xref: /linux/net/mptcp/sockopt.c (revision 505d66d1abfb90853e24ab6cbdf83b611473d6fc)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Multipath TCP
3  *
4  * Copyright (c) 2021, Red Hat.
5  */
6 
7 #define pr_fmt(fmt) "MPTCP: " fmt
8 
9 #include <linux/kernel.h>
10 #include <linux/module.h>
11 #include <net/sock.h>
12 #include <net/protocol.h>
13 #include <net/tcp.h>
14 #include <net/mptcp.h>
15 #include "protocol.h"
16 
17 #define MIN_INFO_OPTLEN_SIZE		16
18 #define MIN_FULL_INFO_OPTLEN_SIZE	40
19 
20 static struct sock *__mptcp_tcp_fallback(struct mptcp_sock *msk)
21 {
22 	msk_owned_by_me(msk);
23 
24 	if (likely(!__mptcp_check_fallback(msk)))
25 		return NULL;
26 
27 	return msk->first;
28 }
29 
30 static u32 sockopt_seq_reset(const struct sock *sk)
31 {
32 	sock_owned_by_me(sk);
33 
34 	/* Highbits contain state.  Allows to distinguish sockopt_seq
35 	 * of listener and established:
36 	 * s0 = new_listener()
37 	 * sockopt(s0) - seq is 1
38 	 * s1 = accept(s0) - s1 inherits seq 1 if listener sk (s0)
39 	 * sockopt(s0) - seq increments to 2 on s0
40 	 * sockopt(s1) // seq increments to 2 on s1 (different option)
41 	 * new ssk completes join, inherits options from s0 // seq 2
42 	 * Needs sync from mptcp join logic, but ssk->seq == msk->seq
43 	 *
44 	 * Set High order bits to sk_state so ssk->seq == msk->seq test
45 	 * will fail.
46 	 */
47 
48 	return (u32)sk->sk_state << 24u;
49 }
50 
51 static void sockopt_seq_inc(struct mptcp_sock *msk)
52 {
53 	u32 seq = (msk->setsockopt_seq + 1) & 0x00ffffff;
54 
55 	msk->setsockopt_seq = sockopt_seq_reset((struct sock *)msk) + seq;
56 }
57 
58 static int mptcp_get_int_option(struct mptcp_sock *msk, sockptr_t optval,
59 				unsigned int optlen, int *val)
60 {
61 	if (optlen < sizeof(int))
62 		return -EINVAL;
63 
64 	if (copy_from_sockptr(val, optval, sizeof(*val)))
65 		return -EFAULT;
66 
67 	return 0;
68 }
69 
70 static void mptcp_sol_socket_sync_intval(struct mptcp_sock *msk, int optname, int val)
71 {
72 	struct mptcp_subflow_context *subflow;
73 	struct sock *sk = (struct sock *)msk;
74 
75 	lock_sock(sk);
76 	sockopt_seq_inc(msk);
77 
78 	mptcp_for_each_subflow(msk, subflow) {
79 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
80 		bool slow = lock_sock_fast(ssk);
81 
82 		switch (optname) {
83 		case SO_DEBUG:
84 			sock_valbool_flag(ssk, SOCK_DBG, !!val);
85 			break;
86 		case SO_KEEPALIVE:
87 			if (ssk->sk_prot->keepalive)
88 				ssk->sk_prot->keepalive(ssk, !!val);
89 			sock_valbool_flag(ssk, SOCK_KEEPOPEN, !!val);
90 			break;
91 		case SO_PRIORITY:
92 			WRITE_ONCE(ssk->sk_priority, val);
93 			break;
94 		case SO_SNDBUF:
95 		case SO_SNDBUFFORCE:
96 			ssk->sk_userlocks |= SOCK_SNDBUF_LOCK;
97 			WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
98 			mptcp_subflow_ctx(ssk)->cached_sndbuf = sk->sk_sndbuf;
99 			break;
100 		case SO_RCVBUF:
101 		case SO_RCVBUFFORCE:
102 			ssk->sk_userlocks |= SOCK_RCVBUF_LOCK;
103 			WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
104 			break;
105 		case SO_MARK:
106 			if (READ_ONCE(ssk->sk_mark) != sk->sk_mark) {
107 				WRITE_ONCE(ssk->sk_mark, sk->sk_mark);
108 				sk_dst_reset(ssk);
109 			}
110 			break;
111 		case SO_INCOMING_CPU:
112 			WRITE_ONCE(ssk->sk_incoming_cpu, val);
113 			break;
114 		}
115 
116 		subflow->setsockopt_seq = msk->setsockopt_seq;
117 		unlock_sock_fast(ssk, slow);
118 	}
119 
120 	release_sock(sk);
121 }
122 
123 static int mptcp_sol_socket_intval(struct mptcp_sock *msk, int optname, int val)
124 {
125 	sockptr_t optval = KERNEL_SOCKPTR(&val);
126 	struct sock *sk = (struct sock *)msk;
127 	int ret;
128 
129 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
130 			      optval, sizeof(val));
131 	if (ret)
132 		return ret;
133 
134 	mptcp_sol_socket_sync_intval(msk, optname, val);
135 	return 0;
136 }
137 
138 static void mptcp_so_incoming_cpu(struct mptcp_sock *msk, int val)
139 {
140 	struct sock *sk = (struct sock *)msk;
141 
142 	WRITE_ONCE(sk->sk_incoming_cpu, val);
143 
144 	mptcp_sol_socket_sync_intval(msk, SO_INCOMING_CPU, val);
145 }
146 
147 static int mptcp_setsockopt_sol_socket_tstamp(struct mptcp_sock *msk, int optname, int val)
148 {
149 	sockptr_t optval = KERNEL_SOCKPTR(&val);
150 	struct mptcp_subflow_context *subflow;
151 	struct sock *sk = (struct sock *)msk;
152 	int ret;
153 
154 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
155 			      optval, sizeof(val));
156 	if (ret)
157 		return ret;
158 
159 	lock_sock(sk);
160 	mptcp_for_each_subflow(msk, subflow) {
161 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
162 		bool slow = lock_sock_fast(ssk);
163 
164 		sock_set_timestamp(sk, optname, !!val);
165 		unlock_sock_fast(ssk, slow);
166 	}
167 
168 	release_sock(sk);
169 	return 0;
170 }
171 
172 static int mptcp_setsockopt_sol_socket_int(struct mptcp_sock *msk, int optname,
173 					   sockptr_t optval,
174 					   unsigned int optlen)
175 {
176 	int val, ret;
177 
178 	ret = mptcp_get_int_option(msk, optval, optlen, &val);
179 	if (ret)
180 		return ret;
181 
182 	switch (optname) {
183 	case SO_KEEPALIVE:
184 	case SO_DEBUG:
185 	case SO_MARK:
186 	case SO_PRIORITY:
187 	case SO_SNDBUF:
188 	case SO_SNDBUFFORCE:
189 	case SO_RCVBUF:
190 	case SO_RCVBUFFORCE:
191 		return mptcp_sol_socket_intval(msk, optname, val);
192 	case SO_INCOMING_CPU:
193 		mptcp_so_incoming_cpu(msk, val);
194 		return 0;
195 	case SO_TIMESTAMP_OLD:
196 	case SO_TIMESTAMP_NEW:
197 	case SO_TIMESTAMPNS_OLD:
198 	case SO_TIMESTAMPNS_NEW:
199 		return mptcp_setsockopt_sol_socket_tstamp(msk, optname, val);
200 	}
201 
202 	return -ENOPROTOOPT;
203 }
204 
205 static int mptcp_setsockopt_sol_socket_timestamping(struct mptcp_sock *msk,
206 						    int optname,
207 						    sockptr_t optval,
208 						    unsigned int optlen)
209 {
210 	struct mptcp_subflow_context *subflow;
211 	struct sock *sk = (struct sock *)msk;
212 	struct so_timestamping timestamping;
213 	int ret;
214 
215 	if (optlen == sizeof(timestamping)) {
216 		if (copy_from_sockptr(&timestamping, optval,
217 				      sizeof(timestamping)))
218 			return -EFAULT;
219 	} else if (optlen == sizeof(int)) {
220 		memset(&timestamping, 0, sizeof(timestamping));
221 
222 		if (copy_from_sockptr(&timestamping.flags, optval, sizeof(int)))
223 			return -EFAULT;
224 	} else {
225 		return -EINVAL;
226 	}
227 
228 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
229 			      KERNEL_SOCKPTR(&timestamping),
230 			      sizeof(timestamping));
231 	if (ret)
232 		return ret;
233 
234 	lock_sock(sk);
235 
236 	mptcp_for_each_subflow(msk, subflow) {
237 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
238 		bool slow = lock_sock_fast(ssk);
239 
240 		sock_set_timestamping(sk, optname, timestamping);
241 		unlock_sock_fast(ssk, slow);
242 	}
243 
244 	release_sock(sk);
245 
246 	return 0;
247 }
248 
249 static int mptcp_setsockopt_sol_socket_linger(struct mptcp_sock *msk, sockptr_t optval,
250 					      unsigned int optlen)
251 {
252 	struct mptcp_subflow_context *subflow;
253 	struct sock *sk = (struct sock *)msk;
254 	struct linger ling;
255 	sockptr_t kopt;
256 	int ret;
257 
258 	if (optlen < sizeof(ling))
259 		return -EINVAL;
260 
261 	if (copy_from_sockptr(&ling, optval, sizeof(ling)))
262 		return -EFAULT;
263 
264 	kopt = KERNEL_SOCKPTR(&ling);
265 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, SO_LINGER, kopt, sizeof(ling));
266 	if (ret)
267 		return ret;
268 
269 	lock_sock(sk);
270 	sockopt_seq_inc(msk);
271 	mptcp_for_each_subflow(msk, subflow) {
272 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
273 		bool slow = lock_sock_fast(ssk);
274 
275 		if (!ling.l_onoff) {
276 			sock_reset_flag(ssk, SOCK_LINGER);
277 		} else {
278 			ssk->sk_lingertime = sk->sk_lingertime;
279 			sock_set_flag(ssk, SOCK_LINGER);
280 		}
281 
282 		subflow->setsockopt_seq = msk->setsockopt_seq;
283 		unlock_sock_fast(ssk, slow);
284 	}
285 
286 	release_sock(sk);
287 	return 0;
288 }
289 
290 static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,
291 				       sockptr_t optval, unsigned int optlen)
292 {
293 	struct sock *sk = (struct sock *)msk;
294 	struct sock *ssk;
295 	int ret;
296 
297 	switch (optname) {
298 	case SO_REUSEPORT:
299 	case SO_REUSEADDR:
300 	case SO_BINDTODEVICE:
301 	case SO_BINDTOIFINDEX:
302 		lock_sock(sk);
303 		ssk = __mptcp_nmpc_sk(msk);
304 		if (IS_ERR(ssk)) {
305 			release_sock(sk);
306 			return PTR_ERR(ssk);
307 		}
308 
309 		ret = sk_setsockopt(ssk, SOL_SOCKET, optname, optval, optlen);
310 		if (ret == 0) {
311 			if (optname == SO_REUSEPORT)
312 				sk->sk_reuseport = ssk->sk_reuseport;
313 			else if (optname == SO_REUSEADDR)
314 				sk->sk_reuse = ssk->sk_reuse;
315 			else if (optname == SO_BINDTODEVICE)
316 				sk->sk_bound_dev_if = ssk->sk_bound_dev_if;
317 			else if (optname == SO_BINDTOIFINDEX)
318 				sk->sk_bound_dev_if = ssk->sk_bound_dev_if;
319 		}
320 		release_sock(sk);
321 		return ret;
322 	case SO_KEEPALIVE:
323 	case SO_PRIORITY:
324 	case SO_SNDBUF:
325 	case SO_SNDBUFFORCE:
326 	case SO_RCVBUF:
327 	case SO_RCVBUFFORCE:
328 	case SO_MARK:
329 	case SO_INCOMING_CPU:
330 	case SO_DEBUG:
331 	case SO_TIMESTAMP_OLD:
332 	case SO_TIMESTAMP_NEW:
333 	case SO_TIMESTAMPNS_OLD:
334 	case SO_TIMESTAMPNS_NEW:
335 		return mptcp_setsockopt_sol_socket_int(msk, optname, optval,
336 						       optlen);
337 	case SO_TIMESTAMPING_OLD:
338 	case SO_TIMESTAMPING_NEW:
339 		return mptcp_setsockopt_sol_socket_timestamping(msk, optname,
340 								optval, optlen);
341 	case SO_LINGER:
342 		return mptcp_setsockopt_sol_socket_linger(msk, optval, optlen);
343 	case SO_RCVLOWAT:
344 	case SO_RCVTIMEO_OLD:
345 	case SO_RCVTIMEO_NEW:
346 	case SO_SNDTIMEO_OLD:
347 	case SO_SNDTIMEO_NEW:
348 	case SO_BUSY_POLL:
349 	case SO_PREFER_BUSY_POLL:
350 	case SO_BUSY_POLL_BUDGET:
351 		/* No need to copy: only relevant for msk */
352 		return sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname, optval, optlen);
353 	case SO_NO_CHECK:
354 	case SO_DONTROUTE:
355 	case SO_BROADCAST:
356 	case SO_BSDCOMPAT:
357 	case SO_PASSCRED:
358 	case SO_PASSPIDFD:
359 	case SO_PASSSEC:
360 	case SO_RXQ_OVFL:
361 	case SO_WIFI_STATUS:
362 	case SO_NOFCS:
363 	case SO_SELECT_ERR_QUEUE:
364 		return 0;
365 	}
366 
367 	/* SO_OOBINLINE is not supported, let's avoid the related mess
368 	 * SO_ATTACH_FILTER, SO_ATTACH_BPF, SO_ATTACH_REUSEPORT_CBPF,
369 	 * SO_DETACH_REUSEPORT_BPF, SO_DETACH_FILTER, SO_LOCK_FILTER,
370 	 * we must be careful with subflows
371 	 *
372 	 * SO_ATTACH_REUSEPORT_EBPF is not supported, at it checks
373 	 * explicitly the sk_protocol field
374 	 *
375 	 * SO_PEEK_OFF is unsupported, as it is for plain TCP
376 	 * SO_MAX_PACING_RATE is unsupported, we must be careful with subflows
377 	 * SO_CNX_ADVICE is currently unsupported, could possibly be relevant,
378 	 * but likely needs careful design
379 	 *
380 	 * SO_ZEROCOPY is currently unsupported, TODO in sndmsg
381 	 * SO_TXTIME is currently unsupported
382 	 */
383 
384 	return -EOPNOTSUPP;
385 }
386 
387 static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,
388 			       sockptr_t optval, unsigned int optlen)
389 {
390 	struct sock *sk = (struct sock *)msk;
391 	int ret = -EOPNOTSUPP;
392 	struct sock *ssk;
393 
394 	switch (optname) {
395 	case IPV6_V6ONLY:
396 	case IPV6_TRANSPARENT:
397 	case IPV6_FREEBIND:
398 		lock_sock(sk);
399 		ssk = __mptcp_nmpc_sk(msk);
400 		if (IS_ERR(ssk)) {
401 			release_sock(sk);
402 			return PTR_ERR(ssk);
403 		}
404 
405 		ret = tcp_setsockopt(ssk, SOL_IPV6, optname, optval, optlen);
406 		if (ret != 0) {
407 			release_sock(sk);
408 			return ret;
409 		}
410 
411 		sockopt_seq_inc(msk);
412 
413 		switch (optname) {
414 		case IPV6_V6ONLY:
415 			sk->sk_ipv6only = ssk->sk_ipv6only;
416 			break;
417 		case IPV6_TRANSPARENT:
418 			inet_assign_bit(TRANSPARENT, sk,
419 					inet_test_bit(TRANSPARENT, ssk));
420 			break;
421 		case IPV6_FREEBIND:
422 			inet_assign_bit(FREEBIND, sk,
423 					inet_test_bit(FREEBIND, ssk));
424 			break;
425 		}
426 
427 		release_sock(sk);
428 		break;
429 	}
430 
431 	return ret;
432 }
433 
434 static bool mptcp_supported_sockopt(int level, int optname)
435 {
436 	if (level == SOL_IP) {
437 		switch (optname) {
438 		/* should work fine */
439 		case IP_FREEBIND:
440 		case IP_TRANSPARENT:
441 		case IP_BIND_ADDRESS_NO_PORT:
442 		case IP_LOCAL_PORT_RANGE:
443 
444 		/* the following are control cmsg related */
445 		case IP_PKTINFO:
446 		case IP_RECVTTL:
447 		case IP_RECVTOS:
448 		case IP_RECVOPTS:
449 		case IP_RETOPTS:
450 		case IP_PASSSEC:
451 		case IP_RECVORIGDSTADDR:
452 		case IP_CHECKSUM:
453 		case IP_RECVFRAGSIZE:
454 
455 		/* common stuff that need some love */
456 		case IP_TOS:
457 		case IP_TTL:
458 		case IP_MTU_DISCOVER:
459 		case IP_RECVERR:
460 
461 		/* possibly less common may deserve some love */
462 		case IP_MINTTL:
463 
464 		/* the following is apparently a no-op for plain TCP */
465 		case IP_RECVERR_RFC4884:
466 			return true;
467 		}
468 
469 		/* IP_OPTIONS is not supported, needs subflow care */
470 		/* IP_HDRINCL, IP_NODEFRAG are not supported, RAW specific */
471 		/* IP_MULTICAST_TTL, IP_MULTICAST_LOOP, IP_UNICAST_IF,
472 		 * IP_ADD_MEMBERSHIP, IP_ADD_SOURCE_MEMBERSHIP, IP_DROP_MEMBERSHIP,
473 		 * IP_DROP_SOURCE_MEMBERSHIP, IP_BLOCK_SOURCE, IP_UNBLOCK_SOURCE,
474 		 * MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP MCAST_JOIN_SOURCE_GROUP,
475 		 * MCAST_LEAVE_SOURCE_GROUP, MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE,
476 		 * MCAST_MSFILTER, IP_MULTICAST_ALL are not supported, better not deal
477 		 * with mcast stuff
478 		 */
479 		/* IP_IPSEC_POLICY, IP_XFRM_POLICY are nut supported, unrelated here */
480 		return false;
481 	}
482 	if (level == SOL_IPV6) {
483 		switch (optname) {
484 		case IPV6_V6ONLY:
485 
486 		/* the following are control cmsg related */
487 		case IPV6_RECVPKTINFO:
488 		case IPV6_2292PKTINFO:
489 		case IPV6_RECVHOPLIMIT:
490 		case IPV6_2292HOPLIMIT:
491 		case IPV6_RECVRTHDR:
492 		case IPV6_2292RTHDR:
493 		case IPV6_RECVHOPOPTS:
494 		case IPV6_2292HOPOPTS:
495 		case IPV6_RECVDSTOPTS:
496 		case IPV6_2292DSTOPTS:
497 		case IPV6_RECVTCLASS:
498 		case IPV6_FLOWINFO:
499 		case IPV6_RECVPATHMTU:
500 		case IPV6_RECVORIGDSTADDR:
501 		case IPV6_RECVFRAGSIZE:
502 
503 		/* the following ones need some love but are quite common */
504 		case IPV6_TCLASS:
505 		case IPV6_TRANSPARENT:
506 		case IPV6_FREEBIND:
507 		case IPV6_PKTINFO:
508 		case IPV6_2292PKTOPTIONS:
509 		case IPV6_UNICAST_HOPS:
510 		case IPV6_MTU_DISCOVER:
511 		case IPV6_MTU:
512 		case IPV6_RECVERR:
513 		case IPV6_FLOWINFO_SEND:
514 		case IPV6_FLOWLABEL_MGR:
515 		case IPV6_MINHOPCOUNT:
516 		case IPV6_DONTFRAG:
517 		case IPV6_AUTOFLOWLABEL:
518 
519 		/* the following one is a no-op for plain TCP */
520 		case IPV6_RECVERR_RFC4884:
521 			return true;
522 		}
523 
524 		/* IPV6_HOPOPTS, IPV6_RTHDRDSTOPTS, IPV6_RTHDR, IPV6_DSTOPTS are
525 		 * not supported
526 		 */
527 		/* IPV6_MULTICAST_HOPS, IPV6_MULTICAST_LOOP, IPV6_UNICAST_IF,
528 		 * IPV6_MULTICAST_IF, IPV6_ADDRFORM,
529 		 * IPV6_ADD_MEMBERSHIP, IPV6_DROP_MEMBERSHIP, IPV6_JOIN_ANYCAST,
530 		 * IPV6_LEAVE_ANYCAST, IPV6_MULTICAST_ALL, MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP,
531 		 * MCAST_JOIN_SOURCE_GROUP, MCAST_LEAVE_SOURCE_GROUP,
532 		 * MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE, MCAST_MSFILTER
533 		 * are not supported better not deal with mcast
534 		 */
535 		/* IPV6_ROUTER_ALERT, IPV6_ROUTER_ALERT_ISOLATE are not supported, since are evil */
536 
537 		/* IPV6_IPSEC_POLICY, IPV6_XFRM_POLICY are not supported */
538 		/* IPV6_ADDR_PREFERENCES is not supported, we must be careful with subflows */
539 		return false;
540 	}
541 	if (level == SOL_TCP) {
542 		switch (optname) {
543 		/* the following are no-op or should work just fine */
544 		case TCP_THIN_DUPACK:
545 		case TCP_DEFER_ACCEPT:
546 
547 		/* the following need some love */
548 		case TCP_MAXSEG:
549 		case TCP_NODELAY:
550 		case TCP_THIN_LINEAR_TIMEOUTS:
551 		case TCP_CONGESTION:
552 		case TCP_CORK:
553 		case TCP_KEEPIDLE:
554 		case TCP_KEEPINTVL:
555 		case TCP_KEEPCNT:
556 		case TCP_SYNCNT:
557 		case TCP_SAVE_SYN:
558 		case TCP_LINGER2:
559 		case TCP_WINDOW_CLAMP:
560 		case TCP_QUICKACK:
561 		case TCP_USER_TIMEOUT:
562 		case TCP_TIMESTAMP:
563 		case TCP_NOTSENT_LOWAT:
564 		case TCP_TX_DELAY:
565 		case TCP_INQ:
566 		case TCP_FASTOPEN:
567 		case TCP_FASTOPEN_CONNECT:
568 		case TCP_FASTOPEN_KEY:
569 		case TCP_FASTOPEN_NO_COOKIE:
570 			return true;
571 		}
572 
573 		/* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
574 
575 		/* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
576 		 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
577 		 */
578 	}
579 	return false;
580 }
581 
582 static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
583 					       unsigned int optlen)
584 {
585 	struct mptcp_subflow_context *subflow;
586 	struct sock *sk = (struct sock *)msk;
587 	char name[TCP_CA_NAME_MAX];
588 	bool cap_net_admin;
589 	int ret;
590 
591 	if (optlen < 1)
592 		return -EINVAL;
593 
594 	ret = strncpy_from_sockptr(name, optval,
595 				   min_t(long, TCP_CA_NAME_MAX - 1, optlen));
596 	if (ret < 0)
597 		return -EFAULT;
598 
599 	name[ret] = 0;
600 
601 	cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
602 
603 	ret = 0;
604 	lock_sock(sk);
605 	sockopt_seq_inc(msk);
606 	mptcp_for_each_subflow(msk, subflow) {
607 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
608 		int err;
609 
610 		lock_sock(ssk);
611 		err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
612 		if (err < 0 && ret == 0)
613 			ret = err;
614 		subflow->setsockopt_seq = msk->setsockopt_seq;
615 		release_sock(ssk);
616 	}
617 
618 	if (ret == 0)
619 		strscpy(msk->ca_name, name, sizeof(msk->ca_name));
620 
621 	release_sock(sk);
622 	return ret;
623 }
624 
625 static int __mptcp_setsockopt_set_val(struct mptcp_sock *msk, int max,
626 				      int (*set_val)(struct sock *, int),
627 				      int *msk_val, int val)
628 {
629 	struct mptcp_subflow_context *subflow;
630 	int err = 0;
631 
632 	mptcp_for_each_subflow(msk, subflow) {
633 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
634 		int ret;
635 
636 		lock_sock(ssk);
637 		ret = set_val(ssk, val);
638 		err = err ? : ret;
639 		release_sock(ssk);
640 	}
641 
642 	if (!err) {
643 		*msk_val = val;
644 		sockopt_seq_inc(msk);
645 	}
646 
647 	return err;
648 }
649 
650 static int __mptcp_setsockopt_sol_tcp_cork(struct mptcp_sock *msk, int val)
651 {
652 	struct mptcp_subflow_context *subflow;
653 	struct sock *sk = (struct sock *)msk;
654 
655 	sockopt_seq_inc(msk);
656 	msk->cork = !!val;
657 	mptcp_for_each_subflow(msk, subflow) {
658 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
659 
660 		lock_sock(ssk);
661 		__tcp_sock_set_cork(ssk, !!val);
662 		release_sock(ssk);
663 	}
664 	if (!val)
665 		mptcp_check_and_set_pending(sk);
666 
667 	return 0;
668 }
669 
670 static int __mptcp_setsockopt_sol_tcp_nodelay(struct mptcp_sock *msk, int val)
671 {
672 	struct mptcp_subflow_context *subflow;
673 	struct sock *sk = (struct sock *)msk;
674 
675 	sockopt_seq_inc(msk);
676 	msk->nodelay = !!val;
677 	mptcp_for_each_subflow(msk, subflow) {
678 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
679 
680 		lock_sock(ssk);
681 		__tcp_sock_set_nodelay(ssk, !!val);
682 		release_sock(ssk);
683 	}
684 	if (val)
685 		mptcp_check_and_set_pending(sk);
686 	return 0;
687 }
688 
689 static int mptcp_setsockopt_sol_ip_set(struct mptcp_sock *msk, int optname,
690 				       sockptr_t optval, unsigned int optlen)
691 {
692 	struct sock *sk = (struct sock *)msk;
693 	struct sock *ssk;
694 	int err;
695 
696 	err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
697 	if (err != 0)
698 		return err;
699 
700 	lock_sock(sk);
701 
702 	ssk = __mptcp_nmpc_sk(msk);
703 	if (IS_ERR(ssk)) {
704 		release_sock(sk);
705 		return PTR_ERR(ssk);
706 	}
707 
708 	switch (optname) {
709 	case IP_FREEBIND:
710 		inet_assign_bit(FREEBIND, ssk, inet_test_bit(FREEBIND, sk));
711 		break;
712 	case IP_TRANSPARENT:
713 		inet_assign_bit(TRANSPARENT, ssk,
714 				inet_test_bit(TRANSPARENT, sk));
715 		break;
716 	case IP_BIND_ADDRESS_NO_PORT:
717 		inet_assign_bit(BIND_ADDRESS_NO_PORT, ssk,
718 				inet_test_bit(BIND_ADDRESS_NO_PORT, sk));
719 		break;
720 	case IP_LOCAL_PORT_RANGE:
721 		WRITE_ONCE(inet_sk(ssk)->local_port_range,
722 			   READ_ONCE(inet_sk(sk)->local_port_range));
723 		break;
724 	default:
725 		release_sock(sk);
726 		WARN_ON_ONCE(1);
727 		return -EOPNOTSUPP;
728 	}
729 
730 	sockopt_seq_inc(msk);
731 	release_sock(sk);
732 	return 0;
733 }
734 
735 static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,
736 				       sockptr_t optval, unsigned int optlen)
737 {
738 	struct mptcp_subflow_context *subflow;
739 	struct sock *sk = (struct sock *)msk;
740 	int err, val;
741 
742 	err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
743 
744 	if (err != 0)
745 		return err;
746 
747 	lock_sock(sk);
748 	sockopt_seq_inc(msk);
749 	val = READ_ONCE(inet_sk(sk)->tos);
750 	mptcp_for_each_subflow(msk, subflow) {
751 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
752 		bool slow;
753 
754 		slow = lock_sock_fast(ssk);
755 		__ip_sock_set_tos(ssk, val);
756 		unlock_sock_fast(ssk, slow);
757 	}
758 	release_sock(sk);
759 
760 	return 0;
761 }
762 
763 static int mptcp_setsockopt_v4(struct mptcp_sock *msk, int optname,
764 			       sockptr_t optval, unsigned int optlen)
765 {
766 	switch (optname) {
767 	case IP_FREEBIND:
768 	case IP_TRANSPARENT:
769 	case IP_BIND_ADDRESS_NO_PORT:
770 	case IP_LOCAL_PORT_RANGE:
771 		return mptcp_setsockopt_sol_ip_set(msk, optname, optval, optlen);
772 	case IP_TOS:
773 		return mptcp_setsockopt_v4_set_tos(msk, optname, optval, optlen);
774 	}
775 
776 	return -EOPNOTSUPP;
777 }
778 
779 static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
780 					  sockptr_t optval, unsigned int optlen)
781 {
782 	struct sock *sk = (struct sock *)msk;
783 	struct sock *ssk;
784 	int ret;
785 
786 	/* Limit to first subflow, before the connection establishment */
787 	lock_sock(sk);
788 	ssk = __mptcp_nmpc_sk(msk);
789 	if (IS_ERR(ssk)) {
790 		ret = PTR_ERR(ssk);
791 		goto unlock;
792 	}
793 
794 	ret = tcp_setsockopt(ssk, level, optname, optval, optlen);
795 
796 unlock:
797 	release_sock(sk);
798 	return ret;
799 }
800 
801 static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
802 				    sockptr_t optval, unsigned int optlen)
803 {
804 	struct sock *sk = (void *)msk;
805 	int ret, val;
806 
807 	switch (optname) {
808 	case TCP_ULP:
809 		return -EOPNOTSUPP;
810 	case TCP_CONGESTION:
811 		return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
812 	case TCP_DEFER_ACCEPT:
813 		/* See tcp.c: TCP_DEFER_ACCEPT does not fail */
814 		mptcp_setsockopt_first_sf_only(msk, SOL_TCP, optname, optval, optlen);
815 		return 0;
816 	case TCP_FASTOPEN:
817 	case TCP_FASTOPEN_CONNECT:
818 	case TCP_FASTOPEN_KEY:
819 	case TCP_FASTOPEN_NO_COOKIE:
820 		return mptcp_setsockopt_first_sf_only(msk, SOL_TCP, optname,
821 						      optval, optlen);
822 	}
823 
824 	ret = mptcp_get_int_option(msk, optval, optlen, &val);
825 	if (ret)
826 		return ret;
827 
828 	lock_sock(sk);
829 	switch (optname) {
830 	case TCP_INQ:
831 		if (val < 0 || val > 1)
832 			ret = -EINVAL;
833 		else
834 			msk->recvmsg_inq = !!val;
835 		break;
836 	case TCP_NOTSENT_LOWAT:
837 		WRITE_ONCE(msk->notsent_lowat, val);
838 		mptcp_write_space(sk);
839 		break;
840 	case TCP_CORK:
841 		ret = __mptcp_setsockopt_sol_tcp_cork(msk, val);
842 		break;
843 	case TCP_NODELAY:
844 		ret = __mptcp_setsockopt_sol_tcp_nodelay(msk, val);
845 		break;
846 	case TCP_KEEPIDLE:
847 		ret = __mptcp_setsockopt_set_val(msk, MAX_TCP_KEEPIDLE,
848 						 &tcp_sock_set_keepidle_locked,
849 						 &msk->keepalive_idle, val);
850 		break;
851 	case TCP_KEEPINTVL:
852 		ret = __mptcp_setsockopt_set_val(msk, MAX_TCP_KEEPINTVL,
853 						 &tcp_sock_set_keepintvl,
854 						 &msk->keepalive_intvl, val);
855 		break;
856 	case TCP_KEEPCNT:
857 		ret = __mptcp_setsockopt_set_val(msk, MAX_TCP_KEEPCNT,
858 						 &tcp_sock_set_keepcnt,
859 						 &msk->keepalive_cnt,
860 						 val);
861 		break;
862 	default:
863 		ret = -ENOPROTOOPT;
864 	}
865 
866 	release_sock(sk);
867 	return ret;
868 }
869 
870 int mptcp_setsockopt(struct sock *sk, int level, int optname,
871 		     sockptr_t optval, unsigned int optlen)
872 {
873 	struct mptcp_sock *msk = mptcp_sk(sk);
874 	struct sock *ssk;
875 
876 	pr_debug("msk=%p", msk);
877 
878 	if (level == SOL_SOCKET)
879 		return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
880 
881 	if (!mptcp_supported_sockopt(level, optname))
882 		return -ENOPROTOOPT;
883 
884 	/* @@ the meaning of setsockopt() when the socket is connected and
885 	 * there are multiple subflows is not yet defined. It is up to the
886 	 * MPTCP-level socket to configure the subflows until the subflow
887 	 * is in TCP fallback, when TCP socket options are passed through
888 	 * to the one remaining subflow.
889 	 */
890 	lock_sock(sk);
891 	ssk = __mptcp_tcp_fallback(msk);
892 	release_sock(sk);
893 	if (ssk)
894 		return tcp_setsockopt(ssk, level, optname, optval, optlen);
895 
896 	if (level == SOL_IP)
897 		return mptcp_setsockopt_v4(msk, optname, optval, optlen);
898 
899 	if (level == SOL_IPV6)
900 		return mptcp_setsockopt_v6(msk, optname, optval, optlen);
901 
902 	if (level == SOL_TCP)
903 		return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
904 
905 	return -EOPNOTSUPP;
906 }
907 
908 static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
909 					  char __user *optval, int __user *optlen)
910 {
911 	struct sock *sk = (struct sock *)msk;
912 	struct sock *ssk;
913 	int ret;
914 
915 	lock_sock(sk);
916 	ssk = msk->first;
917 	if (ssk) {
918 		ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
919 		goto out;
920 	}
921 
922 	ssk = __mptcp_nmpc_sk(msk);
923 	if (IS_ERR(ssk)) {
924 		ret = PTR_ERR(ssk);
925 		goto out;
926 	}
927 
928 	ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
929 
930 out:
931 	release_sock(sk);
932 	return ret;
933 }
934 
935 void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
936 {
937 	struct sock *sk = (struct sock *)msk;
938 	u32 flags = 0;
939 	bool slow;
940 	u32 now;
941 
942 	memset(info, 0, sizeof(*info));
943 
944 	info->mptcpi_subflows = READ_ONCE(msk->pm.subflows);
945 	info->mptcpi_add_addr_signal = READ_ONCE(msk->pm.add_addr_signaled);
946 	info->mptcpi_add_addr_accepted = READ_ONCE(msk->pm.add_addr_accepted);
947 	info->mptcpi_local_addr_used = READ_ONCE(msk->pm.local_addr_used);
948 
949 	if (inet_sk_state_load(sk) == TCP_LISTEN)
950 		return;
951 
952 	/* The following limits only make sense for the in-kernel PM */
953 	if (mptcp_pm_is_kernel(msk)) {
954 		info->mptcpi_subflows_max =
955 			mptcp_pm_get_subflows_max(msk);
956 		info->mptcpi_add_addr_signal_max =
957 			mptcp_pm_get_add_addr_signal_max(msk);
958 		info->mptcpi_add_addr_accepted_max =
959 			mptcp_pm_get_add_addr_accept_max(msk);
960 		info->mptcpi_local_addr_max =
961 			mptcp_pm_get_local_addr_max(msk);
962 	}
963 
964 	if (__mptcp_check_fallback(msk))
965 		flags |= MPTCP_INFO_FLAG_FALLBACK;
966 	if (READ_ONCE(msk->can_ack))
967 		flags |= MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED;
968 	info->mptcpi_flags = flags;
969 
970 	slow = lock_sock_fast(sk);
971 	info->mptcpi_csum_enabled = READ_ONCE(msk->csum_enabled);
972 	info->mptcpi_token = msk->token;
973 	info->mptcpi_write_seq = msk->write_seq;
974 	info->mptcpi_retransmits = inet_csk(sk)->icsk_retransmits;
975 	info->mptcpi_bytes_sent = msk->bytes_sent;
976 	info->mptcpi_bytes_received = msk->bytes_received;
977 	info->mptcpi_bytes_retrans = msk->bytes_retrans;
978 	info->mptcpi_subflows_total = info->mptcpi_subflows +
979 		__mptcp_has_initial_subflow(msk);
980 	now = tcp_jiffies32;
981 	info->mptcpi_last_data_sent = jiffies_to_msecs(now - msk->last_data_sent);
982 	info->mptcpi_last_data_recv = jiffies_to_msecs(now - msk->last_data_recv);
983 	unlock_sock_fast(sk, slow);
984 
985 	mptcp_data_lock(sk);
986 	info->mptcpi_last_ack_recv = jiffies_to_msecs(now - msk->last_ack_recv);
987 	info->mptcpi_snd_una = msk->snd_una;
988 	info->mptcpi_rcv_nxt = msk->ack_seq;
989 	info->mptcpi_bytes_acked = msk->bytes_acked;
990 	mptcp_data_unlock(sk);
991 }
992 EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
993 
994 static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
995 {
996 	struct mptcp_info m_info;
997 	int len;
998 
999 	if (get_user(len, optlen))
1000 		return -EFAULT;
1001 
1002 	/* When used only to check if a fallback to TCP happened. */
1003 	if (len == 0)
1004 		return 0;
1005 
1006 	len = min_t(unsigned int, len, sizeof(struct mptcp_info));
1007 
1008 	mptcp_diag_fill_info(msk, &m_info);
1009 
1010 	if (put_user(len, optlen))
1011 		return -EFAULT;
1012 
1013 	if (copy_to_user(optval, &m_info, len))
1014 		return -EFAULT;
1015 
1016 	return 0;
1017 }
1018 
1019 static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
1020 				  char __user *optval,
1021 				  u32 copied,
1022 				  int __user *optlen)
1023 {
1024 	u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
1025 
1026 	if (copied)
1027 		copied += sfd->size_subflow_data;
1028 	else
1029 		copied = copylen;
1030 
1031 	if (put_user(copied, optlen))
1032 		return -EFAULT;
1033 
1034 	if (copy_to_user(optval, sfd, copylen))
1035 		return -EFAULT;
1036 
1037 	return 0;
1038 }
1039 
1040 static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
1041 				  char __user *optval,
1042 				  int __user *optlen)
1043 {
1044 	int len, copylen;
1045 
1046 	if (get_user(len, optlen))
1047 		return -EFAULT;
1048 
1049 	/* if mptcp_subflow_data size is changed, need to adjust
1050 	 * this function to deal with programs using old version.
1051 	 */
1052 	BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
1053 
1054 	if (len < MIN_INFO_OPTLEN_SIZE)
1055 		return -EINVAL;
1056 
1057 	memset(sfd, 0, sizeof(*sfd));
1058 
1059 	copylen = min_t(unsigned int, len, sizeof(*sfd));
1060 	if (copy_from_user(sfd, optval, copylen))
1061 		return -EFAULT;
1062 
1063 	/* size_subflow_data is u32, but len is signed */
1064 	if (sfd->size_subflow_data > INT_MAX ||
1065 	    sfd->size_user > INT_MAX)
1066 		return -EINVAL;
1067 
1068 	if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
1069 	    sfd->size_subflow_data > len)
1070 		return -EINVAL;
1071 
1072 	if (sfd->num_subflows || sfd->size_kernel)
1073 		return -EINVAL;
1074 
1075 	return len - sfd->size_subflow_data;
1076 }
1077 
1078 static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
1079 				    int __user *optlen)
1080 {
1081 	struct mptcp_subflow_context *subflow;
1082 	struct sock *sk = (struct sock *)msk;
1083 	unsigned int sfcount = 0, copied = 0;
1084 	struct mptcp_subflow_data sfd;
1085 	char __user *infoptr;
1086 	int len;
1087 
1088 	len = mptcp_get_subflow_data(&sfd, optval, optlen);
1089 	if (len < 0)
1090 		return len;
1091 
1092 	sfd.size_kernel = sizeof(struct tcp_info);
1093 	sfd.size_user = min_t(unsigned int, sfd.size_user,
1094 			      sizeof(struct tcp_info));
1095 
1096 	infoptr = optval + sfd.size_subflow_data;
1097 
1098 	lock_sock(sk);
1099 
1100 	mptcp_for_each_subflow(msk, subflow) {
1101 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1102 
1103 		++sfcount;
1104 
1105 		if (len && len >= sfd.size_user) {
1106 			struct tcp_info info;
1107 
1108 			tcp_get_info(ssk, &info);
1109 
1110 			if (copy_to_user(infoptr, &info, sfd.size_user)) {
1111 				release_sock(sk);
1112 				return -EFAULT;
1113 			}
1114 
1115 			infoptr += sfd.size_user;
1116 			copied += sfd.size_user;
1117 			len -= sfd.size_user;
1118 		}
1119 	}
1120 
1121 	release_sock(sk);
1122 
1123 	sfd.num_subflows = sfcount;
1124 
1125 	if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1126 		return -EFAULT;
1127 
1128 	return 0;
1129 }
1130 
1131 static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
1132 {
1133 	const struct inet_sock *inet = inet_sk(sk);
1134 
1135 	memset(a, 0, sizeof(*a));
1136 
1137 	if (sk->sk_family == AF_INET) {
1138 		a->sin_local.sin_family = AF_INET;
1139 		a->sin_local.sin_port = inet->inet_sport;
1140 		a->sin_local.sin_addr.s_addr = inet->inet_rcv_saddr;
1141 
1142 		if (!a->sin_local.sin_addr.s_addr)
1143 			a->sin_local.sin_addr.s_addr = inet->inet_saddr;
1144 
1145 		a->sin_remote.sin_family = AF_INET;
1146 		a->sin_remote.sin_port = inet->inet_dport;
1147 		a->sin_remote.sin_addr.s_addr = inet->inet_daddr;
1148 #if IS_ENABLED(CONFIG_IPV6)
1149 	} else if (sk->sk_family == AF_INET6) {
1150 		const struct ipv6_pinfo *np = inet6_sk(sk);
1151 
1152 		if (WARN_ON_ONCE(!np))
1153 			return;
1154 
1155 		a->sin6_local.sin6_family = AF_INET6;
1156 		a->sin6_local.sin6_port = inet->inet_sport;
1157 
1158 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
1159 			a->sin6_local.sin6_addr = np->saddr;
1160 		else
1161 			a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
1162 
1163 		a->sin6_remote.sin6_family = AF_INET6;
1164 		a->sin6_remote.sin6_port = inet->inet_dport;
1165 		a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
1166 #endif
1167 	}
1168 }
1169 
1170 static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
1171 					  int __user *optlen)
1172 {
1173 	struct mptcp_subflow_context *subflow;
1174 	struct sock *sk = (struct sock *)msk;
1175 	unsigned int sfcount = 0, copied = 0;
1176 	struct mptcp_subflow_data sfd;
1177 	char __user *addrptr;
1178 	int len;
1179 
1180 	len = mptcp_get_subflow_data(&sfd, optval, optlen);
1181 	if (len < 0)
1182 		return len;
1183 
1184 	sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
1185 	sfd.size_user = min_t(unsigned int, sfd.size_user,
1186 			      sizeof(struct mptcp_subflow_addrs));
1187 
1188 	addrptr = optval + sfd.size_subflow_data;
1189 
1190 	lock_sock(sk);
1191 
1192 	mptcp_for_each_subflow(msk, subflow) {
1193 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1194 
1195 		++sfcount;
1196 
1197 		if (len && len >= sfd.size_user) {
1198 			struct mptcp_subflow_addrs a;
1199 
1200 			mptcp_get_sub_addrs(ssk, &a);
1201 
1202 			if (copy_to_user(addrptr, &a, sfd.size_user)) {
1203 				release_sock(sk);
1204 				return -EFAULT;
1205 			}
1206 
1207 			addrptr += sfd.size_user;
1208 			copied += sfd.size_user;
1209 			len -= sfd.size_user;
1210 		}
1211 	}
1212 
1213 	release_sock(sk);
1214 
1215 	sfd.num_subflows = sfcount;
1216 
1217 	if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1218 		return -EFAULT;
1219 
1220 	return 0;
1221 }
1222 
1223 static int mptcp_get_full_info(struct mptcp_full_info *mfi,
1224 			       char __user *optval,
1225 			       int __user *optlen)
1226 {
1227 	int len;
1228 
1229 	BUILD_BUG_ON(offsetof(struct mptcp_full_info, mptcp_info) !=
1230 		     MIN_FULL_INFO_OPTLEN_SIZE);
1231 
1232 	if (get_user(len, optlen))
1233 		return -EFAULT;
1234 
1235 	if (len < MIN_FULL_INFO_OPTLEN_SIZE)
1236 		return -EINVAL;
1237 
1238 	memset(mfi, 0, sizeof(*mfi));
1239 	if (copy_from_user(mfi, optval, MIN_FULL_INFO_OPTLEN_SIZE))
1240 		return -EFAULT;
1241 
1242 	if (mfi->size_tcpinfo_kernel ||
1243 	    mfi->size_sfinfo_kernel ||
1244 	    mfi->num_subflows)
1245 		return -EINVAL;
1246 
1247 	if (mfi->size_sfinfo_user > INT_MAX ||
1248 	    mfi->size_tcpinfo_user > INT_MAX)
1249 		return -EINVAL;
1250 
1251 	return len - MIN_FULL_INFO_OPTLEN_SIZE;
1252 }
1253 
1254 static int mptcp_put_full_info(struct mptcp_full_info *mfi,
1255 			       char __user *optval,
1256 			       u32 copylen,
1257 			       int __user *optlen)
1258 {
1259 	copylen += MIN_FULL_INFO_OPTLEN_SIZE;
1260 	if (put_user(copylen, optlen))
1261 		return -EFAULT;
1262 
1263 	if (copy_to_user(optval, mfi, copylen))
1264 		return -EFAULT;
1265 	return 0;
1266 }
1267 
1268 static int mptcp_getsockopt_full_info(struct mptcp_sock *msk, char __user *optval,
1269 				      int __user *optlen)
1270 {
1271 	unsigned int sfcount = 0, copylen = 0;
1272 	struct mptcp_subflow_context *subflow;
1273 	struct sock *sk = (struct sock *)msk;
1274 	void __user *tcpinfoptr, *sfinfoptr;
1275 	struct mptcp_full_info mfi;
1276 	int len;
1277 
1278 	len = mptcp_get_full_info(&mfi, optval, optlen);
1279 	if (len < 0)
1280 		return len;
1281 
1282 	/* don't bother filling the mptcp info if there is not enough
1283 	 * user-space-provided storage
1284 	 */
1285 	if (len > 0) {
1286 		mptcp_diag_fill_info(msk, &mfi.mptcp_info);
1287 		copylen += min_t(unsigned int, len, sizeof(struct mptcp_info));
1288 	}
1289 
1290 	mfi.size_tcpinfo_kernel = sizeof(struct tcp_info);
1291 	mfi.size_tcpinfo_user = min_t(unsigned int, mfi.size_tcpinfo_user,
1292 				      sizeof(struct tcp_info));
1293 	sfinfoptr = u64_to_user_ptr(mfi.subflow_info);
1294 	mfi.size_sfinfo_kernel = sizeof(struct mptcp_subflow_info);
1295 	mfi.size_sfinfo_user = min_t(unsigned int, mfi.size_sfinfo_user,
1296 				     sizeof(struct mptcp_subflow_info));
1297 	tcpinfoptr = u64_to_user_ptr(mfi.tcp_info);
1298 
1299 	lock_sock(sk);
1300 	mptcp_for_each_subflow(msk, subflow) {
1301 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1302 		struct mptcp_subflow_info sfinfo;
1303 		struct tcp_info tcp_info;
1304 
1305 		if (sfcount++ >= mfi.size_arrays_user)
1306 			continue;
1307 
1308 		/* fetch addr/tcp_info only if the user space buffers
1309 		 * are wide enough
1310 		 */
1311 		memset(&sfinfo, 0, sizeof(sfinfo));
1312 		sfinfo.id = subflow->subflow_id;
1313 		if (mfi.size_sfinfo_user >
1314 		    offsetof(struct mptcp_subflow_info, addrs))
1315 			mptcp_get_sub_addrs(ssk, &sfinfo.addrs);
1316 		if (copy_to_user(sfinfoptr, &sfinfo, mfi.size_sfinfo_user))
1317 			goto fail_release;
1318 
1319 		if (mfi.size_tcpinfo_user) {
1320 			tcp_get_info(ssk, &tcp_info);
1321 			if (copy_to_user(tcpinfoptr, &tcp_info,
1322 					 mfi.size_tcpinfo_user))
1323 				goto fail_release;
1324 		}
1325 
1326 		tcpinfoptr += mfi.size_tcpinfo_user;
1327 		sfinfoptr += mfi.size_sfinfo_user;
1328 	}
1329 	release_sock(sk);
1330 
1331 	mfi.num_subflows = sfcount;
1332 	if (mptcp_put_full_info(&mfi, optval, copylen, optlen))
1333 		return -EFAULT;
1334 
1335 	return 0;
1336 
1337 fail_release:
1338 	release_sock(sk);
1339 	return -EFAULT;
1340 }
1341 
1342 static int mptcp_put_int_option(struct mptcp_sock *msk, char __user *optval,
1343 				int __user *optlen, int val)
1344 {
1345 	int len;
1346 
1347 	if (get_user(len, optlen))
1348 		return -EFAULT;
1349 	if (len < 0)
1350 		return -EINVAL;
1351 
1352 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1353 		unsigned char ucval = (unsigned char)val;
1354 
1355 		len = 1;
1356 		if (put_user(len, optlen))
1357 			return -EFAULT;
1358 		if (copy_to_user(optval, &ucval, 1))
1359 			return -EFAULT;
1360 	} else {
1361 		len = min_t(unsigned int, len, sizeof(int));
1362 		if (put_user(len, optlen))
1363 			return -EFAULT;
1364 		if (copy_to_user(optval, &val, len))
1365 			return -EFAULT;
1366 	}
1367 
1368 	return 0;
1369 }
1370 
1371 static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
1372 				    char __user *optval, int __user *optlen)
1373 {
1374 	struct sock *sk = (void *)msk;
1375 
1376 	switch (optname) {
1377 	case TCP_ULP:
1378 	case TCP_CONGESTION:
1379 	case TCP_INFO:
1380 	case TCP_CC_INFO:
1381 	case TCP_DEFER_ACCEPT:
1382 	case TCP_FASTOPEN:
1383 	case TCP_FASTOPEN_CONNECT:
1384 	case TCP_FASTOPEN_KEY:
1385 	case TCP_FASTOPEN_NO_COOKIE:
1386 		return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
1387 						      optval, optlen);
1388 	case TCP_INQ:
1389 		return mptcp_put_int_option(msk, optval, optlen, msk->recvmsg_inq);
1390 	case TCP_CORK:
1391 		return mptcp_put_int_option(msk, optval, optlen, msk->cork);
1392 	case TCP_NODELAY:
1393 		return mptcp_put_int_option(msk, optval, optlen, msk->nodelay);
1394 	case TCP_KEEPIDLE:
1395 		return mptcp_put_int_option(msk, optval, optlen,
1396 					    msk->keepalive_idle ? :
1397 					    READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_keepalive_time) / HZ);
1398 	case TCP_KEEPINTVL:
1399 		return mptcp_put_int_option(msk, optval, optlen,
1400 					    msk->keepalive_intvl ? :
1401 					    READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_keepalive_intvl) / HZ);
1402 	case TCP_KEEPCNT:
1403 		return mptcp_put_int_option(msk, optval, optlen,
1404 					    msk->keepalive_cnt ? :
1405 					    READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_keepalive_probes));
1406 	case TCP_NOTSENT_LOWAT:
1407 		return mptcp_put_int_option(msk, optval, optlen, msk->notsent_lowat);
1408 	case TCP_IS_MPTCP:
1409 		return mptcp_put_int_option(msk, optval, optlen, 1);
1410 	}
1411 	return -EOPNOTSUPP;
1412 }
1413 
1414 static int mptcp_getsockopt_v4(struct mptcp_sock *msk, int optname,
1415 			       char __user *optval, int __user *optlen)
1416 {
1417 	struct sock *sk = (void *)msk;
1418 
1419 	switch (optname) {
1420 	case IP_TOS:
1421 		return mptcp_put_int_option(msk, optval, optlen, READ_ONCE(inet_sk(sk)->tos));
1422 	case IP_BIND_ADDRESS_NO_PORT:
1423 		return mptcp_put_int_option(msk, optval, optlen,
1424 				inet_test_bit(BIND_ADDRESS_NO_PORT, sk));
1425 	case IP_LOCAL_PORT_RANGE:
1426 		return mptcp_put_int_option(msk, optval, optlen,
1427 				READ_ONCE(inet_sk(sk)->local_port_range));
1428 	}
1429 
1430 	return -EOPNOTSUPP;
1431 }
1432 
1433 static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
1434 				      char __user *optval, int __user *optlen)
1435 {
1436 	switch (optname) {
1437 	case MPTCP_INFO:
1438 		return mptcp_getsockopt_info(msk, optval, optlen);
1439 	case MPTCP_FULL_INFO:
1440 		return mptcp_getsockopt_full_info(msk, optval, optlen);
1441 	case MPTCP_TCPINFO:
1442 		return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
1443 	case MPTCP_SUBFLOW_ADDRS:
1444 		return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
1445 	}
1446 
1447 	return -EOPNOTSUPP;
1448 }
1449 
1450 int mptcp_getsockopt(struct sock *sk, int level, int optname,
1451 		     char __user *optval, int __user *option)
1452 {
1453 	struct mptcp_sock *msk = mptcp_sk(sk);
1454 	struct sock *ssk;
1455 
1456 	pr_debug("msk=%p", msk);
1457 
1458 	/* @@ the meaning of setsockopt() when the socket is connected and
1459 	 * there are multiple subflows is not yet defined. It is up to the
1460 	 * MPTCP-level socket to configure the subflows until the subflow
1461 	 * is in TCP fallback, when socket options are passed through
1462 	 * to the one remaining subflow.
1463 	 */
1464 	lock_sock(sk);
1465 	ssk = __mptcp_tcp_fallback(msk);
1466 	release_sock(sk);
1467 	if (ssk)
1468 		return tcp_getsockopt(ssk, level, optname, optval, option);
1469 
1470 	if (level == SOL_IP)
1471 		return mptcp_getsockopt_v4(msk, optname, optval, option);
1472 	if (level == SOL_TCP)
1473 		return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
1474 	if (level == SOL_MPTCP)
1475 		return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
1476 	return -EOPNOTSUPP;
1477 }
1478 
1479 static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
1480 {
1481 	static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
1482 	struct sock *sk = (struct sock *)msk;
1483 
1484 	if (ssk->sk_prot->keepalive) {
1485 		if (sock_flag(sk, SOCK_KEEPOPEN))
1486 			ssk->sk_prot->keepalive(ssk, 1);
1487 		else
1488 			ssk->sk_prot->keepalive(ssk, 0);
1489 	}
1490 
1491 	ssk->sk_priority = sk->sk_priority;
1492 	ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1493 	ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
1494 	ssk->sk_ipv6only = sk->sk_ipv6only;
1495 	__ip_sock_set_tos(ssk, inet_sk(sk)->tos);
1496 
1497 	if (sk->sk_userlocks & tx_rx_locks) {
1498 		ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1499 		if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) {
1500 			WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1501 			mptcp_subflow_ctx(ssk)->cached_sndbuf = sk->sk_sndbuf;
1502 		}
1503 		if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1504 			WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1505 	}
1506 
1507 	if (sock_flag(sk, SOCK_LINGER)) {
1508 		ssk->sk_lingertime = sk->sk_lingertime;
1509 		sock_set_flag(ssk, SOCK_LINGER);
1510 	} else {
1511 		sock_reset_flag(ssk, SOCK_LINGER);
1512 	}
1513 
1514 	if (sk->sk_mark != ssk->sk_mark) {
1515 		ssk->sk_mark = sk->sk_mark;
1516 		sk_dst_reset(ssk);
1517 	}
1518 
1519 	sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1520 
1521 	if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
1522 		tcp_set_congestion_control(ssk, msk->ca_name, false, true);
1523 	__tcp_sock_set_cork(ssk, !!msk->cork);
1524 	__tcp_sock_set_nodelay(ssk, !!msk->nodelay);
1525 	tcp_sock_set_keepidle_locked(ssk, msk->keepalive_idle);
1526 	tcp_sock_set_keepintvl(ssk, msk->keepalive_intvl);
1527 	tcp_sock_set_keepcnt(ssk, msk->keepalive_cnt);
1528 
1529 	inet_assign_bit(TRANSPARENT, ssk, inet_test_bit(TRANSPARENT, sk));
1530 	inet_assign_bit(FREEBIND, ssk, inet_test_bit(FREEBIND, sk));
1531 	inet_assign_bit(BIND_ADDRESS_NO_PORT, ssk, inet_test_bit(BIND_ADDRESS_NO_PORT, sk));
1532 	WRITE_ONCE(inet_sk(ssk)->local_port_range, READ_ONCE(inet_sk(sk)->local_port_range));
1533 }
1534 
1535 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk)
1536 {
1537 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1538 
1539 	msk_owned_by_me(msk);
1540 
1541 	ssk->sk_rcvlowat = 0;
1542 
1543 	/* subflows must ignore any latency-related settings: will not affect
1544 	 * the user-space - only the msk is relevant - but will foul the
1545 	 * mptcp scheduler
1546 	 */
1547 	tcp_sk(ssk)->notsent_lowat = UINT_MAX;
1548 
1549 	if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1550 		sync_socket_options(msk, ssk);
1551 
1552 		subflow->setsockopt_seq = msk->setsockopt_seq;
1553 	}
1554 }
1555 
1556 /* unfortunately this is different enough from the tcp version so
1557  * that we can't factor it out
1558  */
1559 int mptcp_set_rcvlowat(struct sock *sk, int val)
1560 {
1561 	struct mptcp_subflow_context *subflow;
1562 	int space, cap;
1563 
1564 	/* bpf can land here with a wrong sk type */
1565 	if (sk->sk_protocol == IPPROTO_TCP)
1566 		return -EINVAL;
1567 
1568 	if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1569 		cap = sk->sk_rcvbuf >> 1;
1570 	else
1571 		cap = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_rmem[2]) >> 1;
1572 	val = min(val, cap);
1573 	WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);
1574 
1575 	/* Check if we need to signal EPOLLIN right now */
1576 	if (mptcp_epollin_ready(sk))
1577 		sk->sk_data_ready(sk);
1578 
1579 	if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1580 		return 0;
1581 
1582 	space = __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, val);
1583 	if (space <= sk->sk_rcvbuf)
1584 		return 0;
1585 
1586 	/* propagate the rcvbuf changes to all the subflows */
1587 	WRITE_ONCE(sk->sk_rcvbuf, space);
1588 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
1589 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1590 		bool slow;
1591 
1592 		slow = lock_sock_fast(ssk);
1593 		WRITE_ONCE(ssk->sk_rcvbuf, space);
1594 		WRITE_ONCE(tcp_sk(ssk)->window_clamp, val);
1595 		unlock_sock_fast(ssk, slow);
1596 	}
1597 	return 0;
1598 }
1599