xref: /openbmc/linux/net/core/sock_map.c (revision ef9303fd) 
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 static int sock_map_update_common(struct bpf_map *map, u32 idx,
430 				  struct sock *sk, u64 flags)
431 {
432 	struct bpf_stab *stab = container_of(map, struct bpf_stab, map);
433 	struct sk_psock_link *link;
434 	struct sk_psock *psock;
435 	struct sock *osk;
436 	int ret;
437 
438 	WARN_ON_ONCE(!rcu_read_lock_held());
439 	if (unlikely(flags > BPF_EXIST))
440 		return -EINVAL;
441 	if (unlikely(idx >= map->max_entries))
442 		return -E2BIG;
443 	if (inet_csk_has_ulp(sk))
444 		return -EINVAL;
445 
446 	link = sk_psock_init_link();
447 	if (!link)
448 		return -ENOMEM;
449 
450 	/* Only sockets we can redirect into/from in BPF need to hold
451 	 * refs to parser/verdict progs and have their sk_data_ready
452 	 * and sk_write_space callbacks overridden.
453 	 */
454 	if (sock_map_redirect_allowed(sk))
455 		ret = sock_map_link(map, &stab->progs, sk);
456 	else
457 		ret = sock_map_link_no_progs(map, sk);
458 	if (ret < 0)
459 		goto out_free;
460 
461 	psock = sk_psock(sk);
462 	WARN_ON_ONCE(!psock);
463 
464 	raw_spin_lock_bh(&stab->lock);
465 	osk = stab->sks[idx];
466 	if (osk && flags == BPF_NOEXIST) {
467 		ret = -EEXIST;
468 		goto out_unlock;
469 	} else if (!osk && flags == BPF_EXIST) {
470 		ret = -ENOENT;
471 		goto out_unlock;
472 	}
473 
474 	sock_map_add_link(psock, link, map, &stab->sks[idx]);
475 	stab->sks[idx] = sk;
476 	if (osk)
477 		sock_map_unref(osk, &stab->sks[idx]);
478 	raw_spin_unlock_bh(&stab->lock);
479 	return 0;
480 out_unlock:
481 	raw_spin_unlock_bh(&stab->lock);
482 	if (psock)
483 		sk_psock_put(sk, psock);
484 out_free:
485 	sk_psock_free_link(link);
486 	return ret;
487 }
488 
489 static bool sock_map_op_okay(const struct bpf_sock_ops_kern *ops)
490 {
491 	return ops->op == BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB ||
492 	       ops->op == BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB ||
493 	       ops->op == BPF_SOCK_OPS_TCP_LISTEN_CB;
494 }
495 
496 static bool sk_is_tcp(const struct sock *sk)
497 {
498 	return sk->sk_type == SOCK_STREAM &&
499 	       sk->sk_protocol == IPPROTO_TCP;
500 }
501 
502 static bool sk_is_udp(const struct sock *sk)
503 {
504 	return sk->sk_type == SOCK_DGRAM &&
505 	       sk->sk_protocol == IPPROTO_UDP;
506 }
507 
508 static bool sock_map_redirect_allowed(const struct sock *sk)
509 {
510 	return sk_is_tcp(sk) && sk->sk_state != TCP_LISTEN;
511 }
512 
513 static bool sock_map_sk_is_suitable(const struct sock *sk)
514 {
515 	return sk_is_tcp(sk) || sk_is_udp(sk);
516 }
517 
518 static bool sock_map_sk_state_allowed(const struct sock *sk)
519 {
520 	if (sk_is_tcp(sk))
521 		return (1 << sk->sk_state) & (TCPF_ESTABLISHED | TCPF_LISTEN);
522 	else if (sk_is_udp(sk))
523 		return sk_hashed(sk);
524 
525 	return false;
526 }
527 
528 static int sock_map_update_elem(struct bpf_map *map, void *key,
529 				void *value, u64 flags)
530 {
531 	u32 idx = *(u32 *)key;
532 	struct socket *sock;
533 	struct sock *sk;
534 	int ret;
535 	u64 ufd;
536 
537 	if (map->value_size == sizeof(u64))
538 		ufd = *(u64 *)value;
539 	else
540 		ufd = *(u32 *)value;
541 	if (ufd > S32_MAX)
542 		return -EINVAL;
543 
544 	sock = sockfd_lookup(ufd, &ret);
545 	if (!sock)
546 		return ret;
547 	sk = sock->sk;
548 	if (!sk) {
549 		ret = -EINVAL;
550 		goto out;
551 	}
552 	if (!sock_map_sk_is_suitable(sk)) {
553 		ret = -EOPNOTSUPP;
554 		goto out;
555 	}
556 
557 	sock_map_sk_acquire(sk);
558 	if (!sock_map_sk_state_allowed(sk))
559 		ret = -EOPNOTSUPP;
560 	else
561 		ret = sock_map_update_common(map, idx, sk, flags);
562 	sock_map_sk_release(sk);
563 out:
564 	fput(sock->file);
565 	return ret;
566 }
567 
568 BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, sops,
569 	   struct bpf_map *, map, void *, key, u64, flags)
570 {
571 	WARN_ON_ONCE(!rcu_read_lock_held());
572 
573 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
574 		   sock_map_op_okay(sops)))
575 		return sock_map_update_common(map, *(u32 *)key, sops->sk,
576 					      flags);
577 	return -EOPNOTSUPP;
578 }
579 
580 const struct bpf_func_proto bpf_sock_map_update_proto = {
581 	.func		= bpf_sock_map_update,
582 	.gpl_only	= false,
583 	.pkt_access	= true,
584 	.ret_type	= RET_INTEGER,
585 	.arg1_type	= ARG_PTR_TO_CTX,
586 	.arg2_type	= ARG_CONST_MAP_PTR,
587 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
588 	.arg4_type	= ARG_ANYTHING,
589 };
590 
591 BPF_CALL_4(bpf_sk_redirect_map, struct sk_buff *, skb,
592 	   struct bpf_map *, map, u32, key, u64, flags)
593 {
594 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
595 	struct sock *sk;
596 
597 	if (unlikely(flags & ~(BPF_F_INGRESS)))
598 		return SK_DROP;
599 
600 	sk = __sock_map_lookup_elem(map, key);
601 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
602 		return SK_DROP;
603 
604 	tcb->bpf.flags = flags;
605 	tcb->bpf.sk_redir = sk;
606 	return SK_PASS;
607 }
608 
609 const struct bpf_func_proto bpf_sk_redirect_map_proto = {
610 	.func           = bpf_sk_redirect_map,
611 	.gpl_only       = false,
612 	.ret_type       = RET_INTEGER,
613 	.arg1_type	= ARG_PTR_TO_CTX,
614 	.arg2_type      = ARG_CONST_MAP_PTR,
615 	.arg3_type      = ARG_ANYTHING,
616 	.arg4_type      = ARG_ANYTHING,
617 };
618 
619 BPF_CALL_4(bpf_msg_redirect_map, struct sk_msg *, msg,
620 	   struct bpf_map *, map, u32, key, u64, flags)
621 {
622 	struct sock *sk;
623 
624 	if (unlikely(flags & ~(BPF_F_INGRESS)))
625 		return SK_DROP;
626 
627 	sk = __sock_map_lookup_elem(map, key);
628 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
629 		return SK_DROP;
630 
631 	msg->flags = flags;
632 	msg->sk_redir = sk;
633 	return SK_PASS;
634 }
635 
636 const struct bpf_func_proto bpf_msg_redirect_map_proto = {
637 	.func           = bpf_msg_redirect_map,
638 	.gpl_only       = false,
639 	.ret_type       = RET_INTEGER,
640 	.arg1_type	= ARG_PTR_TO_CTX,
641 	.arg2_type      = ARG_CONST_MAP_PTR,
642 	.arg3_type      = ARG_ANYTHING,
643 	.arg4_type      = ARG_ANYTHING,
644 };
645 
646 const struct bpf_map_ops sock_map_ops = {
647 	.map_alloc		= sock_map_alloc,
648 	.map_free		= sock_map_free,
649 	.map_get_next_key	= sock_map_get_next_key,
650 	.map_lookup_elem_sys_only = sock_map_lookup_sys,
651 	.map_update_elem	= sock_map_update_elem,
652 	.map_delete_elem	= sock_map_delete_elem,
653 	.map_lookup_elem	= sock_map_lookup,
654 	.map_release_uref	= sock_map_release_progs,
655 	.map_check_btf		= map_check_no_btf,
656 };
657 
658 struct bpf_htab_elem {
659 	struct rcu_head rcu;
660 	u32 hash;
661 	struct sock *sk;
662 	struct hlist_node node;
663 	u8 key[];
664 };
665 
666 struct bpf_htab_bucket {
667 	struct hlist_head head;
668 	raw_spinlock_t lock;
669 };
670 
671 struct bpf_htab {
672 	struct bpf_map map;
673 	struct bpf_htab_bucket *buckets;
674 	u32 buckets_num;
675 	u32 elem_size;
676 	struct sk_psock_progs progs;
677 	atomic_t count;
678 };
679 
680 static inline u32 sock_hash_bucket_hash(const void *key, u32 len)
681 {
682 	return jhash(key, len, 0);
683 }
684 
685 static struct bpf_htab_bucket *sock_hash_select_bucket(struct bpf_htab *htab,
686 						       u32 hash)
687 {
688 	return &htab->buckets[hash & (htab->buckets_num - 1)];
689 }
690 
691 static struct bpf_htab_elem *
692 sock_hash_lookup_elem_raw(struct hlist_head *head, u32 hash, void *key,
693 			  u32 key_size)
694 {
695 	struct bpf_htab_elem *elem;
696 
697 	hlist_for_each_entry_rcu(elem, head, node) {
698 		if (elem->hash == hash &&
699 		    !memcmp(&elem->key, key, key_size))
700 			return elem;
701 	}
702 
703 	return NULL;
704 }
705 
706 static struct sock *__sock_hash_lookup_elem(struct bpf_map *map, void *key)
707 {
708 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
709 	u32 key_size = map->key_size, hash;
710 	struct bpf_htab_bucket *bucket;
711 	struct bpf_htab_elem *elem;
712 
713 	WARN_ON_ONCE(!rcu_read_lock_held());
714 
715 	hash = sock_hash_bucket_hash(key, key_size);
716 	bucket = sock_hash_select_bucket(htab, hash);
717 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
718 
719 	return elem ? elem->sk : NULL;
720 }
721 
722 static void sock_hash_free_elem(struct bpf_htab *htab,
723 				struct bpf_htab_elem *elem)
724 {
725 	atomic_dec(&htab->count);
726 	kfree_rcu(elem, rcu);
727 }
728 
729 static void sock_hash_delete_from_link(struct bpf_map *map, struct sock *sk,
730 				       void *link_raw)
731 {
732 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
733 	struct bpf_htab_elem *elem_probe, *elem = link_raw;
734 	struct bpf_htab_bucket *bucket;
735 
736 	WARN_ON_ONCE(!rcu_read_lock_held());
737 	bucket = sock_hash_select_bucket(htab, elem->hash);
738 
739 	/* elem may be deleted in parallel from the map, but access here
740 	 * is okay since it's going away only after RCU grace period.
741 	 * However, we need to check whether it's still present.
742 	 */
743 	raw_spin_lock_bh(&bucket->lock);
744 	elem_probe = sock_hash_lookup_elem_raw(&bucket->head, elem->hash,
745 					       elem->key, map->key_size);
746 	if (elem_probe && elem_probe == elem) {
747 		hlist_del_rcu(&elem->node);
748 		sock_map_unref(elem->sk, elem);
749 		sock_hash_free_elem(htab, elem);
750 	}
751 	raw_spin_unlock_bh(&bucket->lock);
752 }
753 
754 static int sock_hash_delete_elem(struct bpf_map *map, void *key)
755 {
756 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
757 	u32 hash, key_size = map->key_size;
758 	struct bpf_htab_bucket *bucket;
759 	struct bpf_htab_elem *elem;
760 	int ret = -ENOENT;
761 
762 	hash = sock_hash_bucket_hash(key, key_size);
763 	bucket = sock_hash_select_bucket(htab, hash);
764 
765 	raw_spin_lock_bh(&bucket->lock);
766 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
767 	if (elem) {
768 		hlist_del_rcu(&elem->node);
769 		sock_map_unref(elem->sk, elem);
770 		sock_hash_free_elem(htab, elem);
771 		ret = 0;
772 	}
773 	raw_spin_unlock_bh(&bucket->lock);
774 	return ret;
775 }
776 
777 static struct bpf_htab_elem *sock_hash_alloc_elem(struct bpf_htab *htab,
778 						  void *key, u32 key_size,
779 						  u32 hash, struct sock *sk,
780 						  struct bpf_htab_elem *old)
781 {
782 	struct bpf_htab_elem *new;
783 
784 	if (atomic_inc_return(&htab->count) > htab->map.max_entries) {
785 		if (!old) {
786 			atomic_dec(&htab->count);
787 			return ERR_PTR(-E2BIG);
788 		}
789 	}
790 
791 	new = kmalloc_node(htab->elem_size, GFP_ATOMIC | __GFP_NOWARN,
792 			   htab->map.numa_node);
793 	if (!new) {
794 		atomic_dec(&htab->count);
795 		return ERR_PTR(-ENOMEM);
796 	}
797 	memcpy(new->key, key, key_size);
798 	new->sk = sk;
799 	new->hash = hash;
800 	return new;
801 }
802 
803 static int sock_hash_update_common(struct bpf_map *map, void *key,
804 				   struct sock *sk, u64 flags)
805 {
806 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
807 	u32 key_size = map->key_size, hash;
808 	struct bpf_htab_elem *elem, *elem_new;
809 	struct bpf_htab_bucket *bucket;
810 	struct sk_psock_link *link;
811 	struct sk_psock *psock;
812 	int ret;
813 
814 	WARN_ON_ONCE(!rcu_read_lock_held());
815 	if (unlikely(flags > BPF_EXIST))
816 		return -EINVAL;
817 	if (inet_csk_has_ulp(sk))
818 		return -EINVAL;
819 
820 	link = sk_psock_init_link();
821 	if (!link)
822 		return -ENOMEM;
823 
824 	/* Only sockets we can redirect into/from in BPF need to hold
825 	 * refs to parser/verdict progs and have their sk_data_ready
826 	 * and sk_write_space callbacks overridden.
827 	 */
828 	if (sock_map_redirect_allowed(sk))
829 		ret = sock_map_link(map, &htab->progs, sk);
830 	else
831 		ret = sock_map_link_no_progs(map, sk);
832 	if (ret < 0)
833 		goto out_free;
834 
835 	psock = sk_psock(sk);
836 	WARN_ON_ONCE(!psock);
837 
838 	hash = sock_hash_bucket_hash(key, key_size);
839 	bucket = sock_hash_select_bucket(htab, hash);
840 
841 	raw_spin_lock_bh(&bucket->lock);
842 	elem = sock_hash_lookup_elem_raw(&bucket->head, hash, key, key_size);
843 	if (elem && flags == BPF_NOEXIST) {
844 		ret = -EEXIST;
845 		goto out_unlock;
846 	} else if (!elem && flags == BPF_EXIST) {
847 		ret = -ENOENT;
848 		goto out_unlock;
849 	}
850 
851 	elem_new = sock_hash_alloc_elem(htab, key, key_size, hash, sk, elem);
852 	if (IS_ERR(elem_new)) {
853 		ret = PTR_ERR(elem_new);
854 		goto out_unlock;
855 	}
856 
857 	sock_map_add_link(psock, link, map, elem_new);
858 	/* Add new element to the head of the list, so that
859 	 * concurrent search will find it before old elem.
860 	 */
861 	hlist_add_head_rcu(&elem_new->node, &bucket->head);
862 	if (elem) {
863 		hlist_del_rcu(&elem->node);
864 		sock_map_unref(elem->sk, elem);
865 		sock_hash_free_elem(htab, elem);
866 	}
867 	raw_spin_unlock_bh(&bucket->lock);
868 	return 0;
869 out_unlock:
870 	raw_spin_unlock_bh(&bucket->lock);
871 	sk_psock_put(sk, psock);
872 out_free:
873 	sk_psock_free_link(link);
874 	return ret;
875 }
876 
877 static int sock_hash_update_elem(struct bpf_map *map, void *key,
878 				 void *value, u64 flags)
879 {
880 	struct socket *sock;
881 	struct sock *sk;
882 	int ret;
883 	u64 ufd;
884 
885 	if (map->value_size == sizeof(u64))
886 		ufd = *(u64 *)value;
887 	else
888 		ufd = *(u32 *)value;
889 	if (ufd > S32_MAX)
890 		return -EINVAL;
891 
892 	sock = sockfd_lookup(ufd, &ret);
893 	if (!sock)
894 		return ret;
895 	sk = sock->sk;
896 	if (!sk) {
897 		ret = -EINVAL;
898 		goto out;
899 	}
900 	if (!sock_map_sk_is_suitable(sk)) {
901 		ret = -EOPNOTSUPP;
902 		goto out;
903 	}
904 
905 	sock_map_sk_acquire(sk);
906 	if (!sock_map_sk_state_allowed(sk))
907 		ret = -EOPNOTSUPP;
908 	else
909 		ret = sock_hash_update_common(map, key, sk, flags);
910 	sock_map_sk_release(sk);
911 out:
912 	fput(sock->file);
913 	return ret;
914 }
915 
916 static int sock_hash_get_next_key(struct bpf_map *map, void *key,
917 				  void *key_next)
918 {
919 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
920 	struct bpf_htab_elem *elem, *elem_next;
921 	u32 hash, key_size = map->key_size;
922 	struct hlist_head *head;
923 	int i = 0;
924 
925 	if (!key)
926 		goto find_first_elem;
927 	hash = sock_hash_bucket_hash(key, key_size);
928 	head = &sock_hash_select_bucket(htab, hash)->head;
929 	elem = sock_hash_lookup_elem_raw(head, hash, key, key_size);
930 	if (!elem)
931 		goto find_first_elem;
932 
933 	elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_next_rcu(&elem->node)),
934 				     struct bpf_htab_elem, node);
935 	if (elem_next) {
936 		memcpy(key_next, elem_next->key, key_size);
937 		return 0;
938 	}
939 
940 	i = hash & (htab->buckets_num - 1);
941 	i++;
942 find_first_elem:
943 	for (; i < htab->buckets_num; i++) {
944 		head = &sock_hash_select_bucket(htab, i)->head;
945 		elem_next = hlist_entry_safe(rcu_dereference_raw(hlist_first_rcu(head)),
946 					     struct bpf_htab_elem, node);
947 		if (elem_next) {
948 			memcpy(key_next, elem_next->key, key_size);
949 			return 0;
950 		}
951 	}
952 
953 	return -ENOENT;
954 }
955 
956 static struct bpf_map *sock_hash_alloc(union bpf_attr *attr)
957 {
958 	struct bpf_htab *htab;
959 	int i, err;
960 	u64 cost;
961 
962 	if (!capable(CAP_NET_ADMIN))
963 		return ERR_PTR(-EPERM);
964 	if (attr->max_entries == 0 ||
965 	    attr->key_size    == 0 ||
966 	    (attr->value_size != sizeof(u32) &&
967 	     attr->value_size != sizeof(u64)) ||
968 	    attr->map_flags & ~SOCK_CREATE_FLAG_MASK)
969 		return ERR_PTR(-EINVAL);
970 	if (attr->key_size > MAX_BPF_STACK)
971 		return ERR_PTR(-E2BIG);
972 
973 	htab = kzalloc(sizeof(*htab), GFP_USER);
974 	if (!htab)
975 		return ERR_PTR(-ENOMEM);
976 
977 	bpf_map_init_from_attr(&htab->map, attr);
978 
979 	htab->buckets_num = roundup_pow_of_two(htab->map.max_entries);
980 	htab->elem_size = sizeof(struct bpf_htab_elem) +
981 			  round_up(htab->map.key_size, 8);
982 	if (htab->buckets_num == 0 ||
983 	    htab->buckets_num > U32_MAX / sizeof(struct bpf_htab_bucket)) {
984 		err = -EINVAL;
985 		goto free_htab;
986 	}
987 
988 	cost = (u64) htab->buckets_num * sizeof(struct bpf_htab_bucket) +
989 	       (u64) htab->elem_size * htab->map.max_entries;
990 	if (cost >= U32_MAX - PAGE_SIZE) {
991 		err = -EINVAL;
992 		goto free_htab;
993 	}
994 	err = bpf_map_charge_init(&htab->map.memory, cost);
995 	if (err)
996 		goto free_htab;
997 
998 	htab->buckets = bpf_map_area_alloc(htab->buckets_num *
999 					   sizeof(struct bpf_htab_bucket),
1000 					   htab->map.numa_node);
1001 	if (!htab->buckets) {
1002 		bpf_map_charge_finish(&htab->map.memory);
1003 		err = -ENOMEM;
1004 		goto free_htab;
1005 	}
1006 
1007 	for (i = 0; i < htab->buckets_num; i++) {
1008 		INIT_HLIST_HEAD(&htab->buckets[i].head);
1009 		raw_spin_lock_init(&htab->buckets[i].lock);
1010 	}
1011 
1012 	return &htab->map;
1013 free_htab:
1014 	kfree(htab);
1015 	return ERR_PTR(err);
1016 }
1017 
1018 static void sock_hash_free(struct bpf_map *map)
1019 {
1020 	struct bpf_htab *htab = container_of(map, struct bpf_htab, map);
1021 	struct bpf_htab_bucket *bucket;
1022 	struct hlist_head unlink_list;
1023 	struct bpf_htab_elem *elem;
1024 	struct hlist_node *node;
1025 	int i;
1026 
1027 	/* After the sync no updates or deletes will be in-flight so it
1028 	 * is safe to walk map and remove entries without risking a race
1029 	 * in EEXIST update case.
1030 	 */
1031 	synchronize_rcu();
1032 	for (i = 0; i < htab->buckets_num; i++) {
1033 		bucket = sock_hash_select_bucket(htab, i);
1034 
1035 		/* We are racing with sock_hash_delete_from_link to
1036 		 * enter the spin-lock critical section. Every socket on
1037 		 * the list is still linked to sockhash. Since link
1038 		 * exists, psock exists and holds a ref to socket. That
1039 		 * lets us to grab a socket ref too.
1040 		 */
1041 		raw_spin_lock_bh(&bucket->lock);
1042 		hlist_for_each_entry(elem, &bucket->head, node)
1043 			sock_hold(elem->sk);
1044 		hlist_move_list(&bucket->head, &unlink_list);
1045 		raw_spin_unlock_bh(&bucket->lock);
1046 
1047 		/* Process removed entries out of atomic context to
1048 		 * block for socket lock before deleting the psock's
1049 		 * link to sockhash.
1050 		 */
1051 		hlist_for_each_entry_safe(elem, node, &unlink_list, node) {
1052 			hlist_del(&elem->node);
1053 			lock_sock(elem->sk);
1054 			rcu_read_lock();
1055 			sock_map_unref(elem->sk, elem);
1056 			rcu_read_unlock();
1057 			release_sock(elem->sk);
1058 			sock_put(elem->sk);
1059 			sock_hash_free_elem(htab, elem);
1060 		}
1061 	}
1062 
1063 	/* wait for psock readers accessing its map link */
1064 	synchronize_rcu();
1065 
1066 	bpf_map_area_free(htab->buckets);
1067 	kfree(htab);
1068 }
1069 
1070 static void *sock_hash_lookup_sys(struct bpf_map *map, void *key)
1071 {
1072 	struct sock *sk;
1073 
1074 	if (map->value_size != sizeof(u64))
1075 		return ERR_PTR(-ENOSPC);
1076 
1077 	sk = __sock_hash_lookup_elem(map, key);
1078 	if (!sk)
1079 		return ERR_PTR(-ENOENT);
1080 
1081 	sock_gen_cookie(sk);
1082 	return &sk->sk_cookie;
1083 }
1084 
1085 static void *sock_hash_lookup(struct bpf_map *map, void *key)
1086 {
1087 	struct sock *sk;
1088 
1089 	sk = __sock_hash_lookup_elem(map, key);
1090 	if (!sk || !sk_fullsock(sk))
1091 		return NULL;
1092 	if (sk_is_refcounted(sk) && !refcount_inc_not_zero(&sk->sk_refcnt))
1093 		return NULL;
1094 	return sk;
1095 }
1096 
1097 static void sock_hash_release_progs(struct bpf_map *map)
1098 {
1099 	psock_progs_drop(&container_of(map, struct bpf_htab, map)->progs);
1100 }
1101 
1102 BPF_CALL_4(bpf_sock_hash_update, struct bpf_sock_ops_kern *, sops,
1103 	   struct bpf_map *, map, void *, key, u64, flags)
1104 {
1105 	WARN_ON_ONCE(!rcu_read_lock_held());
1106 
1107 	if (likely(sock_map_sk_is_suitable(sops->sk) &&
1108 		   sock_map_op_okay(sops)))
1109 		return sock_hash_update_common(map, key, sops->sk, flags);
1110 	return -EOPNOTSUPP;
1111 }
1112 
1113 const struct bpf_func_proto bpf_sock_hash_update_proto = {
1114 	.func		= bpf_sock_hash_update,
1115 	.gpl_only	= false,
1116 	.pkt_access	= true,
1117 	.ret_type	= RET_INTEGER,
1118 	.arg1_type	= ARG_PTR_TO_CTX,
1119 	.arg2_type	= ARG_CONST_MAP_PTR,
1120 	.arg3_type	= ARG_PTR_TO_MAP_KEY,
1121 	.arg4_type	= ARG_ANYTHING,
1122 };
1123 
1124 BPF_CALL_4(bpf_sk_redirect_hash, struct sk_buff *, skb,
1125 	   struct bpf_map *, map, void *, key, u64, flags)
1126 {
1127 	struct tcp_skb_cb *tcb = TCP_SKB_CB(skb);
1128 	struct sock *sk;
1129 
1130 	if (unlikely(flags & ~(BPF_F_INGRESS)))
1131 		return SK_DROP;
1132 
1133 	sk = __sock_hash_lookup_elem(map, key);
1134 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
1135 		return SK_DROP;
1136 
1137 	tcb->bpf.flags = flags;
1138 	tcb->bpf.sk_redir = sk;
1139 	return SK_PASS;
1140 }
1141 
1142 const struct bpf_func_proto bpf_sk_redirect_hash_proto = {
1143 	.func           = bpf_sk_redirect_hash,
1144 	.gpl_only       = false,
1145 	.ret_type       = RET_INTEGER,
1146 	.arg1_type	= ARG_PTR_TO_CTX,
1147 	.arg2_type      = ARG_CONST_MAP_PTR,
1148 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
1149 	.arg4_type      = ARG_ANYTHING,
1150 };
1151 
1152 BPF_CALL_4(bpf_msg_redirect_hash, struct sk_msg *, msg,
1153 	   struct bpf_map *, map, void *, key, u64, flags)
1154 {
1155 	struct sock *sk;
1156 
1157 	if (unlikely(flags & ~(BPF_F_INGRESS)))
1158 		return SK_DROP;
1159 
1160 	sk = __sock_hash_lookup_elem(map, key);
1161 	if (unlikely(!sk || !sock_map_redirect_allowed(sk)))
1162 		return SK_DROP;
1163 
1164 	msg->flags = flags;
1165 	msg->sk_redir = sk;
1166 	return SK_PASS;
1167 }
1168 
1169 const struct bpf_func_proto bpf_msg_redirect_hash_proto = {
1170 	.func           = bpf_msg_redirect_hash,
1171 	.gpl_only       = false,
1172 	.ret_type       = RET_INTEGER,
1173 	.arg1_type	= ARG_PTR_TO_CTX,
1174 	.arg2_type      = ARG_CONST_MAP_PTR,
1175 	.arg3_type      = ARG_PTR_TO_MAP_KEY,
1176 	.arg4_type      = ARG_ANYTHING,
1177 };
1178 
1179 const struct bpf_map_ops sock_hash_ops = {
1180 	.map_alloc		= sock_hash_alloc,
1181 	.map_free		= sock_hash_free,
1182 	.map_get_next_key	= sock_hash_get_next_key,
1183 	.map_update_elem	= sock_hash_update_elem,
1184 	.map_delete_elem	= sock_hash_delete_elem,
1185 	.map_lookup_elem	= sock_hash_lookup,
1186 	.map_lookup_elem_sys_only = sock_hash_lookup_sys,
1187 	.map_release_uref	= sock_hash_release_progs,
1188 	.map_check_btf		= map_check_no_btf,
1189 };
1190 
1191 static struct sk_psock_progs *sock_map_progs(struct bpf_map *map)
1192 {
1193 	switch (map->map_type) {
1194 	case BPF_MAP_TYPE_SOCKMAP:
1195 		return &container_of(map, struct bpf_stab, map)->progs;
1196 	case BPF_MAP_TYPE_SOCKHASH:
1197 		return &container_of(map, struct bpf_htab, map)->progs;
1198 	default:
1199 		break;
1200 	}
1201 
1202 	return NULL;
1203 }
1204 
1205 int sock_map_prog_update(struct bpf_map *map, struct bpf_prog *prog,
1206 			 u32 which)
1207 {
1208 	struct sk_psock_progs *progs = sock_map_progs(map);
1209 
1210 	if (!progs)
1211 		return -EOPNOTSUPP;
1212 
1213 	switch (which) {
1214 	case BPF_SK_MSG_VERDICT:
1215 		psock_set_prog(&progs->msg_parser, prog);
1216 		break;
1217 	case BPF_SK_SKB_STREAM_PARSER:
1218 		psock_set_prog(&progs->skb_parser, prog);
1219 		break;
1220 	case BPF_SK_SKB_STREAM_VERDICT:
1221 		psock_set_prog(&progs->skb_verdict, prog);
1222 		break;
1223 	default:
1224 		return -EOPNOTSUPP;
1225 	}
1226 
1227 	return 0;
1228 }
1229 
1230 static void sock_map_unlink(struct sock *sk, struct sk_psock_link *link)
1231 {
1232 	switch (link->map->map_type) {
1233 	case BPF_MAP_TYPE_SOCKMAP:
1234 		return sock_map_delete_from_link(link->map, sk,
1235 						 link->link_raw);
1236 	case BPF_MAP_TYPE_SOCKHASH:
1237 		return sock_hash_delete_from_link(link->map, sk,
1238 						  link->link_raw);
1239 	default:
1240 		break;
1241 	}
1242 }
1243 
1244 static void sock_map_remove_links(struct sock *sk, struct sk_psock *psock)
1245 {
1246 	struct sk_psock_link *link;
1247 
1248 	while ((link = sk_psock_link_pop(psock))) {
1249 		sock_map_unlink(sk, link);
1250 		sk_psock_free_link(link);
1251 	}
1252 }
1253 
1254 void sock_map_unhash(struct sock *sk)
1255 {
1256 	void (*saved_unhash)(struct sock *sk);
1257 	struct sk_psock *psock;
1258 
1259 	rcu_read_lock();
1260 	psock = sk_psock(sk);
1261 	if (unlikely(!psock)) {
1262 		rcu_read_unlock();
1263 		if (sk->sk_prot->unhash)
1264 			sk->sk_prot->unhash(sk);
1265 		return;
1266 	}
1267 
1268 	saved_unhash = psock->saved_unhash;
1269 	sock_map_remove_links(sk, psock);
1270 	rcu_read_unlock();
1271 	saved_unhash(sk);
1272 }
1273 
1274 void sock_map_close(struct sock *sk, long timeout)
1275 {
1276 	void (*saved_close)(struct sock *sk, long timeout);
1277 	struct sk_psock *psock;
1278 
1279 	lock_sock(sk);
1280 	rcu_read_lock();
1281 	psock = sk_psock(sk);
1282 	if (unlikely(!psock)) {
1283 		rcu_read_unlock();
1284 		release_sock(sk);
1285 		return sk->sk_prot->close(sk, timeout);
1286 	}
1287 
1288 	saved_close = psock->saved_close;
1289 	sock_map_remove_links(sk, psock);
1290 	rcu_read_unlock();
1291 	release_sock(sk);
1292 	saved_close(sk, timeout);
1293 }
1294