xref: /openbmc/linux/net/core/sock_map.c (revision 7b73a9c8)
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 	synchronize_rcu();
251 
252 	bpf_map_area_free(stab->sks);
253 	kfree(stab);
254 }
255 
256 static void sock_map_release_progs(struct bpf_map *map)
257 {
258 	psock_progs_drop(&container_of(map, struct bpf_stab, map)->progs);
259 }
260 
261 static struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
262 {
263 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
264 
265 	WARN_ON_ONCE(!rcu_read_lock_held());
266 
267 	if (unlikely(key >= map->max_entries))
268 		return NULL;
269 	return READ_ONCE(stab->sks[key]);
270 }
271 
272 static void *sock_map_lookup(struct bpf_map *map, void *key)
273 {
274 	return ERR_PTR(-EOPNOTSUPP);
275 }
276 
277 static int __sock_map_delete(struct bpf_stab *stab, struct sock *sk_test,
278 			     struct sock **psk)
279 {
280 	struct sock *sk;
281 	int err = 0;
282 
283 	raw_spin_lock_bh(&stab->lock);
284 	sk = *psk;
285 	if (!sk_test || sk_test == sk)
286 		sk = xchg(psk, NULL);
287 
288 	if (likely(sk))
289 		sock_map_unref(sk, psk);
290 	else
291 		err = -EINVAL;
292 
293 	raw_spin_unlock_bh(&stab->lock);
294 	return err;
295 }
296 
297 static void sock_map_delete_from_link(struct bpf_map *map, struct sock *sk,
298 				      void *link_raw)
299 {
300 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
301 
302 	__sock_map_delete(stab, sk, link_raw);
303 }
304 
305 static int sock_map_delete_elem(struct bpf_map *map, void *key)
306 {
307 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
308 	u32 i = *(u32 *)key;
309 	struct sock **psk;
310 
311 	if (unlikely(i >= map->max_entries))
312 		return -EINVAL;
313 
314 	psk = &stab->sks[i];
315 	return __sock_map_delete(stab, NULL, psk);
316 }
317 
318 static int sock_map_get_next_key(struct bpf_map *map, void *key, void *next)
319 {
320 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
321 	u32 i = key ? *(u32 *)key : U32_MAX;
322 	u32 *key_next = next;
323 
324 	if (i == stab->map.max_entries - 1)
325 		return -ENOENT;
326 	if (i >= stab->map.max_entries)
327 		*key_next = 0;
328 	else
329 		*key_next = i + 1;
330 	return 0;
331 }
332 
333 static int sock_map_update_common(struct bpf_map *map, u32 idx,
334 				  struct sock *sk, u64 flags)
335 {
336 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
337 	struct inet_connection_sock *icsk = inet_csk(sk);
338 	struct sk_psock_link *link;
339 	struct sk_psock *psock;
340 	struct sock *osk;
341 	int ret;
342 
343 	WARN_ON_ONCE(!rcu_read_lock_held());
344 	if (unlikely(flags > BPF_EXIST))
345 		return -EINVAL;
346 	if (unlikely(idx >= map->max_entries))
347 		return -E2BIG;
348 	if (unlikely(rcu_access_pointer(icsk->icsk_ulp_data)))
349 		return -EINVAL;
350 
351 	link = sk_psock_init_link();
352 	if (!link)
353 		return -ENOMEM;
354 
355 	ret = sock_map_link(map, &stab->progs, sk);
356 	if (ret < 0)
357 		goto out_free;
358 
359 	psock = sk_psock(sk);
360 	WARN_ON_ONCE(!psock);
361 
362 	raw_spin_lock_bh(&stab->lock);
363 	osk = stab->sks[idx];
364 	if (osk && flags == BPF_NOEXIST) {
365 		ret = -EEXIST;
366 		goto out_unlock;
367 	} else if (!osk && flags == BPF_EXIST) {
368 		ret = -ENOENT;
369 		goto out_unlock;
370 	}
371 
372 	sock_map_add_link(psock, link, map, &stab->sks[idx]);
373 	stab->sks[idx] = sk;
374 	if (osk)
375 		sock_map_unref(osk, &stab->sks[idx]);
376 	raw_spin_unlock_bh(&stab->lock);
377 	return 0;
378 out_unlock:
379 	raw_spin_unlock_bh(&stab->lock);
380 	if (psock)
381 		sk_psock_put(sk, psock);
382 out_free:
383 	sk_psock_free_link(link);
384 	return ret;
385 }
386 
387 static bool sock_map_op_okay(const struct bpf_sock_ops_kern *ops)
388 {
389 	return ops->op == BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB ||
390 	       ops->op == BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB;
391 }
392 
393 static bool sock_map_sk_is_suitable(const struct sock *sk)
394 {
395 	return sk->sk_type == SOCK_STREAM &&
396 	       sk->sk_protocol == IPPROTO_TCP;
397 }
398 
399 static int sock_map_update_elem(struct bpf_map *map, void *key,
400 				void *value, u64 flags)
401 {
402 	u32 ufd = *(u32 *)value;
403 	u32 idx = *(u32 *)key;
404 	struct socket *sock;
405 	struct sock *sk;
406 	int ret;
407 
408 	sock = sockfd_lookup(ufd, &ret);
409 	if (!sock)
410 		return ret;
411 	sk = sock->sk;
412 	if (!sk) {
413 		ret = -EINVAL;
414 		goto out;
415 	}
416 	if (!sock_map_sk_is_suitable(sk) ||
417 	    sk->sk_state != TCP_ESTABLISHED) {
418 		ret = -EOPNOTSUPP;
419 		goto out;
420 	}
421 
422 	sock_map_sk_acquire(sk);
423 	ret = sock_map_update_common(map, idx, sk, flags);
424 	sock_map_sk_release(sk);
425 out:
426 	fput(sock->file);
427 	return ret;
428 }
429 
430 BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, sops,
431 	   struct bpf_map *, map, void *, key, u64, flags)
432 {
433 	WARN_ON_ONCE(!rcu_read_lock_held());
434 
435 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
436 		   sock_map_op_okay(sops)))
437 		return sock_map_update_common(map, *(u32 *)key, sops->sk,
438 					      flags);
439 	return -EOPNOTSUPP;
440 }
441 
442 const struct bpf_func_proto bpf_sock_map_update_proto = {
443 	.func		= bpf_sock_map_update,
444 	.gpl_only	= false,
445 	.pkt_access	= true,
446 	.ret_type	= RET_INTEGER,
447 	.arg1_type	= ARG_PTR_TO_CTX,
448 	.arg2_type	= ARG_CONST_MAP_PTR,
449 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
450 	.arg4_type	= ARG_ANYTHING,
451 };
452 
453 BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb,
454 	   struct bpf_map *, map, u32, key, u64, flags)
455 {
456 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
457 
458 	if (unlikely(flags & ~(BPF_F_INGRESS)))
459 		return SK_DROP;
460 	tcb->bpf.flags = flags;
461 	tcb->bpf.sk_redir = __sock_map_lookup_elem(map, key);
462 	if (!tcb->bpf.sk_redir)
463 		return SK_DROP;
464 	return SK_PASS;
465 }
466 
467 const struct bpf_func_proto bpf_sk_redirect_map_proto = {
468 	.func           = bpf_sk_redirect_map,
469 	.gpl_only       = false,
470 	.ret_type       = RET_INTEGER,
471 	.arg1_type	= ARG_PTR_TO_CTX,
472 	.arg2_type      = ARG_CONST_MAP_PTR,
473 	.arg3_type      = ARG_ANYTHING,
474 	.arg4_type      = ARG_ANYTHING,
475 };
476 
477 BPF_CALL_4(bpf_msg_redirect_map, struct sk_msg *, msg,
478 	   struct bpf_map *, map, u32, key, u64, flags)
479 {
480 	if (unlikely(flags & ~(BPF_F_INGRESS)))
481 		return SK_DROP;
482 	msg->flags = flags;
483 	msg->sk_redir = __sock_map_lookup_elem(map, key);
484 	if (!msg->sk_redir)
485 		return SK_DROP;
486 	return SK_PASS;
487 }
488 
489 const struct bpf_func_proto bpf_msg_redirect_map_proto = {
490 	.func           = bpf_msg_redirect_map,
491 	.gpl_only       = false,
492 	.ret_type       = RET_INTEGER,
493 	.arg1_type	= ARG_PTR_TO_CTX,
494 	.arg2_type      = ARG_CONST_MAP_PTR,
495 	.arg3_type      = ARG_ANYTHING,
496 	.arg4_type      = ARG_ANYTHING,
497 };
498 
499 const struct bpf_map_ops sock_map_ops = {
500 	.map_alloc		= sock_map_alloc,
501 	.map_free		= sock_map_free,
502 	.map_get_next_key	= sock_map_get_next_key,
503 	.map_update_elem	= sock_map_update_elem,
504 	.map_delete_elem	= sock_map_delete_elem,
505 	.map_lookup_elem	= sock_map_lookup,
506 	.map_release_uref	= sock_map_release_progs,
507 	.map_check_btf		= map_check_no_btf,
508 };
509 
510 struct bpf_htab_elem {
511 	struct rcu_head rcu;
512 	u32 hash;
513 	struct sock *sk;
514 	struct hlist_node node;
515 	u8 key[0];
516 };
517 
518 struct bpf_htab_bucket {
519 	struct hlist_head head;
520 	raw_spinlock_t lock;
521 };
522 
523 struct bpf_htab {
524 	struct bpf_map map;
525 	struct bpf_htab_bucket *buckets;
526 	u32 buckets_num;
527 	u32 elem_size;
528 	struct sk_psock_progs progs;
529 	atomic_t count;
530 };
531 
532 static inline u32 sock_hash_bucket_hash(const void *key, u32 len)
533 {
534 	return jhash(key, len, 0);
535 }
536 
537 static struct bpf_htab_bucket *sock_hash_select_bucket(struct bpf_htab *htab,
538 						       u32 hash)
539 {
540 	return &htab->buckets[hash & (htab->buckets_num - 1)];
541 }
542 
543 static struct bpf_htab_elem *
544 sock_hash_lookup_elem_raw(struct hlist_head *head, u32 hash, void *key,
545 			  u32 key_size)
546 {
547 	struct bpf_htab_elem *elem;
548 
549 	hlist_for_each_entry_rcu(elem, head, node) {
550 		if (elem->hash == hash &&
551 		    !memcmp(&elem->key, key, key_size))
552 			return elem;
553 	}
554 
555 	return NULL;
556 }
557 
558 static struct sock *__sock_hash_lookup_elem(struct bpf_map *map, void *key)
559 {
560 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
561 	u32 key_size = map->key_size, hash;
562 	struct bpf_htab_bucket *bucket;
563 	struct bpf_htab_elem *elem;
564 
565 	WARN_ON_ONCE(!rcu_read_lock_held());
566 
567 	hash = sock_hash_bucket_hash(key, key_size);
568 	bucket = sock_hash_select_bucket(htab, hash);
569 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
570 
571 	return elem ? elem->sk : NULL;
572 }
573 
574 static void sock_hash_free_elem(struct bpf_htab *htab,
575 				struct bpf_htab_elem *elem)
576 {
577 	atomic_dec(&htab->count);
578 	kfree_rcu(elem, rcu);
579 }
580 
581 static void sock_hash_delete_from_link(struct bpf_map *map, struct sock *sk,
582 				       void *link_raw)
583 {
584 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
585 	struct bpf_htab_elem *elem_probe, *elem = link_raw;
586 	struct bpf_htab_bucket *bucket;
587 
588 	WARN_ON_ONCE(!rcu_read_lock_held());
589 	bucket = sock_hash_select_bucket(htab, elem->hash);
590 
591 	/* elem may be deleted in parallel from the map, but access here
592 	 * is okay since it's going away only after RCU grace period.
593 	 * However, we need to check whether it's still present.
594 	 */
595 	raw_spin_lock_bh(&bucket->lock);
596 	elem_probe = sock_hash_lookup_elem_raw(&bucket->head, elem->hash,
597 					       elem->key, map->key_size);
598 	if (elem_probe && elem_probe == elem) {
599 		hlist_del_rcu(&elem->node);
600 		sock_map_unref(elem->sk, elem);
601 		sock_hash_free_elem(htab, elem);
602 	}
603 	raw_spin_unlock_bh(&bucket->lock);
604 }
605 
606 static int sock_hash_delete_elem(struct bpf_map *map, void *key)
607 {
608 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
609 	u32 hash, key_size = map->key_size;
610 	struct bpf_htab_bucket *bucket;
611 	struct bpf_htab_elem *elem;
612 	int ret = -ENOENT;
613 
614 	hash = sock_hash_bucket_hash(key, key_size);
615 	bucket = sock_hash_select_bucket(htab, hash);
616 
617 	raw_spin_lock_bh(&bucket->lock);
618 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
619 	if (elem) {
620 		hlist_del_rcu(&elem->node);
621 		sock_map_unref(elem->sk, elem);
622 		sock_hash_free_elem(htab, elem);
623 		ret = 0;
624 	}
625 	raw_spin_unlock_bh(&bucket->lock);
626 	return ret;
627 }
628 
629 static struct bpf_htab_elem *sock_hash_alloc_elem(struct bpf_htab *htab,
630 						  void *key, u32 key_size,
631 						  u32 hash, struct sock *sk,
632 						  struct bpf_htab_elem *old)
633 {
634 	struct bpf_htab_elem *new;
635 
636 	if (atomic_inc_return(&htab->count) > htab->map.max_entries) {
637 		if (!old) {
638 			atomic_dec(&htab->count);
639 			return ERR_PTR(-E2BIG);
640 		}
641 	}
642 
643 	new = kmalloc_node(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN,
644 			   htab->map.numa_node);
645 	if (!new) {
646 		atomic_dec(&htab->count);
647 		return ERR_PTR(-ENOMEM);
648 	}
649 	memcpy(new->key, key, key_size);
650 	new->sk = sk;
651 	new->hash = hash;
652 	return new;
653 }
654 
655 static int sock_hash_update_common(struct bpf_map *map, void *key,
656 				   struct sock *sk, u64 flags)
657 {
658 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
659 	struct inet_connection_sock *icsk = inet_csk(sk);
660 	u32 key_size = map->key_size, hash;
661 	struct bpf_htab_elem *elem, *elem_new;
662 	struct bpf_htab_bucket *bucket;
663 	struct sk_psock_link *link;
664 	struct sk_psock *psock;
665 	int ret;
666 
667 	WARN_ON_ONCE(!rcu_read_lock_held());
668 	if (unlikely(flags > BPF_EXIST))
669 		return -EINVAL;
670 	if (unlikely(icsk->icsk_ulp_data))
671 		return -EINVAL;
672 
673 	link = sk_psock_init_link();
674 	if (!link)
675 		return -ENOMEM;
676 
677 	ret = sock_map_link(map, &htab->progs, sk);
678 	if (ret < 0)
679 		goto out_free;
680 
681 	psock = sk_psock(sk);
682 	WARN_ON_ONCE(!psock);
683 
684 	hash = sock_hash_bucket_hash(key, key_size);
685 	bucket = sock_hash_select_bucket(htab, hash);
686 
687 	raw_spin_lock_bh(&bucket->lock);
688 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
689 	if (elem && flags == BPF_NOEXIST) {
690 		ret = -EEXIST;
691 		goto out_unlock;
692 	} else if (!elem && flags == BPF_EXIST) {
693 		ret = -ENOENT;
694 		goto out_unlock;
695 	}
696 
697 	elem_new = sock_hash_alloc_elem(htab, key, key_size, hash, sk, elem);
698 	if (IS_ERR(elem_new)) {
699 		ret = PTR_ERR(elem_new);
700 		goto out_unlock;
701 	}
702 
703 	sock_map_add_link(psock, link, map, elem_new);
704 	/* Add new element to the head of the list, so that
705 	 * concurrent search will find it before old elem.
706 	 */
707 	hlist_add_head_rcu(&elem_new->node, &bucket->head);
708 	if (elem) {
709 		hlist_del_rcu(&elem->node);
710 		sock_map_unref(elem->sk, elem);
711 		sock_hash_free_elem(htab, elem);
712 	}
713 	raw_spin_unlock_bh(&bucket->lock);
714 	return 0;
715 out_unlock:
716 	raw_spin_unlock_bh(&bucket->lock);
717 	sk_psock_put(sk, psock);
718 out_free:
719 	sk_psock_free_link(link);
720 	return ret;
721 }
722 
723 static int sock_hash_update_elem(struct bpf_map *map, void *key,
724 				 void *value, u64 flags)
725 {
726 	u32 ufd = *(u32 *)value;
727 	struct socket *sock;
728 	struct sock *sk;
729 	int ret;
730 
731 	sock = sockfd_lookup(ufd, &ret);
732 	if (!sock)
733 		return ret;
734 	sk = sock->sk;
735 	if (!sk) {
736 		ret = -EINVAL;
737 		goto out;
738 	}
739 	if (!sock_map_sk_is_suitable(sk) ||
740 	    sk->sk_state != TCP_ESTABLISHED) {
741 		ret = -EOPNOTSUPP;
742 		goto out;
743 	}
744 
745 	sock_map_sk_acquire(sk);
746 	ret = sock_hash_update_common(map, key, sk, flags);
747 	sock_map_sk_release(sk);
748 out:
749 	fput(sock->file);
750 	return ret;
751 }
752 
753 static int sock_hash_get_next_key(struct bpf_map *map, void *key,
754 				  void *key_next)
755 {
756 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
757 	struct bpf_htab_elem *elem, *elem_next;
758 	u32 hash, key_size = map->key_size;
759 	struct hlist_head *head;
760 	int i = 0;
761 
762 	if (!key)
763 		goto find_first_elem;
764 	hash = sock_hash_bucket_hash(key, key_size);
765 	head = &sock_hash_select_bucket(htab, hash)->head;
766 	elem = sock_hash_lookup_elem_raw(head, hash, key, key_size);
767 	if (!elem)
768 		goto find_first_elem;
769 
770 	elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&elem->node)),
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 	i = hash & (htab->buckets_num - 1);
778 	i++;
779 find_first_elem:
780 	for (; i < htab->buckets_num; i++) {
781 		head = &sock_hash_select_bucket(htab, i)->head;
782 		elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
783 					     struct bpf_htab_elem, node);
784 		if (elem_next) {
785 			memcpy(key_next, elem_next->key, key_size);
786 			return 0;
787 		}
788 	}
789 
790 	return -ENOENT;
791 }
792 
793 static struct bpf_map *sock_hash_alloc(union bpf_attr *attr)
794 {
795 	struct bpf_htab *htab;
796 	int i, err;
797 	u64 cost;
798 
799 	if (!capable(CAP_NET_ADMIN))
800 		return ERR_PTR(-EPERM);
801 	if (attr->max_entries == 0 ||
802 	    attr->key_size    == 0 ||
803 	    attr->value_size  != 4 ||
804 	    attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
805 		return ERR_PTR(-EINVAL);
806 	if (attr->key_size > MAX_BPF_STACK)
807 		return ERR_PTR(-E2BIG);
808 
809 	htab = kzalloc(sizeof(*htab), GFP_USER);
810 	if (!htab)
811 		return ERR_PTR(-ENOMEM);
812 
813 	bpf_map_init_from_attr(&htab->map, attr);
814 
815 	htab->buckets_num = roundup_pow_of_two(htab->map.max_entries);
816 	htab->elem_size = sizeof(struct bpf_htab_elem) +
817 			  round_up(htab->map.key_size, 8);
818 	if (htab->buckets_num == 0 ||
819 	    htab->buckets_num > U32_MAX / sizeof(struct bpf_htab_bucket)) {
820 		err = -EINVAL;
821 		goto free_htab;
822 	}
823 
824 	cost = (u64) htab->buckets_num * sizeof(struct bpf_htab_bucket) +
825 	       (u64) htab->elem_size * htab->map.max_entries;
826 	if (cost >= U32_MAX - PAGE_SIZE) {
827 		err = -EINVAL;
828 		goto free_htab;
829 	}
830 
831 	htab->buckets = bpf_map_area_alloc(htab->buckets_num *
832 					   sizeof(struct bpf_htab_bucket),
833 					   htab->map.numa_node);
834 	if (!htab->buckets) {
835 		err = -ENOMEM;
836 		goto free_htab;
837 	}
838 
839 	for (i = 0; i < htab->buckets_num; i++) {
840 		INIT_HLIST_HEAD(&htab->buckets[i].head);
841 		raw_spin_lock_init(&htab->buckets[i].lock);
842 	}
843 
844 	return &htab->map;
845 free_htab:
846 	kfree(htab);
847 	return ERR_PTR(err);
848 }
849 
850 static void sock_hash_free(struct bpf_map *map)
851 {
852 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
853 	struct bpf_htab_bucket *bucket;
854 	struct bpf_htab_elem *elem;
855 	struct hlist_node *node;
856 	int i;
857 
858 	synchronize_rcu();
859 	rcu_read_lock();
860 	for (i = 0; i < htab->buckets_num; i++) {
861 		bucket = sock_hash_select_bucket(htab, i);
862 		raw_spin_lock_bh(&bucket->lock);
863 		hlist_for_each_entry_safe(elem, node, &bucket->head, node) {
864 			hlist_del_rcu(&elem->node);
865 			sock_map_unref(elem->sk, elem);
866 		}
867 		raw_spin_unlock_bh(&bucket->lock);
868 	}
869 	rcu_read_unlock();
870 
871 	bpf_map_area_free(htab->buckets);
872 	kfree(htab);
873 }
874 
875 static void sock_hash_release_progs(struct bpf_map *map)
876 {
877 	psock_progs_drop(&container_of(map, struct bpf_htab, map)->progs);
878 }
879 
880 BPF_CALL_4(bpf_sock_hash_update, struct bpf_sock_ops_kern *, sops,
881 	   struct bpf_map *, map, void *, key, u64, flags)
882 {
883 	WARN_ON_ONCE(!rcu_read_lock_held());
884 
885 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
886 		   sock_map_op_okay(sops)))
887 		return sock_hash_update_common(map, key, sops->sk, flags);
888 	return -EOPNOTSUPP;
889 }
890 
891 const struct bpf_func_proto bpf_sock_hash_update_proto = {
892 	.func		= bpf_sock_hash_update,
893 	.gpl_only	= false,
894 	.pkt_access	= true,
895 	.ret_type	= RET_INTEGER,
896 	.arg1_type	= ARG_PTR_TO_CTX,
897 	.arg2_type	= ARG_CONST_MAP_PTR,
898 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
899 	.arg4_type	= ARG_ANYTHING,
900 };
901 
902 BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb,
903 	   struct bpf_map *, map, void *, key, u64, flags)
904 {
905 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
906 
907 	if (unlikely(flags & ~(BPF_F_INGRESS)))
908 		return SK_DROP;
909 	tcb->bpf.flags = flags;
910 	tcb->bpf.sk_redir = __sock_hash_lookup_elem(map, key);
911 	if (!tcb->bpf.sk_redir)
912 		return SK_DROP;
913 	return SK_PASS;
914 }
915 
916 const struct bpf_func_proto bpf_sk_redirect_hash_proto = {
917 	.func           = bpf_sk_redirect_hash,
918 	.gpl_only       = false,
919 	.ret_type       = RET_INTEGER,
920 	.arg1_type	= ARG_PTR_TO_CTX,
921 	.arg2_type      = ARG_CONST_MAP_PTR,
922 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
923 	.arg4_type      = ARG_ANYTHING,
924 };
925 
926 BPF_CALL_4(bpf_msg_redirect_hash, struct sk_msg *, msg,
927 	   struct bpf_map *, map, void *, key, u64, flags)
928 {
929 	if (unlikely(flags & ~(BPF_F_INGRESS)))
930 		return SK_DROP;
931 	msg->flags = flags;
932 	msg->sk_redir = __sock_hash_lookup_elem(map, key);
933 	if (!msg->sk_redir)
934 		return SK_DROP;
935 	return SK_PASS;
936 }
937 
938 const struct bpf_func_proto bpf_msg_redirect_hash_proto = {
939 	.func           = bpf_msg_redirect_hash,
940 	.gpl_only       = false,
941 	.ret_type       = RET_INTEGER,
942 	.arg1_type	= ARG_PTR_TO_CTX,
943 	.arg2_type      = ARG_CONST_MAP_PTR,
944 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
945 	.arg4_type      = ARG_ANYTHING,
946 };
947 
948 const struct bpf_map_ops sock_hash_ops = {
949 	.map_alloc		= sock_hash_alloc,
950 	.map_free		= sock_hash_free,
951 	.map_get_next_key	= sock_hash_get_next_key,
952 	.map_update_elem	= sock_hash_update_elem,
953 	.map_delete_elem	= sock_hash_delete_elem,
954 	.map_lookup_elem	= sock_map_lookup,
955 	.map_release_uref	= sock_hash_release_progs,
956 	.map_check_btf		= map_check_no_btf,
957 };
958 
959 static struct sk_psock_progs *sock_map_progs(struct bpf_map *map)
960 {
961 	switch (map->map_type) {
962 	case BPF_MAP_TYPE_SOCKMAP:
963 		return &container_of(map, struct bpf_stab, map)->progs;
964 	case BPF_MAP_TYPE_SOCKHASH:
965 		return &container_of(map, struct bpf_htab, map)->progs;
966 	default:
967 		break;
968 	}
969 
970 	return NULL;
971 }
972 
973 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
974 			 u32 which)
975 {
976 	struct sk_psock_progs *progs = sock_map_progs(map);
977 
978 	if (!progs)
979 		return -EOPNOTSUPP;
980 
981 	switch (which) {
982 	case BPF_SK_MSG_VERDICT:
983 		psock_set_prog(&progs->msg_parser, prog);
984 		break;
985 	case BPF_SK_SKB_STREAM_PARSER:
986 		psock_set_prog(&progs->skb_parser, prog);
987 		break;
988 	case BPF_SK_SKB_STREAM_VERDICT:
989 		psock_set_prog(&progs->skb_verdict, prog);
990 		break;
991 	default:
992 		return -EOPNOTSUPP;
993 	}
994 
995 	return 0;
996 }
997 
998 void sk_psock_unlink(struct sock *sk, struct sk_psock_link *link)
999 {
1000 	switch (link->map->map_type) {
1001 	case BPF_MAP_TYPE_SOCKMAP:
1002 		return sock_map_delete_from_link(link->map, sk,
1003 						 link->link_raw);
1004 	case BPF_MAP_TYPE_SOCKHASH:
1005 		return sock_hash_delete_from_link(link->map, sk,
1006 						  link->link_raw);
1007 	default:
1008 		break;
1009 	}
1010 }
1011