xref: /openbmc/linux/net/core/sock_map.c (revision e8ec0493)
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 #include <linux/sock_diag.h>
14 #include <net/udp.h>
15 
16 struct bpf_stab {
17 	struct bpf_map map;
18 	struct sock **sks;
19 	struct sk_psock_progs progs;
20 	raw_spinlock_t lock;
21 };
22 
23 #define SOCK_CREATE_FLAG_MASK				\
24 	(BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY)
25 
26 static struct bpf_map *sock_map_alloc(union bpf_attr *attr)
27 {
28 	struct bpf_stab *stab;
29 	u64 cost;
30 	int err;
31 
32 	if (!capable(CAP_NET_ADMIN))
33 		return ERR_PTR(-EPERM);
34 	if (attr->max_entries == 0 ||
35 	    attr->key_size    != 4 ||
36 	    (attr->value_size != sizeof(u32) &&
37 	     attr->value_size != sizeof(u64)) ||
38 	    attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
39 		return ERR_PTR(-EINVAL);
40 
41 	stab = kzalloc(sizeof(*stab), GFP_USER);
42 	if (!stab)
43 		return ERR_PTR(-ENOMEM);
44 
45 	bpf_map_init_from_attr(&stab->map, attr);
46 	raw_spin_lock_init(&stab->lock);
47 
48 	/* Make sure page count doesn't overflow. */
49 	cost = (u64) stab->map.max_entries * sizeof(struct sock *);
50 	err = bpf_map_charge_init(&stab->map.memory, cost);
51 	if (err)
52 		goto free_stab;
53 
54 	stab->sks = bpf_map_area_alloc(stab->map.max_entries *
55 				       sizeof(struct sock *),
56 				       stab->map.numa_node);
57 	if (stab->sks)
58 		return &stab->map;
59 	err = -ENOMEM;
60 	bpf_map_charge_finish(&stab->map.memory);
61 free_stab:
62 	kfree(stab);
63 	return ERR_PTR(err);
64 }
65 
66 int sock_map_get_from_fd(const union bpf_attr *attr, struct bpf_prog *prog)
67 {
68 	u32 ufd = attr->target_fd;
69 	struct bpf_map *map;
70 	struct fd f;
71 	int ret;
72 
73 	f = fdget(ufd);
74 	map = __bpf_map_get(f);
75 	if (IS_ERR(map))
76 		return PTR_ERR(map);
77 	ret = sock_map_prog_update(map, prog, attr->attach_type);
78 	fdput(f);
79 	return ret;
80 }
81 
82 static void sock_map_sk_acquire(struct sock *sk)
83 	__acquires(&sk->sk_lock.slock)
84 {
85 	lock_sock(sk);
86 	preempt_disable();
87 	rcu_read_lock();
88 }
89 
90 static void sock_map_sk_release(struct sock *sk)
91 	__releases(&sk->sk_lock.slock)
92 {
93 	rcu_read_unlock();
94 	preempt_enable();
95 	release_sock(sk);
96 }
97 
98 static void sock_map_add_link(struct sk_psock *psock,
99 			      struct sk_psock_link *link,
100 			      struct bpf_map *map, void *link_raw)
101 {
102 	link->link_raw = link_raw;
103 	link->map = map;
104 	spin_lock_bh(&psock->link_lock);
105 	list_add_tail(&link->list, &psock->link);
106 	spin_unlock_bh(&psock->link_lock);
107 }
108 
109 static void sock_map_del_link(struct sock *sk,
110 			      struct sk_psock *psock, void *link_raw)
111 {
112 	struct sk_psock_link *link, *tmp;
113 	bool strp_stop = false;
114 
115 	spin_lock_bh(&psock->link_lock);
116 	list_for_each_entry_safe(link, tmp, &psock->link, list) {
117 		if (link->link_raw == link_raw) {
118 			struct bpf_map *map = link->map;
119 			struct bpf_stab *stab = container_of(map, struct bpf_stab,
120 							     map);
121 			if (psock->parser.enabled && stab->progs.skb_parser)
122 				strp_stop = true;
123 			list_del(&link->list);
124 			sk_psock_free_link(link);
125 		}
126 	}
127 	spin_unlock_bh(&psock->link_lock);
128 	if (strp_stop) {
129 		write_lock_bh(&sk->sk_callback_lock);
130 		sk_psock_stop_strp(sk, psock);
131 		write_unlock_bh(&sk->sk_callback_lock);
132 	}
133 }
134 
135 static void sock_map_unref(struct sock *sk, void *link_raw)
136 {
137 	struct sk_psock *psock = sk_psock(sk);
138 
139 	if (likely(psock)) {
140 		sock_map_del_link(sk, psock, link_raw);
141 		sk_psock_put(sk, psock);
142 	}
143 }
144 
145 static int sock_map_init_proto(struct sock *sk, struct sk_psock *psock)
146 {
147 	struct proto *prot;
148 
149 	sock_owned_by_me(sk);
150 
151 	switch (sk->sk_type) {
152 	case SOCK_STREAM:
153 		prot = tcp_bpf_get_proto(sk, psock);
154 		break;
155 
156 	case SOCK_DGRAM:
157 		prot = udp_bpf_get_proto(sk, psock);
158 		break;
159 
160 	default:
161 		return -EINVAL;
162 	}
163 
164 	if (IS_ERR(prot))
165 		return PTR_ERR(prot);
166 
167 	sk_psock_update_proto(sk, psock, prot);
168 	return 0;
169 }
170 
171 static struct sk_psock *sock_map_psock_get_checked(struct sock *sk)
172 {
173 	struct sk_psock *psock;
174 
175 	rcu_read_lock();
176 	psock = sk_psock(sk);
177 	if (psock) {
178 		if (sk->sk_prot->close != sock_map_close) {
179 			psock = ERR_PTR(-EBUSY);
180 			goto out;
181 		}
182 
183 		if (!refcount_inc_not_zero(&psock->refcnt))
184 			psock = ERR_PTR(-EBUSY);
185 	}
186 out:
187 	rcu_read_unlock();
188 	return psock;
189 }
190 
191 static int sock_map_link(struct bpf_map *map, struct sk_psock_progs *progs,
192 			 struct sock *sk)
193 {
194 	struct bpf_prog *msg_parser, *skb_parser, *skb_verdict;
195 	struct sk_psock *psock;
196 	bool skb_progs;
197 	int ret;
198 
199 	skb_verdict = READ_ONCE(progs->skb_verdict);
200 	skb_parser = READ_ONCE(progs->skb_parser);
201 	skb_progs = skb_parser && skb_verdict;
202 	if (skb_progs) {
203 		skb_verdict = bpf_prog_inc_not_zero(skb_verdict);
204 		if (IS_ERR(skb_verdict))
205 			return PTR_ERR(skb_verdict);
206 		skb_parser = bpf_prog_inc_not_zero(skb_parser);
207 		if (IS_ERR(skb_parser)) {
208 			bpf_prog_put(skb_verdict);
209 			return PTR_ERR(skb_parser);
210 		}
211 	}
212 
213 	msg_parser = READ_ONCE(progs->msg_parser);
214 	if (msg_parser) {
215 		msg_parser = bpf_prog_inc_not_zero(msg_parser);
216 		if (IS_ERR(msg_parser)) {
217 			ret = PTR_ERR(msg_parser);
218 			goto out;
219 		}
220 	}
221 
222 	psock = sock_map_psock_get_checked(sk);
223 	if (IS_ERR(psock)) {
224 		ret = PTR_ERR(psock);
225 		goto out_progs;
226 	}
227 
228 	if (psock) {
229 		if ((msg_parser && READ_ONCE(psock->progs.msg_parser)) ||
230 		    (skb_progs  && READ_ONCE(psock->progs.skb_parser))) {
231 			sk_psock_put(sk, psock);
232 			ret = -EBUSY;
233 			goto out_progs;
234 		}
235 	} else {
236 		psock = sk_psock_init(sk, map->numa_node);
237 		if (!psock) {
238 			ret = -ENOMEM;
239 			goto out_progs;
240 		}
241 	}
242 
243 	if (msg_parser)
244 		psock_set_prog(&psock->progs.msg_parser, msg_parser);
245 
246 	ret = sock_map_init_proto(sk, psock);
247 	if (ret < 0)
248 		goto out_drop;
249 
250 	write_lock_bh(&sk->sk_callback_lock);
251 	if (skb_progs && !psock->parser.enabled) {
252 		ret = sk_psock_init_strp(sk, psock);
253 		if (ret) {
254 			write_unlock_bh(&sk->sk_callback_lock);
255 			goto out_drop;
256 		}
257 		psock_set_prog(&psock->progs.skb_verdict, skb_verdict);
258 		psock_set_prog(&psock->progs.skb_parser, skb_parser);
259 		sk_psock_start_strp(sk, psock);
260 	}
261 	write_unlock_bh(&sk->sk_callback_lock);
262 	return 0;
263 out_drop:
264 	sk_psock_put(sk, psock);
265 out_progs:
266 	if (msg_parser)
267 		bpf_prog_put(msg_parser);
268 out:
269 	if (skb_progs) {
270 		bpf_prog_put(skb_verdict);
271 		bpf_prog_put(skb_parser);
272 	}
273 	return ret;
274 }
275 
276 static int sock_map_link_no_progs(struct bpf_map *map, struct sock *sk)
277 {
278 	struct sk_psock *psock;
279 	int ret;
280 
281 	psock = sock_map_psock_get_checked(sk);
282 	if (IS_ERR(psock))
283 		return PTR_ERR(psock);
284 
285 	if (!psock) {
286 		psock = sk_psock_init(sk, map->numa_node);
287 		if (!psock)
288 			return -ENOMEM;
289 	}
290 
291 	ret = sock_map_init_proto(sk, psock);
292 	if (ret < 0)
293 		sk_psock_put(sk, psock);
294 	return ret;
295 }
296 
297 static void sock_map_free(struct bpf_map *map)
298 {
299 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
300 	int i;
301 
302 	/* After the sync no updates or deletes will be in-flight so it
303 	 * is safe to walk map and remove entries without risking a race
304 	 * in EEXIST update case.
305 	 */
306 	synchronize_rcu();
307 	for (i = 0; i < stab->map.max_entries; i++) {
308 		struct sock **psk = &stab->sks[i];
309 		struct sock *sk;
310 
311 		sk = xchg(psk, NULL);
312 		if (sk) {
313 			lock_sock(sk);
314 			rcu_read_lock();
315 			sock_map_unref(sk, psk);
316 			rcu_read_unlock();
317 			release_sock(sk);
318 		}
319 	}
320 
321 	/* wait for psock readers accessing its map link */
322 	synchronize_rcu();
323 
324 	bpf_map_area_free(stab->sks);
325 	kfree(stab);
326 }
327 
328 static void sock_map_release_progs(struct bpf_map *map)
329 {
330 	psock_progs_drop(&container_of(map, struct bpf_stab, map)->progs);
331 }
332 
333 static struct sock *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
334 {
335 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
336 
337 	WARN_ON_ONCE(!rcu_read_lock_held());
338 
339 	if (unlikely(key >= map->max_entries))
340 		return NULL;
341 	return READ_ONCE(stab->sks[key]);
342 }
343 
344 static void *sock_map_lookup(struct bpf_map *map, void *key)
345 {
346 	struct sock *sk;
347 
348 	sk = __sock_map_lookup_elem(map, *(u32 *)key);
349 	if (!sk || !sk_fullsock(sk))
350 		return NULL;
351 	if (sk_is_refcounted(sk) && !refcount_inc_not_zero(&sk->sk_refcnt))
352 		return NULL;
353 	return sk;
354 }
355 
356 static void *sock_map_lookup_sys(struct bpf_map *map, void *key)
357 {
358 	struct sock *sk;
359 
360 	if (map->value_size != sizeof(u64))
361 		return ERR_PTR(-ENOSPC);
362 
363 	sk = __sock_map_lookup_elem(map, *(u32 *)key);
364 	if (!sk)
365 		return ERR_PTR(-ENOENT);
366 
367 	sock_gen_cookie(sk);
368 	return &sk->sk_cookie;
369 }
370 
371 static int __sock_map_delete(struct bpf_stab *stab, struct sock *sk_test,
372 			     struct sock **psk)
373 {
374 	struct sock *sk;
375 	int err = 0;
376 
377 	raw_spin_lock_bh(&stab->lock);
378 	sk = *psk;
379 	if (!sk_test || sk_test == sk)
380 		sk = xchg(psk, NULL);
381 
382 	if (likely(sk))
383 		sock_map_unref(sk, psk);
384 	else
385 		err = -EINVAL;
386 
387 	raw_spin_unlock_bh(&stab->lock);
388 	return err;
389 }
390 
391 static void sock_map_delete_from_link(struct bpf_map *map, struct sock *sk,
392 				      void *link_raw)
393 {
394 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
395 
396 	__sock_map_delete(stab, sk, link_raw);
397 }
398 
399 static int sock_map_delete_elem(struct bpf_map *map, void *key)
400 {
401 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
402 	u32 i = *(u32 *)key;
403 	struct sock **psk;
404 
405 	if (unlikely(i >= map->max_entries))
406 		return -EINVAL;
407 
408 	psk = &stab->sks[i];
409 	return __sock_map_delete(stab, NULL, psk);
410 }
411 
412 static int sock_map_get_next_key(struct bpf_map *map, void *key, void *next)
413 {
414 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
415 	u32 i = key ? *(u32 *)key : U32_MAX;
416 	u32 *key_next = next;
417 
418 	if (i == stab->map.max_entries - 1)
419 		return -ENOENT;
420 	if (i >= stab->map.max_entries)
421 		*key_next = 0;
422 	else
423 		*key_next = i + 1;
424 	return 0;
425 }
426 
427 static bool sock_map_redirect_allowed(const struct sock *sk)
428 {
429 	return sk->sk_state != TCP_LISTEN;
430 }
431 
432 static int sock_map_update_common(struct bpf_map *map, u32 idx,
433 				  struct sock *sk, u64 flags)
434 {
435 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
436 	struct sk_psock_link *link;
437 	struct sk_psock *psock;
438 	struct sock *osk;
439 	int ret;
440 
441 	WARN_ON_ONCE(!rcu_read_lock_held());
442 	if (unlikely(flags > BPF_EXIST))
443 		return -EINVAL;
444 	if (unlikely(idx >= map->max_entries))
445 		return -E2BIG;
446 	if (inet_csk_has_ulp(sk))
447 		return -EINVAL;
448 
449 	link = sk_psock_init_link();
450 	if (!link)
451 		return -ENOMEM;
452 
453 	/* Only sockets we can redirect into/from in BPF need to hold
454 	 * refs to parser/verdict progs and have their sk_data_ready
455 	 * and sk_write_space callbacks overridden.
456 	 */
457 	if (sock_map_redirect_allowed(sk))
458 		ret = sock_map_link(map, &stab->progs, sk);
459 	else
460 		ret = sock_map_link_no_progs(map, sk);
461 	if (ret < 0)
462 		goto out_free;
463 
464 	psock = sk_psock(sk);
465 	WARN_ON_ONCE(!psock);
466 
467 	raw_spin_lock_bh(&stab->lock);
468 	osk = stab->sks[idx];
469 	if (osk && flags == BPF_NOEXIST) {
470 		ret = -EEXIST;
471 		goto out_unlock;
472 	} else if (!osk && flags == BPF_EXIST) {
473 		ret = -ENOENT;
474 		goto out_unlock;
475 	}
476 
477 	sock_map_add_link(psock, link, map, &stab->sks[idx]);
478 	stab->sks[idx] = sk;
479 	if (osk)
480 		sock_map_unref(osk, &stab->sks[idx]);
481 	raw_spin_unlock_bh(&stab->lock);
482 	return 0;
483 out_unlock:
484 	raw_spin_unlock_bh(&stab->lock);
485 	if (psock)
486 		sk_psock_put(sk, psock);
487 out_free:
488 	sk_psock_free_link(link);
489 	return ret;
490 }
491 
492 static bool sock_map_op_okay(const struct bpf_sock_ops_kern *ops)
493 {
494 	return ops->op == BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB ||
495 	       ops->op == BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB ||
496 	       ops->op == BPF_SOCK_OPS_TCP_LISTEN_CB;
497 }
498 
499 static bool sk_is_tcp(const struct sock *sk)
500 {
501 	return sk->sk_type == SOCK_STREAM &&
502 	       sk->sk_protocol == IPPROTO_TCP;
503 }
504 
505 static bool sk_is_udp(const struct sock *sk)
506 {
507 	return sk->sk_type == SOCK_DGRAM &&
508 	       sk->sk_protocol == IPPROTO_UDP;
509 }
510 
511 static bool sock_map_sk_is_suitable(const struct sock *sk)
512 {
513 	return sk_is_tcp(sk) || sk_is_udp(sk);
514 }
515 
516 static bool sock_map_sk_state_allowed(const struct sock *sk)
517 {
518 	if (sk_is_tcp(sk))
519 		return (1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_LISTEN);
520 	else if (sk_is_udp(sk))
521 		return sk_hashed(sk);
522 
523 	return false;
524 }
525 
526 static int sock_map_update_elem(struct bpf_map *map, void *key,
527 				void *value, u64 flags)
528 {
529 	u32 idx = *(u32 *)key;
530 	struct socket *sock;
531 	struct sock *sk;
532 	int ret;
533 	u64 ufd;
534 
535 	if (map->value_size == sizeof(u64))
536 		ufd = *(u64 *)value;
537 	else
538 		ufd = *(u32 *)value;
539 	if (ufd > S32_MAX)
540 		return -EINVAL;
541 
542 	sock = sockfd_lookup(ufd, &ret);
543 	if (!sock)
544 		return ret;
545 	sk = sock->sk;
546 	if (!sk) {
547 		ret = -EINVAL;
548 		goto out;
549 	}
550 	if (!sock_map_sk_is_suitable(sk)) {
551 		ret = -EOPNOTSUPP;
552 		goto out;
553 	}
554 
555 	sock_map_sk_acquire(sk);
556 	if (!sock_map_sk_state_allowed(sk))
557 		ret = -EOPNOTSUPP;
558 	else
559 		ret = sock_map_update_common(map, idx, sk, flags);
560 	sock_map_sk_release(sk);
561 out:
562 	fput(sock->file);
563 	return ret;
564 }
565 
566 BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, sops,
567 	   struct bpf_map *, map, void *, key, u64, flags)
568 {
569 	WARN_ON_ONCE(!rcu_read_lock_held());
570 
571 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
572 		   sock_map_op_okay(sops)))
573 		return sock_map_update_common(map, *(u32 *)key, sops->sk,
574 					      flags);
575 	return -EOPNOTSUPP;
576 }
577 
578 const struct bpf_func_proto bpf_sock_map_update_proto = {
579 	.func		= bpf_sock_map_update,
580 	.gpl_only	= false,
581 	.pkt_access	= true,
582 	.ret_type	= RET_INTEGER,
583 	.arg1_type	= ARG_PTR_TO_CTX,
584 	.arg2_type	= ARG_CONST_MAP_PTR,
585 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
586 	.arg4_type	= ARG_ANYTHING,
587 };
588 
589 BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb,
590 	   struct bpf_map *, map, u32, key, u64, flags)
591 {
592 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
593 	struct sock *sk;
594 
595 	if (unlikely(flags & ~(BPF_F_INGRESS)))
596 		return SK_DROP;
597 
598 	sk = __sock_map_lookup_elem(map, key);
599 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
600 		return SK_DROP;
601 
602 	tcb->bpf.flags = flags;
603 	tcb->bpf.sk_redir = sk;
604 	return SK_PASS;
605 }
606 
607 const struct bpf_func_proto bpf_sk_redirect_map_proto = {
608 	.func           = bpf_sk_redirect_map,
609 	.gpl_only       = false,
610 	.ret_type       = RET_INTEGER,
611 	.arg1_type	= ARG_PTR_TO_CTX,
612 	.arg2_type      = ARG_CONST_MAP_PTR,
613 	.arg3_type      = ARG_ANYTHING,
614 	.arg4_type      = ARG_ANYTHING,
615 };
616 
617 BPF_CALL_4(bpf_msg_redirect_map, struct sk_msg *, msg,
618 	   struct bpf_map *, map, u32, key, u64, flags)
619 {
620 	struct sock *sk;
621 
622 	if (unlikely(flags & ~(BPF_F_INGRESS)))
623 		return SK_DROP;
624 
625 	sk = __sock_map_lookup_elem(map, key);
626 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
627 		return SK_DROP;
628 
629 	msg->flags = flags;
630 	msg->sk_redir = sk;
631 	return SK_PASS;
632 }
633 
634 const struct bpf_func_proto bpf_msg_redirect_map_proto = {
635 	.func           = bpf_msg_redirect_map,
636 	.gpl_only       = false,
637 	.ret_type       = RET_INTEGER,
638 	.arg1_type	= ARG_PTR_TO_CTX,
639 	.arg2_type      = ARG_CONST_MAP_PTR,
640 	.arg3_type      = ARG_ANYTHING,
641 	.arg4_type      = ARG_ANYTHING,
642 };
643 
644 const struct bpf_map_ops sock_map_ops = {
645 	.map_alloc		= sock_map_alloc,
646 	.map_free		= sock_map_free,
647 	.map_get_next_key	= sock_map_get_next_key,
648 	.map_lookup_elem_sys_only = sock_map_lookup_sys,
649 	.map_update_elem	= sock_map_update_elem,
650 	.map_delete_elem	= sock_map_delete_elem,
651 	.map_lookup_elem	= sock_map_lookup,
652 	.map_release_uref	= sock_map_release_progs,
653 	.map_check_btf		= map_check_no_btf,
654 };
655 
656 struct bpf_htab_elem {
657 	struct rcu_head rcu;
658 	u32 hash;
659 	struct sock *sk;
660 	struct hlist_node node;
661 	u8 key[];
662 };
663 
664 struct bpf_htab_bucket {
665 	struct hlist_head head;
666 	raw_spinlock_t lock;
667 };
668 
669 struct bpf_htab {
670 	struct bpf_map map;
671 	struct bpf_htab_bucket *buckets;
672 	u32 buckets_num;
673 	u32 elem_size;
674 	struct sk_psock_progs progs;
675 	atomic_t count;
676 };
677 
678 static inline u32 sock_hash_bucket_hash(const void *key, u32 len)
679 {
680 	return jhash(key, len, 0);
681 }
682 
683 static struct bpf_htab_bucket *sock_hash_select_bucket(struct bpf_htab *htab,
684 						       u32 hash)
685 {
686 	return &htab->buckets[hash & (htab->buckets_num - 1)];
687 }
688 
689 static struct bpf_htab_elem *
690 sock_hash_lookup_elem_raw(struct hlist_head *head, u32 hash, void *key,
691 			  u32 key_size)
692 {
693 	struct bpf_htab_elem *elem;
694 
695 	hlist_for_each_entry_rcu(elem, head, node) {
696 		if (elem->hash == hash &&
697 		    !memcmp(&elem->key, key, key_size))
698 			return elem;
699 	}
700 
701 	return NULL;
702 }
703 
704 static struct sock *__sock_hash_lookup_elem(struct bpf_map *map, void *key)
705 {
706 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
707 	u32 key_size = map->key_size, hash;
708 	struct bpf_htab_bucket *bucket;
709 	struct bpf_htab_elem *elem;
710 
711 	WARN_ON_ONCE(!rcu_read_lock_held());
712 
713 	hash = sock_hash_bucket_hash(key, key_size);
714 	bucket = sock_hash_select_bucket(htab, hash);
715 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
716 
717 	return elem ? elem->sk : NULL;
718 }
719 
720 static void sock_hash_free_elem(struct bpf_htab *htab,
721 				struct bpf_htab_elem *elem)
722 {
723 	atomic_dec(&htab->count);
724 	kfree_rcu(elem, rcu);
725 }
726 
727 static void sock_hash_delete_from_link(struct bpf_map *map, struct sock *sk,
728 				       void *link_raw)
729 {
730 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
731 	struct bpf_htab_elem *elem_probe, *elem = link_raw;
732 	struct bpf_htab_bucket *bucket;
733 
734 	WARN_ON_ONCE(!rcu_read_lock_held());
735 	bucket = sock_hash_select_bucket(htab, elem->hash);
736 
737 	/* elem may be deleted in parallel from the map, but access here
738 	 * is okay since it's going away only after RCU grace period.
739 	 * However, we need to check whether it's still present.
740 	 */
741 	raw_spin_lock_bh(&bucket->lock);
742 	elem_probe = sock_hash_lookup_elem_raw(&bucket->head, elem->hash,
743 					       elem->key, map->key_size);
744 	if (elem_probe && elem_probe == elem) {
745 		hlist_del_rcu(&elem->node);
746 		sock_map_unref(elem->sk, elem);
747 		sock_hash_free_elem(htab, elem);
748 	}
749 	raw_spin_unlock_bh(&bucket->lock);
750 }
751 
752 static int sock_hash_delete_elem(struct bpf_map *map, void *key)
753 {
754 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
755 	u32 hash, key_size = map->key_size;
756 	struct bpf_htab_bucket *bucket;
757 	struct bpf_htab_elem *elem;
758 	int ret = -ENOENT;
759 
760 	hash = sock_hash_bucket_hash(key, key_size);
761 	bucket = sock_hash_select_bucket(htab, hash);
762 
763 	raw_spin_lock_bh(&bucket->lock);
764 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
765 	if (elem) {
766 		hlist_del_rcu(&elem->node);
767 		sock_map_unref(elem->sk, elem);
768 		sock_hash_free_elem(htab, elem);
769 		ret = 0;
770 	}
771 	raw_spin_unlock_bh(&bucket->lock);
772 	return ret;
773 }
774 
775 static struct bpf_htab_elem *sock_hash_alloc_elem(struct bpf_htab *htab,
776 						  void *key, u32 key_size,
777 						  u32 hash, struct sock *sk,
778 						  struct bpf_htab_elem *old)
779 {
780 	struct bpf_htab_elem *new;
781 
782 	if (atomic_inc_return(&htab->count) > htab->map.max_entries) {
783 		if (!old) {
784 			atomic_dec(&htab->count);
785 			return ERR_PTR(-E2BIG);
786 		}
787 	}
788 
789 	new = kmalloc_node(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN,
790 			   htab->map.numa_node);
791 	if (!new) {
792 		atomic_dec(&htab->count);
793 		return ERR_PTR(-ENOMEM);
794 	}
795 	memcpy(new->key, key, key_size);
796 	new->sk = sk;
797 	new->hash = hash;
798 	return new;
799 }
800 
801 static int sock_hash_update_common(struct bpf_map *map, void *key,
802 				   struct sock *sk, u64 flags)
803 {
804 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
805 	u32 key_size = map->key_size, hash;
806 	struct bpf_htab_elem *elem, *elem_new;
807 	struct bpf_htab_bucket *bucket;
808 	struct sk_psock_link *link;
809 	struct sk_psock *psock;
810 	int ret;
811 
812 	WARN_ON_ONCE(!rcu_read_lock_held());
813 	if (unlikely(flags > BPF_EXIST))
814 		return -EINVAL;
815 	if (inet_csk_has_ulp(sk))
816 		return -EINVAL;
817 
818 	link = sk_psock_init_link();
819 	if (!link)
820 		return -ENOMEM;
821 
822 	/* Only sockets we can redirect into/from in BPF need to hold
823 	 * refs to parser/verdict progs and have their sk_data_ready
824 	 * and sk_write_space callbacks overridden.
825 	 */
826 	if (sock_map_redirect_allowed(sk))
827 		ret = sock_map_link(map, &htab->progs, sk);
828 	else
829 		ret = sock_map_link_no_progs(map, sk);
830 	if (ret < 0)
831 		goto out_free;
832 
833 	psock = sk_psock(sk);
834 	WARN_ON_ONCE(!psock);
835 
836 	hash = sock_hash_bucket_hash(key, key_size);
837 	bucket = sock_hash_select_bucket(htab, hash);
838 
839 	raw_spin_lock_bh(&bucket->lock);
840 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
841 	if (elem && flags == BPF_NOEXIST) {
842 		ret = -EEXIST;
843 		goto out_unlock;
844 	} else if (!elem && flags == BPF_EXIST) {
845 		ret = -ENOENT;
846 		goto out_unlock;
847 	}
848 
849 	elem_new = sock_hash_alloc_elem(htab, key, key_size, hash, sk, elem);
850 	if (IS_ERR(elem_new)) {
851 		ret = PTR_ERR(elem_new);
852 		goto out_unlock;
853 	}
854 
855 	sock_map_add_link(psock, link, map, elem_new);
856 	/* Add new element to the head of the list, so that
857 	 * concurrent search will find it before old elem.
858 	 */
859 	hlist_add_head_rcu(&elem_new->node, &bucket->head);
860 	if (elem) {
861 		hlist_del_rcu(&elem->node);
862 		sock_map_unref(elem->sk, elem);
863 		sock_hash_free_elem(htab, elem);
864 	}
865 	raw_spin_unlock_bh(&bucket->lock);
866 	return 0;
867 out_unlock:
868 	raw_spin_unlock_bh(&bucket->lock);
869 	sk_psock_put(sk, psock);
870 out_free:
871 	sk_psock_free_link(link);
872 	return ret;
873 }
874 
875 static int sock_hash_update_elem(struct bpf_map *map, void *key,
876 				 void *value, u64 flags)
877 {
878 	struct socket *sock;
879 	struct sock *sk;
880 	int ret;
881 	u64 ufd;
882 
883 	if (map->value_size == sizeof(u64))
884 		ufd = *(u64 *)value;
885 	else
886 		ufd = *(u32 *)value;
887 	if (ufd > S32_MAX)
888 		return -EINVAL;
889 
890 	sock = sockfd_lookup(ufd, &ret);
891 	if (!sock)
892 		return ret;
893 	sk = sock->sk;
894 	if (!sk) {
895 		ret = -EINVAL;
896 		goto out;
897 	}
898 	if (!sock_map_sk_is_suitable(sk)) {
899 		ret = -EOPNOTSUPP;
900 		goto out;
901 	}
902 
903 	sock_map_sk_acquire(sk);
904 	if (!sock_map_sk_state_allowed(sk))
905 		ret = -EOPNOTSUPP;
906 	else
907 		ret = sock_hash_update_common(map, key, sk, flags);
908 	sock_map_sk_release(sk);
909 out:
910 	fput(sock->file);
911 	return ret;
912 }
913 
914 static int sock_hash_get_next_key(struct bpf_map *map, void *key,
915 				  void *key_next)
916 {
917 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
918 	struct bpf_htab_elem *elem, *elem_next;
919 	u32 hash, key_size = map->key_size;
920 	struct hlist_head *head;
921 	int i = 0;
922 
923 	if (!key)
924 		goto find_first_elem;
925 	hash = sock_hash_bucket_hash(key, key_size);
926 	head = &sock_hash_select_bucket(htab, hash)->head;
927 	elem = sock_hash_lookup_elem_raw(head, hash, key, key_size);
928 	if (!elem)
929 		goto find_first_elem;
930 
931 	elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&elem->node)),
932 				     struct bpf_htab_elem, node);
933 	if (elem_next) {
934 		memcpy(key_next, elem_next->key, key_size);
935 		return 0;
936 	}
937 
938 	i = hash & (htab->buckets_num - 1);
939 	i++;
940 find_first_elem:
941 	for (; i < htab->buckets_num; i++) {
942 		head = &sock_hash_select_bucket(htab, i)->head;
943 		elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
944 					     struct bpf_htab_elem, node);
945 		if (elem_next) {
946 			memcpy(key_next, elem_next->key, key_size);
947 			return 0;
948 		}
949 	}
950 
951 	return -ENOENT;
952 }
953 
954 static struct bpf_map *sock_hash_alloc(union bpf_attr *attr)
955 {
956 	struct bpf_htab *htab;
957 	int i, err;
958 	u64 cost;
959 
960 	if (!capable(CAP_NET_ADMIN))
961 		return ERR_PTR(-EPERM);
962 	if (attr->max_entries == 0 ||
963 	    attr->key_size    == 0 ||
964 	    (attr->value_size != sizeof(u32) &&
965 	     attr->value_size != sizeof(u64)) ||
966 	    attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
967 		return ERR_PTR(-EINVAL);
968 	if (attr->key_size > MAX_BPF_STACK)
969 		return ERR_PTR(-E2BIG);
970 
971 	htab = kzalloc(sizeof(*htab), GFP_USER);
972 	if (!htab)
973 		return ERR_PTR(-ENOMEM);
974 
975 	bpf_map_init_from_attr(&htab->map, attr);
976 
977 	htab->buckets_num = roundup_pow_of_two(htab->map.max_entries);
978 	htab->elem_size = sizeof(struct bpf_htab_elem) +
979 			  round_up(htab->map.key_size, 8);
980 	if (htab->buckets_num == 0 ||
981 	    htab->buckets_num > U32_MAX / sizeof(struct bpf_htab_bucket)) {
982 		err = -EINVAL;
983 		goto free_htab;
984 	}
985 
986 	cost = (u64) htab->buckets_num * sizeof(struct bpf_htab_bucket) +
987 	       (u64) htab->elem_size * htab->map.max_entries;
988 	if (cost >= U32_MAX - PAGE_SIZE) {
989 		err = -EINVAL;
990 		goto free_htab;
991 	}
992 
993 	htab->buckets = bpf_map_area_alloc(htab->buckets_num *
994 					   sizeof(struct bpf_htab_bucket),
995 					   htab->map.numa_node);
996 	if (!htab->buckets) {
997 		err = -ENOMEM;
998 		goto free_htab;
999 	}
1000 
1001 	for (i = 0; i < htab->buckets_num; i++) {
1002 		INIT_HLIST_HEAD(&htab->buckets[i].head);
1003 		raw_spin_lock_init(&htab->buckets[i].lock);
1004 	}
1005 
1006 	return &htab->map;
1007 free_htab:
1008 	kfree(htab);
1009 	return ERR_PTR(err);
1010 }
1011 
1012 static void sock_hash_free(struct bpf_map *map)
1013 {
1014 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
1015 	struct bpf_htab_bucket *bucket;
1016 	struct bpf_htab_elem *elem;
1017 	struct hlist_node *node;
1018 	int i;
1019 
1020 	/* After the sync no updates or deletes will be in-flight so it
1021 	 * is safe to walk map and remove entries without risking a race
1022 	 * in EEXIST update case.
1023 	 */
1024 	synchronize_rcu();
1025 	for (i = 0; i < htab->buckets_num; i++) {
1026 		bucket = sock_hash_select_bucket(htab, i);
1027 		hlist_for_each_entry_safe(elem, node, &bucket->head, node) {
1028 			hlist_del_rcu(&elem->node);
1029 			lock_sock(elem->sk);
1030 			rcu_read_lock();
1031 			sock_map_unref(elem->sk, elem);
1032 			rcu_read_unlock();
1033 			release_sock(elem->sk);
1034 		}
1035 	}
1036 
1037 	/* wait for psock readers accessing its map link */
1038 	synchronize_rcu();
1039 
1040 	bpf_map_area_free(htab->buckets);
1041 	kfree(htab);
1042 }
1043 
1044 static void *sock_hash_lookup_sys(struct bpf_map *map, void *key)
1045 {
1046 	struct sock *sk;
1047 
1048 	if (map->value_size != sizeof(u64))
1049 		return ERR_PTR(-ENOSPC);
1050 
1051 	sk = __sock_hash_lookup_elem(map, key);
1052 	if (!sk)
1053 		return ERR_PTR(-ENOENT);
1054 
1055 	sock_gen_cookie(sk);
1056 	return &sk->sk_cookie;
1057 }
1058 
1059 static void *sock_hash_lookup(struct bpf_map *map, void *key)
1060 {
1061 	struct sock *sk;
1062 
1063 	sk = __sock_hash_lookup_elem(map, key);
1064 	if (!sk || !sk_fullsock(sk))
1065 		return NULL;
1066 	if (sk_is_refcounted(sk) && !refcount_inc_not_zero(&sk->sk_refcnt))
1067 		return NULL;
1068 	return sk;
1069 }
1070 
1071 static void sock_hash_release_progs(struct bpf_map *map)
1072 {
1073 	psock_progs_drop(&container_of(map, struct bpf_htab, map)->progs);
1074 }
1075 
1076 BPF_CALL_4(bpf_sock_hash_update, struct bpf_sock_ops_kern *, sops,
1077 	   struct bpf_map *, map, void *, key, u64, flags)
1078 {
1079 	WARN_ON_ONCE(!rcu_read_lock_held());
1080 
1081 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
1082 		   sock_map_op_okay(sops)))
1083 		return sock_hash_update_common(map, key, sops->sk, flags);
1084 	return -EOPNOTSUPP;
1085 }
1086 
1087 const struct bpf_func_proto bpf_sock_hash_update_proto = {
1088 	.func		= bpf_sock_hash_update,
1089 	.gpl_only	= false,
1090 	.pkt_access	= true,
1091 	.ret_type	= RET_INTEGER,
1092 	.arg1_type	= ARG_PTR_TO_CTX,
1093 	.arg2_type	= ARG_CONST_MAP_PTR,
1094 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
1095 	.arg4_type	= ARG_ANYTHING,
1096 };
1097 
1098 BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb,
1099 	   struct bpf_map *, map, void *, key, u64, flags)
1100 {
1101 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
1102 	struct sock *sk;
1103 
1104 	if (unlikely(flags & ~(BPF_F_INGRESS)))
1105 		return SK_DROP;
1106 
1107 	sk = __sock_hash_lookup_elem(map, key);
1108 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
1109 		return SK_DROP;
1110 
1111 	tcb->bpf.flags = flags;
1112 	tcb->bpf.sk_redir = sk;
1113 	return SK_PASS;
1114 }
1115 
1116 const struct bpf_func_proto bpf_sk_redirect_hash_proto = {
1117 	.func           = bpf_sk_redirect_hash,
1118 	.gpl_only       = false,
1119 	.ret_type       = RET_INTEGER,
1120 	.arg1_type	= ARG_PTR_TO_CTX,
1121 	.arg2_type      = ARG_CONST_MAP_PTR,
1122 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
1123 	.arg4_type      = ARG_ANYTHING,
1124 };
1125 
1126 BPF_CALL_4(bpf_msg_redirect_hash, struct sk_msg *, msg,
1127 	   struct bpf_map *, map, void *, key, u64, flags)
1128 {
1129 	struct sock *sk;
1130 
1131 	if (unlikely(flags & ~(BPF_F_INGRESS)))
1132 		return SK_DROP;
1133 
1134 	sk = __sock_hash_lookup_elem(map, key);
1135 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
1136 		return SK_DROP;
1137 
1138 	msg->flags = flags;
1139 	msg->sk_redir = sk;
1140 	return SK_PASS;
1141 }
1142 
1143 const struct bpf_func_proto bpf_msg_redirect_hash_proto = {
1144 	.func           = bpf_msg_redirect_hash,
1145 	.gpl_only       = false,
1146 	.ret_type       = RET_INTEGER,
1147 	.arg1_type	= ARG_PTR_TO_CTX,
1148 	.arg2_type      = ARG_CONST_MAP_PTR,
1149 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
1150 	.arg4_type      = ARG_ANYTHING,
1151 };
1152 
1153 const struct bpf_map_ops sock_hash_ops = {
1154 	.map_alloc		= sock_hash_alloc,
1155 	.map_free		= sock_hash_free,
1156 	.map_get_next_key	= sock_hash_get_next_key,
1157 	.map_update_elem	= sock_hash_update_elem,
1158 	.map_delete_elem	= sock_hash_delete_elem,
1159 	.map_lookup_elem	= sock_hash_lookup,
1160 	.map_lookup_elem_sys_only = sock_hash_lookup_sys,
1161 	.map_release_uref	= sock_hash_release_progs,
1162 	.map_check_btf		= map_check_no_btf,
1163 };
1164 
1165 static struct sk_psock_progs *sock_map_progs(struct bpf_map *map)
1166 {
1167 	switch (map->map_type) {
1168 	case BPF_MAP_TYPE_SOCKMAP:
1169 		return &container_of(map, struct bpf_stab, map)->progs;
1170 	case BPF_MAP_TYPE_SOCKHASH:
1171 		return &container_of(map, struct bpf_htab, map)->progs;
1172 	default:
1173 		break;
1174 	}
1175 
1176 	return NULL;
1177 }
1178 
1179 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
1180 			 u32 which)
1181 {
1182 	struct sk_psock_progs *progs = sock_map_progs(map);
1183 
1184 	if (!progs)
1185 		return -EOPNOTSUPP;
1186 
1187 	switch (which) {
1188 	case BPF_SK_MSG_VERDICT:
1189 		psock_set_prog(&progs->msg_parser, prog);
1190 		break;
1191 	case BPF_SK_SKB_STREAM_PARSER:
1192 		psock_set_prog(&progs->skb_parser, prog);
1193 		break;
1194 	case BPF_SK_SKB_STREAM_VERDICT:
1195 		psock_set_prog(&progs->skb_verdict, prog);
1196 		break;
1197 	default:
1198 		return -EOPNOTSUPP;
1199 	}
1200 
1201 	return 0;
1202 }
1203 
1204 static void sock_map_unlink(struct sock *sk, struct sk_psock_link *link)
1205 {
1206 	switch (link->map->map_type) {
1207 	case BPF_MAP_TYPE_SOCKMAP:
1208 		return sock_map_delete_from_link(link->map, sk,
1209 						 link->link_raw);
1210 	case BPF_MAP_TYPE_SOCKHASH:
1211 		return sock_hash_delete_from_link(link->map, sk,
1212 						  link->link_raw);
1213 	default:
1214 		break;
1215 	}
1216 }
1217 
1218 static void sock_map_remove_links(struct sock *sk, struct sk_psock *psock)
1219 {
1220 	struct sk_psock_link *link;
1221 
1222 	while ((link = sk_psock_link_pop(psock))) {
1223 		sock_map_unlink(sk, link);
1224 		sk_psock_free_link(link);
1225 	}
1226 }
1227 
1228 void sock_map_unhash(struct sock *sk)
1229 {
1230 	void (*saved_unhash)(struct sock *sk);
1231 	struct sk_psock *psock;
1232 
1233 	rcu_read_lock();
1234 	psock = sk_psock(sk);
1235 	if (unlikely(!psock)) {
1236 		rcu_read_unlock();
1237 		if (sk->sk_prot->unhash)
1238 			sk->sk_prot->unhash(sk);
1239 		return;
1240 	}
1241 
1242 	saved_unhash = psock->saved_unhash;
1243 	sock_map_remove_links(sk, psock);
1244 	rcu_read_unlock();
1245 	saved_unhash(sk);
1246 }
1247 
1248 void sock_map_close(struct sock *sk, long timeout)
1249 {
1250 	void (*saved_close)(struct sock *sk, long timeout);
1251 	struct sk_psock *psock;
1252 
1253 	lock_sock(sk);
1254 	rcu_read_lock();
1255 	psock = sk_psock(sk);
1256 	if (unlikely(!psock)) {
1257 		rcu_read_unlock();
1258 		release_sock(sk);
1259 		return sk->sk_prot->close(sk, timeout);
1260 	}
1261 
1262 	saved_close = psock->saved_close;
1263 	sock_map_remove_links(sk, psock);
1264 	rcu_read_unlock();
1265 	release_sock(sk);
1266 	saved_close(sk, timeout);
1267 }
1268