count);
581 	kfree_rcu(elem, rcu);
582 }
583 
584 static void sock_hash_delete_from_link(struct bpf_map *map, struct sock *sk,
585 				       void *link_raw)
586 {
587 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
588 	struct bpf_htab_elem *elem_probe, *elem = link_raw;
589 	struct bpf_htab_bucket *bucket;
590 
591 	WARN_ON_ONCE(!rcu_read_lock_held());
592 	bucket = sock_hash_select_bucket(htab, elem->hash);
593 
594 	/* elem may be deleted in parallel from the map, but access here
595 	 * is okay since it's going away only after RCU grace period.
596 	 * However, we need to check whether it's still present.
597 	 */
598 	raw_spin_lock_bh(&bucket->lock);
599 	elem_probe = sock_hash_lookup_elem_raw(&bucket->head, elem->hash,
600 					       elem->key, map->key_size);
601 	if (elem_probe && elem_probe == elem) {
602 		hlist_del_rcu(&elem->node);
603 		sock_map_unref(elem->sk, elem);
604 		sock_hash_free_elem(htab, elem);
605 	}
606 	raw_spin_unlock_bh(&bucket->lock);
607 }
608 
609 static int sock_hash_delete_elem(struct bpf_map *map, void *key)
610 {
611 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
612 	u32 hash, key_size = map->key_size;
613 	struct bpf_htab_bucket *bucket;
614 	struct bpf_htab_elem *elem;
615 	int ret = -ENOENT;
616 
617 	hash = sock_hash_bucket_hash(key, key_size);
618 	bucket = sock_hash_select_bucket(htab, hash);
619 
620 	raw_spin_lock_bh(&bucket->lock);
621 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
622 	if (elem) {
623 		hlist_del_rcu(&elem->node);
624 		sock_map_unref(elem->sk, elem);
625 		sock_hash_free_elem(htab, elem);
626 		ret = 0;
627 	}
628 	raw_spin_unlock_bh(&bucket->lock);
629 	return ret;
630 }
631 
632 static struct bpf_htab_elem *sock_hash_alloc_elem(struct bpf_htab *htab,
633 						  void *key, u32 key_size,
634 						  u32 hash, struct sock *sk,
635 						  struct bpf_htab_elem *old)
636 {
637 	struct bpf_htab_elem *new;
638 
639 	if (atomic_inc_return(&htab->count) > htab->map.max_entries) {
640 		if (!old) {
641 			atomic_dec(&htab->count);
642 			return ERR_PTR(-E2BIG);
643 		}
644 	}
645 
646 	new = kmalloc_node(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN,
647 			   htab->map.numa_node);
648 	if (!new) {
649 		atomic_dec(&htab->count);
650 		return ERR_PTR(-ENOMEM);
651 	}
652 	memcpy(new->key, key, key_size);
653 	new->sk = sk;
654 	new->hash = hash;
655 	return new;
656 }
657 
658 static int sock_hash_update_common(struct bpf_map *map, void *key,
659 				   struct sock *sk, u64 flags)
660 {
661 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
662 	struct inet_connection_sock *icsk = inet_csk(sk);
663 	u32 key_size = map->key_size, hash;
664 	struct bpf_htab_elem *elem, *elem_new;
665 	struct bpf_htab_bucket *bucket;
666 	struct sk_psock_link *link;
667 	struct sk_psock *psock;
668 	int ret;
669 
670 	WARN_ON_ONCE(!rcu_read_lock_held());
671 	if (unlikely(flags > BPF_EXIST))
672 		return -EINVAL;
673 	if (unlikely(icsk->icsk_ulp_data))
674 		return -EINVAL;
675 
676 	link = sk_psock_init_link();
677 	if (!link)
678 		return -ENOMEM;
679 
680 	ret = sock_map_link(map, &htab->progs, sk);
681 	if (ret < 0)
682 		goto out_free;
683 
684 	psock = sk_psock(sk);
685 	WARN_ON_ONCE(!psock);
686 
687 	hash = sock_hash_bucket_hash(key, key_size);
688 	bucket = sock_hash_select_bucket(htab, hash);
689 
690 	raw_spin_lock_bh(&bucket->lock);
691 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
692 	if (elem && flags == BPF_NOEXIST) {
693 		ret = -EEXIST;
694 		goto out_unlock;
695 	} else if (!elem && flags == BPF_EXIST) {
696 		ret = -ENOENT;
697 		goto out_unlock;
698 	}
699 
700 	elem_new = sock_hash_alloc_elem(htab, key, key_size, hash, sk, elem);
701 	if (IS_ERR(elem_new)) {
702 		ret = PTR_ERR(elem_new);
703 		goto out_unlock;
704 	}
705 
706 	sock_map_add_link(psock, link, map, elem_new);
707 	/* Add new element to the head of the list, so that
708 	 * concurrent search will find it before old elem.
709 	 */
710 	hlist_add_head_rcu(&elem_new->node, &bucket->head);
711 	if (elem) {
712 		hlist_del_rcu(&elem->node);
713 		sock_map_unref(elem->sk, elem);
714 		sock_hash_free_elem(htab, elem);
715 	}
716 	raw_spin_unlock_bh(&bucket->lock);
717 	return 0;
718 out_unlock:
719 	raw_spin_unlock_bh(&bucket->lock);
720 	sk_psock_put(sk, psock);
721 out_free:
722 	sk_psock_free_link(link);
723 	return ret;
724 }
725 
726 static int sock_hash_update_elem(struct bpf_map *map, void *key,
727 				 void *value, u64 flags)
728 {
729 	u32 ufd = *(u32 *)value;
730 	struct socket *sock;
731 	struct sock *sk;
732 	int ret;
733 
734 	sock = sockfd_lookup(ufd, &ret);
735 	if (!sock)
736 		return ret;
737 	sk = sock->sk;
738 	if (!sk) {
739 		ret = -EINVAL;
740 		goto out;
741 	}
742 	if (!sock_map_sk_is_suitable(sk) ||
743 	    sk->sk_state != TCP_ESTABLISHED) {
744 		ret = -EOPNOTSUPP;
745 		goto out;
746 	}
747 
748 	sock_map_sk_acquire(sk);
749 	ret = sock_hash_update_common(map, key, sk, flags);
750 	sock_map_sk_release(sk);
751 out:
752 	fput(sock->file);
753 	return ret;
754 }
755 
756 static int sock_hash_get_next_key(struct bpf_map *map, void *key,
757 				  void *key_next)
758 {
759 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
760 	struct bpf_htab_elem *elem, *elem_next;
761 	u32 hash, key_size = map->key_size;
762 	struct hlist_head *head;
763 	int i = 0;
764 
765 	if (!key)
766 		goto find_first_elem;
767 	hash = sock_hash_bucket_hash(key, key_size);
768 	head = &sock_hash_select_bucket(htab, hash)->head;
769 	elem = sock_hash_lookup_elem_raw(head, hash, key, key_size);
770 	if (!elem)
771 		goto find_first_elem;
772 
773 	elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&elem->node)),
774 				     struct bpf_htab_elem, node);
775 	if (elem_next) {
776 		memcpy(key_next, elem_next->key, key_size);
777 		return 0;
778 	}
779 
780 	i = hash & (htab->buckets_num - 1);
781 	i++;
782 find_first_elem:
783 	for (; i < htab->buckets_num; i++) {
784 		head = &sock_hash_select_bucket(htab, i)->head;
785 		elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
786 					     struct bpf_htab_elem, node);
787 		if (elem_next) {
788 			memcpy(key_next, elem_next->key, key_size);
789 			return 0;
790 		}
791 	}
792 
793 	return -ENOENT;
794 }
795 
796 static struct bpf_map *sock_hash_alloc(union bpf_attr *attr)
797 {
798 	struct bpf_htab *htab;
799 	int i, err;
800 	u64 cost;
801 
802 	if (!capable(CAP_NET_ADMIN))
803 		return ERR_PTR(-EPERM);
804 	if (attr->max_entries == 0 ||
805 	    attr->key_size    == 0 ||
806 	    attr->value_size  != 4 ||
807 	    attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
808 		return ERR_PTR(-EINVAL);
809 	if (attr->key_size > MAX_BPF_STACK)
810 		return ERR_PTR(-E2BIG);
811 
812 	htab = kzalloc(sizeof(*htab), GFP_USER);
813 	if (!htab)
814 		return ERR_PTR(-ENOMEM);
815 
816 	bpf_map_init_from_attr(&htab->map, attr);
817 
818 	htab->buckets_num = roundup_pow_of_two(htab->map.max_entries);
819 	htab->elem_size = sizeof(struct bpf_htab_elem) +
820 			  round_up(htab->map.key_size, 8);
821 	if (htab->buckets_num == 0 ||
822 	    htab->buckets_num > U32_MAX / sizeof(struct bpf_htab_bucket)) {
823 		err = -EINVAL;
824 		goto free_htab;
825 	}
826 
827 	cost = (u64) htab->buckets_num * sizeof(struct bpf_htab_bucket) +
828 	       (u64) htab->elem_size * htab->map.max_entries;
829 	if (cost >= U32_MAX - PAGE_SIZE) {
830 		err = -EINVAL;
831 		goto free_htab;
832 	}
833 
834 	htab->buckets = bpf_map_area_alloc(htab->buckets_num *
835 					   sizeof(struct bpf_htab_bucket),
836 					   htab->map.numa_node);
837 	if (!htab->buckets) {
838 		err = -ENOMEM;
839 		goto free_htab;
840 	}
841 
842 	for (i = 0; i < htab->buckets_num; i++) {
843 		INIT_HLIST_HEAD(&htab->buckets[i].head);
844 		raw_spin_lock_init(&htab->buckets[i].lock);
845 	}
846 
847 	return &htab->map;
848 free_htab:
849 	kfree(htab);
850 	return ERR_PTR(err);
851 }
852 
853 static void sock_hash_free(struct bpf_map *map)
854 {
855 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
856 	struct bpf_htab_bucket *bucket;
857 	struct bpf_htab_elem *elem;
858 	struct hlist_node *node;
859 	int i;
860 
861 	synchronize_rcu();
862 	rcu_read_lock();
863 	for (i = 0; i < htab->buckets_num; i++) {
864 		bucket = sock_hash_select_bucket(htab, i);
865 		raw_spin_lock_bh(&bucket->lock);
866 		hlist_for_each_entry_safe(elem, node, &bucket->head, node) {
867 			hlist_del_rcu(&elem->node);
868 			lock_sock(elem->sk);
869 			sock_map_unref(elem->sk, elem);
870 			release_sock(elem->sk);
871 		}
872 		raw_spin_unlock_bh(&bucket->lock);
873 	}
874 	rcu_read_unlock();
875 
876 	bpf_map_area_free(htab->buckets);
877 	kfree(htab);
878 }
879 
880 static void sock_hash_release_progs(struct bpf_map *map)
881 {
882 	psock_progs_drop(&container_of(map, struct bpf_htab, map)->progs);
883 }
884 
885 BPF_CALL_4(bpf_sock_hash_update, struct bpf_sock_ops_kern *, sops,
886 	   struct bpf_map *, map, void *, key, u64, flags)
887 {
888 	WARN_ON_ONCE(!rcu_read_lock_held());
889 
890 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
891 		   sock_map_op_okay(sops)))
892 		return sock_hash_update_common(map, key, sops->sk, flags);
893 	return -EOPNOTSUPP;
894 }
895 
896 const struct bpf_func_proto bpf_sock_hash_update_proto = {
897 	.func		= bpf_sock_hash_update,
898 	.gpl_only	= false,
899 	.pkt_access	= true,
900 	.ret_type	= RET_INTEGER,
901 	.arg1_type	= ARG_PTR_TO_CTX,
902 	.arg2_type	= ARG_CONST_MAP_PTR,
903 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
904 	.arg4_type	= ARG_ANYTHING,
905 };
906 
907 BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb,
908 	   struct bpf_map *, map, void *, key, u64, flags)
909 {
910 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
911 
912 	if (unlikely(flags & ~(BPF_F_INGRESS)))
913 		return SK_DROP;
914 	tcb->bpf.flags = flags;
915 	tcb->bpf.sk_redir = __sock_hash_lookup_elem(map, key);
916 	if (!tcb->bpf.sk_redir)
917 		return SK_DROP;
918 	return SK_PASS;
919 }
920 
921 const struct bpf_func_proto bpf_sk_redirect_hash_proto = {
922 	.func           = bpf_sk_redirect_hash,
923 	.gpl_only       = false,
924 	.ret_type       = RET_INTEGER,
925 	.arg1_type	= ARG_PTR_TO_CTX,
926 	.arg2_type      = ARG_CONST_MAP_PTR,
927 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
928 	.arg4_type      = ARG_ANYTHING,
929 };
930 
931 BPF_CALL_4(bpf_msg_redirect_hash, struct sk_msg *, msg,
932 	   struct bpf_map *, map, void *, key, u64, flags)
933 {
934 	if (unlikely(flags & ~(BPF_F_INGRESS)))
935 		return SK_DROP;
936 	msg->flags = flags;
937 	msg->sk_redir = __sock_hash_lookup_elem(map, key);
938 	if (!msg->sk_redir)
939 		return SK_DROP;
940 	return SK_PASS;
941 }
942 
943 const struct bpf_func_proto bpf_msg_redirect_hash_proto = {
944 	.func           = bpf_msg_redirect_hash,
945 	.gpl_only       = false,
946 	.ret_type       = RET_INTEGER,
947 	.arg1_type	= ARG_PTR_TO_CTX,
948 	.arg2_type      = ARG_CONST_MAP_PTR,
949 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
950 	.arg4_type      = ARG_ANYTHING,
951 };
952 
953 const struct bpf_map_ops sock_hash_ops = {
954 	.map_alloc		= sock_hash_alloc,
955 	.map_free		= sock_hash_free,
956 	.map_get_next_key	= sock_hash_get_next_key,
957 	.map_update_elem	= sock_hash_update_elem,
958 	.map_delete_elem	= sock_hash_delete_elem,
959 	.map_lookup_elem	= sock_map_lookup,
960 	.map_release_uref	= sock_hash_release_progs,
961 	.map_check_btf		= map_check_no_btf,
962 };
963 
964 static struct sk_psock_progs *sock_map_progs(struct bpf_map *map)
965 {
966 	switch (map->map_type) {
967 	case BPF_MAP_TYPE_SOCKMAP:
968 		return &container_of(map, struct bpf_stab, map)->progs;
969 	case BPF_MAP_TYPE_SOCKHASH:
970 		return &container_of(map, struct bpf_htab, map)->progs;
971 	default:
972 		break;
973 	}
974 
975 	return NULL;
976 }
977 
978 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
979 			 u32 which)
980 {
981 	struct sk_psock_progs *progs = sock_map_progs(map);
982 
983 	if (!progs)
984 		return -EOPNOTSUPP;
985 
986 	switch (which) {
987 	case BPF_SK_MSG_VERDICT:
988 		psock_set_prog(&progs->msg_parser, prog);
989 		break;
990 	case BPF_SK_SKB_STREAM_PARSER:
991 		psock_set_prog(&progs->skb_parser, prog);
992 		break;
993 	case BPF_SK_SKB_STREAM_VERDICT:
994 		psock_set_prog(&progs->skb_verdict, prog);
995 		break;
996 	default:
997 		return -EOPNOTSUPP;
998 	}
999 
1000 	return 0;
1001 }
1002 
1003 void sk_psock_unlink(struct sock *sk, struct sk_psock_link *link)
1004 {
1005 	switch (link->map->map_type) {
1006 	case BPF_MAP_TYPE_SOCKMAP:
1007 		return sock_map_delete_from_link(link->map, sk,
1008 						 link->link_raw);
1009 	case BPF_MAP_TYPE_SOCKHASH:
1010 		return sock_hash_delete_from_link(link->map, sk,
1011 						  link->link_raw);
1012 	default:
1013 		break;
1014 	}
1015 }
1016