xref: /linux/net/mptcp/sockopt.c (revision e7b2b108cdeab76a7e7324459e50b0c1214c0386)
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 		mptcp_sol_socket_sync_intval(msk, optname, val);
185 		return 0;
186 	case SO_DEBUG:
187 	case SO_MARK:
188 	case SO_PRIORITY:
189 	case SO_SNDBUF:
190 	case SO_SNDBUFFORCE:
191 	case SO_RCVBUF:
192 	case SO_RCVBUFFORCE:
193 		return mptcp_sol_socket_intval(msk, optname, val);
194 	case SO_INCOMING_CPU:
195 		mptcp_so_incoming_cpu(msk, val);
196 		return 0;
197 	case SO_TIMESTAMP_OLD:
198 	case SO_TIMESTAMP_NEW:
199 	case SO_TIMESTAMPNS_OLD:
200 	case SO_TIMESTAMPNS_NEW:
201 		return mptcp_setsockopt_sol_socket_tstamp(msk, optname, val);
202 	}
203 
204 	return -ENOPROTOOPT;
205 }
206 
207 static int mptcp_setsockopt_sol_socket_timestamping(struct mptcp_sock *msk,
208 						    int optname,
209 						    sockptr_t optval,
210 						    unsigned int optlen)
211 {
212 	struct mptcp_subflow_context *subflow;
213 	struct sock *sk = (struct sock *)msk;
214 	struct so_timestamping timestamping;
215 	int ret;
216 
217 	if (optlen == sizeof(timestamping)) {
218 		if (copy_from_sockptr(&timestamping, optval,
219 				      sizeof(timestamping)))
220 			return -EFAULT;
221 	} else if (optlen == sizeof(int)) {
222 		memset(&timestamping, 0, sizeof(timestamping));
223 
224 		if (copy_from_sockptr(&timestamping.flags, optval, sizeof(int)))
225 			return -EFAULT;
226 	} else {
227 		return -EINVAL;
228 	}
229 
230 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
231 			      KERNEL_SOCKPTR(&timestamping),
232 			      sizeof(timestamping));
233 	if (ret)
234 		return ret;
235 
236 	lock_sock(sk);
237 
238 	mptcp_for_each_subflow(msk, subflow) {
239 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
240 		bool slow = lock_sock_fast(ssk);
241 
242 		sock_set_timestamping(sk, optname, timestamping);
243 		unlock_sock_fast(ssk, slow);
244 	}
245 
246 	release_sock(sk);
247 
248 	return 0;
249 }
250 
251 static int mptcp_setsockopt_sol_socket_linger(struct mptcp_sock *msk, sockptr_t optval,
252 					      unsigned int optlen)
253 {
254 	struct mptcp_subflow_context *subflow;
255 	struct sock *sk = (struct sock *)msk;
256 	struct linger ling;
257 	sockptr_t kopt;
258 	int ret;
259 
260 	if (optlen < sizeof(ling))
261 		return -EINVAL;
262 
263 	if (copy_from_sockptr(&ling, optval, sizeof(ling)))
264 		return -EFAULT;
265 
266 	kopt = KERNEL_SOCKPTR(&ling);
267 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, SO_LINGER, kopt, sizeof(ling));
268 	if (ret)
269 		return ret;
270 
271 	lock_sock(sk);
272 	sockopt_seq_inc(msk);
273 	mptcp_for_each_subflow(msk, subflow) {
274 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
275 		bool slow = lock_sock_fast(ssk);
276 
277 		if (!ling.l_onoff) {
278 			sock_reset_flag(ssk, SOCK_LINGER);
279 		} else {
280 			ssk->sk_lingertime = sk->sk_lingertime;
281 			sock_set_flag(ssk, SOCK_LINGER);
282 		}
283 
284 		subflow->setsockopt_seq = msk->setsockopt_seq;
285 		unlock_sock_fast(ssk, slow);
286 	}
287 
288 	release_sock(sk);
289 	return 0;
290 }
291 
292 static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,
293 				       sockptr_t optval, unsigned int optlen)
294 {
295 	struct sock *sk = (struct sock *)msk;
296 	struct sock *ssk;
297 	int ret;
298 
299 	switch (optname) {
300 	case SO_REUSEPORT:
301 	case SO_REUSEADDR:
302 	case SO_BINDTODEVICE:
303 	case SO_BINDTOIFINDEX:
304 		lock_sock(sk);
305 		ssk = __mptcp_nmpc_sk(msk);
306 		if (IS_ERR(ssk)) {
307 			release_sock(sk);
308 			return PTR_ERR(ssk);
309 		}
310 
311 		ret = sk_setsockopt(ssk, SOL_SOCKET, optname, optval, optlen);
312 		if (ret == 0) {
313 			if (optname == SO_REUSEPORT)
314 				sk->sk_reuseport = ssk->sk_reuseport;
315 			else if (optname == SO_REUSEADDR)
316 				sk->sk_reuse = ssk->sk_reuse;
317 			else if (optname == SO_BINDTODEVICE)
318 				sk->sk_bound_dev_if = ssk->sk_bound_dev_if;
319 			else if (optname == SO_BINDTOIFINDEX)
320 				sk->sk_bound_dev_if = ssk->sk_bound_dev_if;
321 		}
322 		release_sock(sk);
323 		return ret;
324 	case SO_KEEPALIVE:
325 	case SO_PRIORITY:
326 	case SO_SNDBUF:
327 	case SO_SNDBUFFORCE:
328 	case SO_RCVBUF:
329 	case SO_RCVBUFFORCE:
330 	case SO_MARK:
331 	case SO_INCOMING_CPU:
332 	case SO_DEBUG:
333 	case SO_TIMESTAMP_OLD:
334 	case SO_TIMESTAMP_NEW:
335 	case SO_TIMESTAMPNS_OLD:
336 	case SO_TIMESTAMPNS_NEW:
337 		return mptcp_setsockopt_sol_socket_int(msk, optname, optval,
338 						       optlen);
339 	case SO_TIMESTAMPING_OLD:
340 	case SO_TIMESTAMPING_NEW:
341 		return mptcp_setsockopt_sol_socket_timestamping(msk, optname,
342 								optval, optlen);
343 	case SO_LINGER:
344 		return mptcp_setsockopt_sol_socket_linger(msk, optval, optlen);
345 	case SO_RCVLOWAT:
346 	case SO_RCVTIMEO_OLD:
347 	case SO_RCVTIMEO_NEW:
348 	case SO_SNDTIMEO_OLD:
349 	case SO_SNDTIMEO_NEW:
350 	case SO_BUSY_POLL:
351 	case SO_PREFER_BUSY_POLL:
352 	case SO_BUSY_POLL_BUDGET:
353 		/* No need to copy: only relevant for msk */
354 		return sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname, optval, optlen);
355 	case SO_NO_CHECK:
356 	case SO_DONTROUTE:
357 	case SO_BROADCAST:
358 	case SO_BSDCOMPAT:
359 	case SO_PASSCRED:
360 	case SO_PASSPIDFD:
361 	case SO_PASSSEC:
362 	case SO_RXQ_OVFL:
363 	case SO_WIFI_STATUS:
364 	case SO_NOFCS:
365 	case SO_SELECT_ERR_QUEUE:
366 		return 0;
367 	}
368 
369 	/* SO_OOBINLINE is not supported, let's avoid the related mess
370 	 * SO_ATTACH_FILTER, SO_ATTACH_BPF, SO_ATTACH_REUSEPORT_CBPF,
371 	 * SO_DETACH_REUSEPORT_BPF, SO_DETACH_FILTER, SO_LOCK_FILTER,
372 	 * we must be careful with subflows
373 	 *
374 	 * SO_ATTACH_REUSEPORT_EBPF is not supported, at it checks
375 	 * explicitly the sk_protocol field
376 	 *
377 	 * SO_PEEK_OFF is unsupported, as it is for plain TCP
378 	 * SO_MAX_PACING_RATE is unsupported, we must be careful with subflows
379 	 * SO_CNX_ADVICE is currently unsupported, could possibly be relevant,
380 	 * but likely needs careful design
381 	 *
382 	 * SO_ZEROCOPY is currently unsupported, TODO in sndmsg
383 	 * SO_TXTIME is currently unsupported
384 	 */
385 
386 	return -EOPNOTSUPP;
387 }
388 
389 static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,
390 			       sockptr_t optval, unsigned int optlen)
391 {
392 	struct sock *sk = (struct sock *)msk;
393 	int ret = -EOPNOTSUPP;
394 	struct sock *ssk;
395 
396 	switch (optname) {
397 	case IPV6_V6ONLY:
398 	case IPV6_TRANSPARENT:
399 	case IPV6_FREEBIND:
400 		lock_sock(sk);
401 		ssk = __mptcp_nmpc_sk(msk);
402 		if (IS_ERR(ssk)) {
403 			release_sock(sk);
404 			return PTR_ERR(ssk);
405 		}
406 
407 		ret = tcp_setsockopt(ssk, SOL_IPV6, optname, optval, optlen);
408 		if (ret != 0) {
409 			release_sock(sk);
410 			return ret;
411 		}
412 
413 		sockopt_seq_inc(msk);
414 
415 		switch (optname) {
416 		case IPV6_V6ONLY:
417 			sk->sk_ipv6only = ssk->sk_ipv6only;
418 			break;
419 		case IPV6_TRANSPARENT:
420 			inet_assign_bit(TRANSPARENT, sk,
421 					inet_test_bit(TRANSPARENT, ssk));
422 			break;
423 		case IPV6_FREEBIND:
424 			inet_assign_bit(FREEBIND, sk,
425 					inet_test_bit(FREEBIND, ssk));
426 			break;
427 		}
428 
429 		release_sock(sk);
430 		break;
431 	}
432 
433 	return ret;
434 }
435 
436 static bool mptcp_supported_sockopt(int level, int optname)
437 {
438 	if (level == SOL_IP) {
439 		switch (optname) {
440 		/* should work fine */
441 		case IP_FREEBIND:
442 		case IP_TRANSPARENT:
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_BIND_ADDRESS_NO_PORT:
459 		case IP_MTU_DISCOVER:
460 		case IP_RECVERR:
461 
462 		/* possibly less common may deserve some love */
463 		case IP_MINTTL:
464 
465 		/* the following is apparently a no-op for plain TCP */
466 		case IP_RECVERR_RFC4884:
467 			return true;
468 		}
469 
470 		/* IP_OPTIONS is not supported, needs subflow care */
471 		/* IP_HDRINCL, IP_NODEFRAG are not supported, RAW specific */
472 		/* IP_MULTICAST_TTL, IP_MULTICAST_LOOP, IP_UNICAST_IF,
473 		 * IP_ADD_MEMBERSHIP, IP_ADD_SOURCE_MEMBERSHIP, IP_DROP_MEMBERSHIP,
474 		 * IP_DROP_SOURCE_MEMBERSHIP, IP_BLOCK_SOURCE, IP_UNBLOCK_SOURCE,
475 		 * MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP MCAST_JOIN_SOURCE_GROUP,
476 		 * MCAST_LEAVE_SOURCE_GROUP, MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE,
477 		 * MCAST_MSFILTER, IP_MULTICAST_ALL are not supported, better not deal
478 		 * with mcast stuff
479 		 */
480 		/* IP_IPSEC_POLICY, IP_XFRM_POLICY are nut supported, unrelated here */
481 		return false;
482 	}
483 	if (level == SOL_IPV6) {
484 		switch (optname) {
485 		case IPV6_V6ONLY:
486 
487 		/* the following are control cmsg related */
488 		case IPV6_RECVPKTINFO:
489 		case IPV6_2292PKTINFO:
490 		case IPV6_RECVHOPLIMIT:
491 		case IPV6_2292HOPLIMIT:
492 		case IPV6_RECVRTHDR:
493 		case IPV6_2292RTHDR:
494 		case IPV6_RECVHOPOPTS:
495 		case IPV6_2292HOPOPTS:
496 		case IPV6_RECVDSTOPTS:
497 		case IPV6_2292DSTOPTS:
498 		case IPV6_RECVTCLASS:
499 		case IPV6_FLOWINFO:
500 		case IPV6_RECVPATHMTU:
501 		case IPV6_RECVORIGDSTADDR:
502 		case IPV6_RECVFRAGSIZE:
503 
504 		/* the following ones need some love but are quite common */
505 		case IPV6_TCLASS:
506 		case IPV6_TRANSPARENT:
507 		case IPV6_FREEBIND:
508 		case IPV6_PKTINFO:
509 		case IPV6_2292PKTOPTIONS:
510 		case IPV6_UNICAST_HOPS:
511 		case IPV6_MTU_DISCOVER:
512 		case IPV6_MTU:
513 		case IPV6_RECVERR:
514 		case IPV6_FLOWINFO_SEND:
515 		case IPV6_FLOWLABEL_MGR:
516 		case IPV6_MINHOPCOUNT:
517 		case IPV6_DONTFRAG:
518 		case IPV6_AUTOFLOWLABEL:
519 
520 		/* the following one is a no-op for plain TCP */
521 		case IPV6_RECVERR_RFC4884:
522 			return true;
523 		}
524 
525 		/* IPV6_HOPOPTS, IPV6_RTHDRDSTOPTS, IPV6_RTHDR, IPV6_DSTOPTS are
526 		 * not supported
527 		 */
528 		/* IPV6_MULTICAST_HOPS, IPV6_MULTICAST_LOOP, IPV6_UNICAST_IF,
529 		 * IPV6_MULTICAST_IF, IPV6_ADDRFORM,
530 		 * IPV6_ADD_MEMBERSHIP, IPV6_DROP_MEMBERSHIP, IPV6_JOIN_ANYCAST,
531 		 * IPV6_LEAVE_ANYCAST, IPV6_MULTICAST_ALL, MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP,
532 		 * MCAST_JOIN_SOURCE_GROUP, MCAST_LEAVE_SOURCE_GROUP,
533 		 * MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE, MCAST_MSFILTER
534 		 * are not supported better not deal with mcast
535 		 */
536 		/* IPV6_ROUTER_ALERT, IPV6_ROUTER_ALERT_ISOLATE are not supported, since are evil */
537 
538 		/* IPV6_IPSEC_POLICY, IPV6_XFRM_POLICY are not supported */
539 		/* IPV6_ADDR_PREFERENCES is not supported, we must be careful with subflows */
540 		return false;
541 	}
542 	if (level == SOL_TCP) {
543 		switch (optname) {
544 		/* the following are no-op or should work just fine */
545 		case TCP_THIN_DUPACK:
546 		case TCP_DEFER_ACCEPT:
547 
548 		/* the following need some love */
549 		case TCP_MAXSEG:
550 		case TCP_NODELAY:
551 		case TCP_THIN_LINEAR_TIMEOUTS:
552 		case TCP_CONGESTION:
553 		case TCP_CORK:
554 		case TCP_KEEPIDLE:
555 		case TCP_KEEPINTVL:
556 		case TCP_KEEPCNT:
557 		case TCP_SYNCNT:
558 		case TCP_SAVE_SYN:
559 		case TCP_LINGER2:
560 		case TCP_WINDOW_CLAMP:
561 		case TCP_QUICKACK:
562 		case TCP_USER_TIMEOUT:
563 		case TCP_TIMESTAMP:
564 		case TCP_NOTSENT_LOWAT:
565 		case TCP_TX_DELAY:
566 		case TCP_INQ:
567 		case TCP_FASTOPEN:
568 		case TCP_FASTOPEN_CONNECT:
569 		case TCP_FASTOPEN_KEY:
570 		case TCP_FASTOPEN_NO_COOKIE:
571 			return true;
572 		}
573 
574 		/* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
575 
576 		/* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
577 		 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
578 		 */
579 	}
580 	return false;
581 }
582 
583 static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
584 					       unsigned int optlen)
585 {
586 	struct mptcp_subflow_context *subflow;
587 	struct sock *sk = (struct sock *)msk;
588 	char name[TCP_CA_NAME_MAX];
589 	bool cap_net_admin;
590 	int ret;
591 
592 	if (optlen < 1)
593 		return -EINVAL;
594 
595 	ret = strncpy_from_sockptr(name, optval,
596 				   min_t(long, TCP_CA_NAME_MAX - 1, optlen));
597 	if (ret < 0)
598 		return -EFAULT;
599 
600 	name[ret] = 0;
601 
602 	cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
603 
604 	ret = 0;
605 	lock_sock(sk);
606 	sockopt_seq_inc(msk);
607 	mptcp_for_each_subflow(msk, subflow) {
608 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
609 		int err;
610 
611 		lock_sock(ssk);
612 		err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
613 		if (err < 0 && ret == 0)
614 			ret = err;
615 		subflow->setsockopt_seq = msk->setsockopt_seq;
616 		release_sock(ssk);
617 	}
618 
619 	if (ret == 0)
620 		strcpy(msk->ca_name, name);
621 
622 	release_sock(sk);
623 	return ret;
624 }
625 
626 static int mptcp_setsockopt_sol_tcp_cork(struct mptcp_sock *msk, sockptr_t optval,
627 					 unsigned int optlen)
628 {
629 	struct mptcp_subflow_context *subflow;
630 	struct sock *sk = (struct sock *)msk;
631 	int val;
632 
633 	if (optlen < sizeof(int))
634 		return -EINVAL;
635 
636 	if (copy_from_sockptr(&val, optval, sizeof(val)))
637 		return -EFAULT;
638 
639 	lock_sock(sk);
640 	sockopt_seq_inc(msk);
641 	msk->cork = !!val;
642 	mptcp_for_each_subflow(msk, subflow) {
643 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
644 
645 		lock_sock(ssk);
646 		__tcp_sock_set_cork(ssk, !!val);
647 		release_sock(ssk);
648 	}
649 	if (!val)
650 		mptcp_check_and_set_pending(sk);
651 	release_sock(sk);
652 
653 	return 0;
654 }
655 
656 static int mptcp_setsockopt_sol_tcp_nodelay(struct mptcp_sock *msk, sockptr_t optval,
657 					    unsigned int optlen)
658 {
659 	struct mptcp_subflow_context *subflow;
660 	struct sock *sk = (struct sock *)msk;
661 	int val;
662 
663 	if (optlen < sizeof(int))
664 		return -EINVAL;
665 
666 	if (copy_from_sockptr(&val, optval, sizeof(val)))
667 		return -EFAULT;
668 
669 	lock_sock(sk);
670 	sockopt_seq_inc(msk);
671 	msk->nodelay = !!val;
672 	mptcp_for_each_subflow(msk, subflow) {
673 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
674 
675 		lock_sock(ssk);
676 		__tcp_sock_set_nodelay(ssk, !!val);
677 		release_sock(ssk);
678 	}
679 	if (val)
680 		mptcp_check_and_set_pending(sk);
681 	release_sock(sk);
682 
683 	return 0;
684 }
685 
686 static int mptcp_setsockopt_sol_ip_set_transparent(struct mptcp_sock *msk, int optname,
687 						   sockptr_t optval, unsigned int optlen)
688 {
689 	struct sock *sk = (struct sock *)msk;
690 	struct sock *ssk;
691 	int err;
692 
693 	err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
694 	if (err != 0)
695 		return err;
696 
697 	lock_sock(sk);
698 
699 	ssk = __mptcp_nmpc_sk(msk);
700 	if (IS_ERR(ssk)) {
701 		release_sock(sk);
702 		return PTR_ERR(ssk);
703 	}
704 
705 	switch (optname) {
706 	case IP_FREEBIND:
707 		inet_assign_bit(FREEBIND, ssk, inet_test_bit(FREEBIND, sk));
708 		break;
709 	case IP_TRANSPARENT:
710 		inet_assign_bit(TRANSPARENT, ssk,
711 				inet_test_bit(TRANSPARENT, sk));
712 		break;
713 	default:
714 		release_sock(sk);
715 		WARN_ON_ONCE(1);
716 		return -EOPNOTSUPP;
717 	}
718 
719 	sockopt_seq_inc(msk);
720 	release_sock(sk);
721 	return 0;
722 }
723 
724 static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,
725 				       sockptr_t optval, unsigned int optlen)
726 {
727 	struct mptcp_subflow_context *subflow;
728 	struct sock *sk = (struct sock *)msk;
729 	int err, val;
730 
731 	err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
732 
733 	if (err != 0)
734 		return err;
735 
736 	lock_sock(sk);
737 	sockopt_seq_inc(msk);
738 	val = READ_ONCE(inet_sk(sk)->tos);
739 	mptcp_for_each_subflow(msk, subflow) {
740 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
741 		bool slow;
742 
743 		slow = lock_sock_fast(ssk);
744 		__ip_sock_set_tos(ssk, val);
745 		unlock_sock_fast(ssk, slow);
746 	}
747 	release_sock(sk);
748 
749 	return 0;
750 }
751 
752 static int mptcp_setsockopt_v4(struct mptcp_sock *msk, int optname,
753 			       sockptr_t optval, unsigned int optlen)
754 {
755 	switch (optname) {
756 	case IP_FREEBIND:
757 	case IP_TRANSPARENT:
758 		return mptcp_setsockopt_sol_ip_set_transparent(msk, optname, optval, optlen);
759 	case IP_TOS:
760 		return mptcp_setsockopt_v4_set_tos(msk, optname, optval, optlen);
761 	}
762 
763 	return -EOPNOTSUPP;
764 }
765 
766 static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
767 					  sockptr_t optval, unsigned int optlen)
768 {
769 	struct sock *sk = (struct sock *)msk;
770 	struct sock *ssk;
771 	int ret;
772 
773 	/* Limit to first subflow, before the connection establishment */
774 	lock_sock(sk);
775 	ssk = __mptcp_nmpc_sk(msk);
776 	if (IS_ERR(ssk)) {
777 		ret = PTR_ERR(ssk);
778 		goto unlock;
779 	}
780 
781 	ret = tcp_setsockopt(ssk, level, optname, optval, optlen);
782 
783 unlock:
784 	release_sock(sk);
785 	return ret;
786 }
787 
788 static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
789 				    sockptr_t optval, unsigned int optlen)
790 {
791 	struct sock *sk = (void *)msk;
792 	int ret, val;
793 
794 	switch (optname) {
795 	case TCP_INQ:
796 		ret = mptcp_get_int_option(msk, optval, optlen, &val);
797 		if (ret)
798 			return ret;
799 		if (val < 0 || val > 1)
800 			return -EINVAL;
801 
802 		lock_sock(sk);
803 		msk->recvmsg_inq = !!val;
804 		release_sock(sk);
805 		return 0;
806 	case TCP_ULP:
807 		return -EOPNOTSUPP;
808 	case TCP_CONGESTION:
809 		return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
810 	case TCP_CORK:
811 		return mptcp_setsockopt_sol_tcp_cork(msk, optval, optlen);
812 	case TCP_NODELAY:
813 		return mptcp_setsockopt_sol_tcp_nodelay(msk, optval, optlen);
814 	case TCP_DEFER_ACCEPT:
815 		/* See tcp.c: TCP_DEFER_ACCEPT does not fail */
816 		mptcp_setsockopt_first_sf_only(msk, SOL_TCP, optname, optval, optlen);
817 		return 0;
818 	case TCP_FASTOPEN:
819 	case TCP_FASTOPEN_CONNECT:
820 	case TCP_FASTOPEN_KEY:
821 	case TCP_FASTOPEN_NO_COOKIE:
822 		return mptcp_setsockopt_first_sf_only(msk, SOL_TCP, optname,
823 						      optval, optlen);
824 	}
825 
826 	return -EOPNOTSUPP;
827 }
828 
829 int mptcp_setsockopt(struct sock *sk, int level, int optname,
830 		     sockptr_t optval, unsigned int optlen)
831 {
832 	struct mptcp_sock *msk = mptcp_sk(sk);
833 	struct sock *ssk;
834 
835 	pr_debug("msk=%p", msk);
836 
837 	if (level == SOL_SOCKET)
838 		return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
839 
840 	if (!mptcp_supported_sockopt(level, optname))
841 		return -ENOPROTOOPT;
842 
843 	/* @@ the meaning of setsockopt() when the socket is connected and
844 	 * there are multiple subflows is not yet defined. It is up to the
845 	 * MPTCP-level socket to configure the subflows until the subflow
846 	 * is in TCP fallback, when TCP socket options are passed through
847 	 * to the one remaining subflow.
848 	 */
849 	lock_sock(sk);
850 	ssk = __mptcp_tcp_fallback(msk);
851 	release_sock(sk);
852 	if (ssk)
853 		return tcp_setsockopt(ssk, level, optname, optval, optlen);
854 
855 	if (level == SOL_IP)
856 		return mptcp_setsockopt_v4(msk, optname, optval, optlen);
857 
858 	if (level == SOL_IPV6)
859 		return mptcp_setsockopt_v6(msk, optname, optval, optlen);
860 
861 	if (level == SOL_TCP)
862 		return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
863 
864 	return -EOPNOTSUPP;
865 }
866 
867 static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
868 					  char __user *optval, int __user *optlen)
869 {
870 	struct sock *sk = (struct sock *)msk;
871 	struct sock *ssk;
872 	int ret;
873 
874 	lock_sock(sk);
875 	ssk = msk->first;
876 	if (ssk) {
877 		ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
878 		goto out;
879 	}
880 
881 	ssk = __mptcp_nmpc_sk(msk);
882 	if (IS_ERR(ssk)) {
883 		ret = PTR_ERR(ssk);
884 		goto out;
885 	}
886 
887 	ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
888 
889 out:
890 	release_sock(sk);
891 	return ret;
892 }
893 
894 void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
895 {
896 	struct sock *sk = (struct sock *)msk;
897 	u32 flags = 0;
898 	bool slow;
899 
900 	memset(info, 0, sizeof(*info));
901 
902 	info->mptcpi_subflows = READ_ONCE(msk->pm.subflows);
903 	info->mptcpi_add_addr_signal = READ_ONCE(msk->pm.add_addr_signaled);
904 	info->mptcpi_add_addr_accepted = READ_ONCE(msk->pm.add_addr_accepted);
905 	info->mptcpi_local_addr_used = READ_ONCE(msk->pm.local_addr_used);
906 
907 	if (inet_sk_state_load(sk) == TCP_LISTEN)
908 		return;
909 
910 	/* The following limits only make sense for the in-kernel PM */
911 	if (mptcp_pm_is_kernel(msk)) {
912 		info->mptcpi_subflows_max =
913 			mptcp_pm_get_subflows_max(msk);
914 		info->mptcpi_add_addr_signal_max =
915 			mptcp_pm_get_add_addr_signal_max(msk);
916 		info->mptcpi_add_addr_accepted_max =
917 			mptcp_pm_get_add_addr_accept_max(msk);
918 		info->mptcpi_local_addr_max =
919 			mptcp_pm_get_local_addr_max(msk);
920 	}
921 
922 	if (__mptcp_check_fallback(msk))
923 		flags |= MPTCP_INFO_FLAG_FALLBACK;
924 	if (READ_ONCE(msk->can_ack))
925 		flags |= MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED;
926 	info->mptcpi_flags = flags;
927 	mptcp_data_lock(sk);
928 	info->mptcpi_snd_una = msk->snd_una;
929 	info->mptcpi_rcv_nxt = msk->ack_seq;
930 	info->mptcpi_bytes_acked = msk->bytes_acked;
931 	mptcp_data_unlock(sk);
932 
933 	slow = lock_sock_fast(sk);
934 	info->mptcpi_csum_enabled = msk->csum_enabled;
935 	info->mptcpi_token = msk->token;
936 	info->mptcpi_write_seq = msk->write_seq;
937 	info->mptcpi_retransmits = inet_csk(sk)->icsk_retransmits;
938 	info->mptcpi_bytes_sent = msk->bytes_sent;
939 	info->mptcpi_bytes_received = msk->bytes_received;
940 	info->mptcpi_bytes_retrans = msk->bytes_retrans;
941 	unlock_sock_fast(sk, slow);
942 }
943 EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
944 
945 static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
946 {
947 	struct mptcp_info m_info;
948 	int len;
949 
950 	if (get_user(len, optlen))
951 		return -EFAULT;
952 
953 	len = min_t(unsigned int, len, sizeof(struct mptcp_info));
954 
955 	mptcp_diag_fill_info(msk, &m_info);
956 
957 	if (put_user(len, optlen))
958 		return -EFAULT;
959 
960 	if (copy_to_user(optval, &m_info, len))
961 		return -EFAULT;
962 
963 	return 0;
964 }
965 
966 static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
967 				  char __user *optval,
968 				  u32 copied,
969 				  int __user *optlen)
970 {
971 	u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
972 
973 	if (copied)
974 		copied += sfd->size_subflow_data;
975 	else
976 		copied = copylen;
977 
978 	if (put_user(copied, optlen))
979 		return -EFAULT;
980 
981 	if (copy_to_user(optval, sfd, copylen))
982 		return -EFAULT;
983 
984 	return 0;
985 }
986 
987 static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
988 				  char __user *optval,
989 				  int __user *optlen)
990 {
991 	int len, copylen;
992 
993 	if (get_user(len, optlen))
994 		return -EFAULT;
995 
996 	/* if mptcp_subflow_data size is changed, need to adjust
997 	 * this function to deal with programs using old version.
998 	 */
999 	BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
1000 
1001 	if (len < MIN_INFO_OPTLEN_SIZE)
1002 		return -EINVAL;
1003 
1004 	memset(sfd, 0, sizeof(*sfd));
1005 
1006 	copylen = min_t(unsigned int, len, sizeof(*sfd));
1007 	if (copy_from_user(sfd, optval, copylen))
1008 		return -EFAULT;
1009 
1010 	/* size_subflow_data is u32, but len is signed */
1011 	if (sfd->size_subflow_data > INT_MAX ||
1012 	    sfd->size_user > INT_MAX)
1013 		return -EINVAL;
1014 
1015 	if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
1016 	    sfd->size_subflow_data > len)
1017 		return -EINVAL;
1018 
1019 	if (sfd->num_subflows || sfd->size_kernel)
1020 		return -EINVAL;
1021 
1022 	return len - sfd->size_subflow_data;
1023 }
1024 
1025 static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
1026 				    int __user *optlen)
1027 {
1028 	struct mptcp_subflow_context *subflow;
1029 	struct sock *sk = (struct sock *)msk;
1030 	unsigned int sfcount = 0, copied = 0;
1031 	struct mptcp_subflow_data sfd;
1032 	char __user *infoptr;
1033 	int len;
1034 
1035 	len = mptcp_get_subflow_data(&sfd, optval, optlen);
1036 	if (len < 0)
1037 		return len;
1038 
1039 	sfd.size_kernel = sizeof(struct tcp_info);
1040 	sfd.size_user = min_t(unsigned int, sfd.size_user,
1041 			      sizeof(struct tcp_info));
1042 
1043 	infoptr = optval + sfd.size_subflow_data;
1044 
1045 	lock_sock(sk);
1046 
1047 	mptcp_for_each_subflow(msk, subflow) {
1048 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1049 
1050 		++sfcount;
1051 
1052 		if (len && len >= sfd.size_user) {
1053 			struct tcp_info info;
1054 
1055 			tcp_get_info(ssk, &info);
1056 
1057 			if (copy_to_user(infoptr, &info, sfd.size_user)) {
1058 				release_sock(sk);
1059 				return -EFAULT;
1060 			}
1061 
1062 			infoptr += sfd.size_user;
1063 			copied += sfd.size_user;
1064 			len -= sfd.size_user;
1065 		}
1066 	}
1067 
1068 	release_sock(sk);
1069 
1070 	sfd.num_subflows = sfcount;
1071 
1072 	if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1073 		return -EFAULT;
1074 
1075 	return 0;
1076 }
1077 
1078 static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
1079 {
1080 	const struct inet_sock *inet = inet_sk(sk);
1081 
1082 	memset(a, 0, sizeof(*a));
1083 
1084 	if (sk->sk_family == AF_INET) {
1085 		a->sin_local.sin_family = AF_INET;
1086 		a->sin_local.sin_port = inet->inet_sport;
1087 		a->sin_local.sin_addr.s_addr = inet->inet_rcv_saddr;
1088 
1089 		if (!a->sin_local.sin_addr.s_addr)
1090 			a->sin_local.sin_addr.s_addr = inet->inet_saddr;
1091 
1092 		a->sin_remote.sin_family = AF_INET;
1093 		a->sin_remote.sin_port = inet->inet_dport;
1094 		a->sin_remote.sin_addr.s_addr = inet->inet_daddr;
1095 #if IS_ENABLED(CONFIG_IPV6)
1096 	} else if (sk->sk_family == AF_INET6) {
1097 		const struct ipv6_pinfo *np = inet6_sk(sk);
1098 
1099 		if (WARN_ON_ONCE(!np))
1100 			return;
1101 
1102 		a->sin6_local.sin6_family = AF_INET6;
1103 		a->sin6_local.sin6_port = inet->inet_sport;
1104 
1105 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
1106 			a->sin6_local.sin6_addr = np->saddr;
1107 		else
1108 			a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
1109 
1110 		a->sin6_remote.sin6_family = AF_INET6;
1111 		a->sin6_remote.sin6_port = inet->inet_dport;
1112 		a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
1113 #endif
1114 	}
1115 }
1116 
1117 static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
1118 					  int __user *optlen)
1119 {
1120 	struct mptcp_subflow_context *subflow;
1121 	struct sock *sk = (struct sock *)msk;
1122 	unsigned int sfcount = 0, copied = 0;
1123 	struct mptcp_subflow_data sfd;
1124 	char __user *addrptr;
1125 	int len;
1126 
1127 	len = mptcp_get_subflow_data(&sfd, optval, optlen);
1128 	if (len < 0)
1129 		return len;
1130 
1131 	sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
1132 	sfd.size_user = min_t(unsigned int, sfd.size_user,
1133 			      sizeof(struct mptcp_subflow_addrs));
1134 
1135 	addrptr = optval + sfd.size_subflow_data;
1136 
1137 	lock_sock(sk);
1138 
1139 	mptcp_for_each_subflow(msk, subflow) {
1140 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1141 
1142 		++sfcount;
1143 
1144 		if (len && len >= sfd.size_user) {
1145 			struct mptcp_subflow_addrs a;
1146 
1147 			mptcp_get_sub_addrs(ssk, &a);
1148 
1149 			if (copy_to_user(addrptr, &a, sfd.size_user)) {
1150 				release_sock(sk);
1151 				return -EFAULT;
1152 			}
1153 
1154 			addrptr += sfd.size_user;
1155 			copied += sfd.size_user;
1156 			len -= sfd.size_user;
1157 		}
1158 	}
1159 
1160 	release_sock(sk);
1161 
1162 	sfd.num_subflows = sfcount;
1163 
1164 	if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1165 		return -EFAULT;
1166 
1167 	return 0;
1168 }
1169 
1170 static int mptcp_get_full_info(struct mptcp_full_info *mfi,
1171 			       char __user *optval,
1172 			       int __user *optlen)
1173 {
1174 	int len;
1175 
1176 	BUILD_BUG_ON(offsetof(struct mptcp_full_info, mptcp_info) !=
1177 		     MIN_FULL_INFO_OPTLEN_SIZE);
1178 
1179 	if (get_user(len, optlen))
1180 		return -EFAULT;
1181 
1182 	if (len < MIN_FULL_INFO_OPTLEN_SIZE)
1183 		return -EINVAL;
1184 
1185 	memset(mfi, 0, sizeof(*mfi));
1186 	if (copy_from_user(mfi, optval, MIN_FULL_INFO_OPTLEN_SIZE))
1187 		return -EFAULT;
1188 
1189 	if (mfi->size_tcpinfo_kernel ||
1190 	    mfi->size_sfinfo_kernel ||
1191 	    mfi->num_subflows)
1192 		return -EINVAL;
1193 
1194 	if (mfi->size_sfinfo_user > INT_MAX ||
1195 	    mfi->size_tcpinfo_user > INT_MAX)
1196 		return -EINVAL;
1197 
1198 	return len - MIN_FULL_INFO_OPTLEN_SIZE;
1199 }
1200 
1201 static int mptcp_put_full_info(struct mptcp_full_info *mfi,
1202 			       char __user *optval,
1203 			       u32 copylen,
1204 			       int __user *optlen)
1205 {
1206 	copylen += MIN_FULL_INFO_OPTLEN_SIZE;
1207 	if (put_user(copylen, optlen))
1208 		return -EFAULT;
1209 
1210 	if (copy_to_user(optval, mfi, copylen))
1211 		return -EFAULT;
1212 	return 0;
1213 }
1214 
1215 static int mptcp_getsockopt_full_info(struct mptcp_sock *msk, char __user *optval,
1216 				      int __user *optlen)
1217 {
1218 	unsigned int sfcount = 0, copylen = 0;
1219 	struct mptcp_subflow_context *subflow;
1220 	struct sock *sk = (struct sock *)msk;
1221 	void __user *tcpinfoptr, *sfinfoptr;
1222 	struct mptcp_full_info mfi;
1223 	int len;
1224 
1225 	len = mptcp_get_full_info(&mfi, optval, optlen);
1226 	if (len < 0)
1227 		return len;
1228 
1229 	/* don't bother filling the mptcp info if there is not enough
1230 	 * user-space-provided storage
1231 	 */
1232 	if (len > 0) {
1233 		mptcp_diag_fill_info(msk, &mfi.mptcp_info);
1234 		copylen += min_t(unsigned int, len, sizeof(struct mptcp_info));
1235 	}
1236 
1237 	mfi.size_tcpinfo_kernel = sizeof(struct tcp_info);
1238 	mfi.size_tcpinfo_user = min_t(unsigned int, mfi.size_tcpinfo_user,
1239 				      sizeof(struct tcp_info));
1240 	sfinfoptr = u64_to_user_ptr(mfi.subflow_info);
1241 	mfi.size_sfinfo_kernel = sizeof(struct mptcp_subflow_info);
1242 	mfi.size_sfinfo_user = min_t(unsigned int, mfi.size_sfinfo_user,
1243 				     sizeof(struct mptcp_subflow_info));
1244 	tcpinfoptr = u64_to_user_ptr(mfi.tcp_info);
1245 
1246 	lock_sock(sk);
1247 	mptcp_for_each_subflow(msk, subflow) {
1248 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1249 		struct mptcp_subflow_info sfinfo;
1250 		struct tcp_info tcp_info;
1251 
1252 		if (sfcount++ >= mfi.size_arrays_user)
1253 			continue;
1254 
1255 		/* fetch addr/tcp_info only if the user space buffers
1256 		 * are wide enough
1257 		 */
1258 		memset(&sfinfo, 0, sizeof(sfinfo));
1259 		sfinfo.id = subflow->subflow_id;
1260 		if (mfi.size_sfinfo_user >
1261 		    offsetof(struct mptcp_subflow_info, addrs))
1262 			mptcp_get_sub_addrs(ssk, &sfinfo.addrs);
1263 		if (copy_to_user(sfinfoptr, &sfinfo, mfi.size_sfinfo_user))
1264 			goto fail_release;
1265 
1266 		if (mfi.size_tcpinfo_user) {
1267 			tcp_get_info(ssk, &tcp_info);
1268 			if (copy_to_user(tcpinfoptr, &tcp_info,
1269 					 mfi.size_tcpinfo_user))
1270 				goto fail_release;
1271 		}
1272 
1273 		tcpinfoptr += mfi.size_tcpinfo_user;
1274 		sfinfoptr += mfi.size_sfinfo_user;
1275 	}
1276 	release_sock(sk);
1277 
1278 	mfi.num_subflows = sfcount;
1279 	if (mptcp_put_full_info(&mfi, optval, copylen, optlen))
1280 		return -EFAULT;
1281 
1282 	return 0;
1283 
1284 fail_release:
1285 	release_sock(sk);
1286 	return -EFAULT;
1287 }
1288 
1289 static int mptcp_put_int_option(struct mptcp_sock *msk, char __user *optval,
1290 				int __user *optlen, int val)
1291 {
1292 	int len;
1293 
1294 	if (get_user(len, optlen))
1295 		return -EFAULT;
1296 	if (len < 0)
1297 		return -EINVAL;
1298 
1299 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1300 		unsigned char ucval = (unsigned char)val;
1301 
1302 		len = 1;
1303 		if (put_user(len, optlen))
1304 			return -EFAULT;
1305 		if (copy_to_user(optval, &ucval, 1))
1306 			return -EFAULT;
1307 	} else {
1308 		len = min_t(unsigned int, len, sizeof(int));
1309 		if (put_user(len, optlen))
1310 			return -EFAULT;
1311 		if (copy_to_user(optval, &val, len))
1312 			return -EFAULT;
1313 	}
1314 
1315 	return 0;
1316 }
1317 
1318 static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
1319 				    char __user *optval, int __user *optlen)
1320 {
1321 	switch (optname) {
1322 	case TCP_ULP:
1323 	case TCP_CONGESTION:
1324 	case TCP_INFO:
1325 	case TCP_CC_INFO:
1326 	case TCP_DEFER_ACCEPT:
1327 	case TCP_FASTOPEN:
1328 	case TCP_FASTOPEN_CONNECT:
1329 	case TCP_FASTOPEN_KEY:
1330 	case TCP_FASTOPEN_NO_COOKIE:
1331 		return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
1332 						      optval, optlen);
1333 	case TCP_INQ:
1334 		return mptcp_put_int_option(msk, optval, optlen, msk->recvmsg_inq);
1335 	case TCP_CORK:
1336 		return mptcp_put_int_option(msk, optval, optlen, msk->cork);
1337 	case TCP_NODELAY:
1338 		return mptcp_put_int_option(msk, optval, optlen, msk->nodelay);
1339 	}
1340 	return -EOPNOTSUPP;
1341 }
1342 
1343 static int mptcp_getsockopt_v4(struct mptcp_sock *msk, int optname,
1344 			       char __user *optval, int __user *optlen)
1345 {
1346 	struct sock *sk = (void *)msk;
1347 
1348 	switch (optname) {
1349 	case IP_TOS:
1350 		return mptcp_put_int_option(msk, optval, optlen, READ_ONCE(inet_sk(sk)->tos));
1351 	}
1352 
1353 	return -EOPNOTSUPP;
1354 }
1355 
1356 static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
1357 				      char __user *optval, int __user *optlen)
1358 {
1359 	switch (optname) {
1360 	case MPTCP_INFO:
1361 		return mptcp_getsockopt_info(msk, optval, optlen);
1362 	case MPTCP_FULL_INFO:
1363 		return mptcp_getsockopt_full_info(msk, optval, optlen);
1364 	case MPTCP_TCPINFO:
1365 		return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
1366 	case MPTCP_SUBFLOW_ADDRS:
1367 		return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
1368 	}
1369 
1370 	return -EOPNOTSUPP;
1371 }
1372 
1373 int mptcp_getsockopt(struct sock *sk, int level, int optname,
1374 		     char __user *optval, int __user *option)
1375 {
1376 	struct mptcp_sock *msk = mptcp_sk(sk);
1377 	struct sock *ssk;
1378 
1379 	pr_debug("msk=%p", msk);
1380 
1381 	/* @@ the meaning of setsockopt() when the socket is connected and
1382 	 * there are multiple subflows is not yet defined. It is up to the
1383 	 * MPTCP-level socket to configure the subflows until the subflow
1384 	 * is in TCP fallback, when socket options are passed through
1385 	 * to the one remaining subflow.
1386 	 */
1387 	lock_sock(sk);
1388 	ssk = __mptcp_tcp_fallback(msk);
1389 	release_sock(sk);
1390 	if (ssk)
1391 		return tcp_getsockopt(ssk, level, optname, optval, option);
1392 
1393 	if (level == SOL_IP)
1394 		return mptcp_getsockopt_v4(msk, optname, optval, option);
1395 	if (level == SOL_TCP)
1396 		return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
1397 	if (level == SOL_MPTCP)
1398 		return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
1399 	return -EOPNOTSUPP;
1400 }
1401 
1402 static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
1403 {
1404 	static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
1405 	struct sock *sk = (struct sock *)msk;
1406 
1407 	if (ssk->sk_prot->keepalive) {
1408 		if (sock_flag(sk, SOCK_KEEPOPEN))
1409 			ssk->sk_prot->keepalive(ssk, 1);
1410 		else
1411 			ssk->sk_prot->keepalive(ssk, 0);
1412 	}
1413 
1414 	ssk->sk_priority = sk->sk_priority;
1415 	ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1416 	ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
1417 	ssk->sk_ipv6only = sk->sk_ipv6only;
1418 	__ip_sock_set_tos(ssk, inet_sk(sk)->tos);
1419 
1420 	if (sk->sk_userlocks & tx_rx_locks) {
1421 		ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1422 		if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) {
1423 			WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1424 			mptcp_subflow_ctx(ssk)->cached_sndbuf = sk->sk_sndbuf;
1425 		}
1426 		if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1427 			WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1428 	}
1429 
1430 	if (sock_flag(sk, SOCK_LINGER)) {
1431 		ssk->sk_lingertime = sk->sk_lingertime;
1432 		sock_set_flag(ssk, SOCK_LINGER);
1433 	} else {
1434 		sock_reset_flag(ssk, SOCK_LINGER);
1435 	}
1436 
1437 	if (sk->sk_mark != ssk->sk_mark) {
1438 		ssk->sk_mark = sk->sk_mark;
1439 		sk_dst_reset(ssk);
1440 	}
1441 
1442 	sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1443 
1444 	if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
1445 		tcp_set_congestion_control(ssk, msk->ca_name, false, true);
1446 	__tcp_sock_set_cork(ssk, !!msk->cork);
1447 	__tcp_sock_set_nodelay(ssk, !!msk->nodelay);
1448 
1449 	inet_assign_bit(TRANSPARENT, ssk, inet_test_bit(TRANSPARENT, sk));
1450 	inet_assign_bit(FREEBIND, ssk, inet_test_bit(FREEBIND, sk));
1451 }
1452 
1453 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk)
1454 {
1455 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1456 
1457 	msk_owned_by_me(msk);
1458 
1459 	ssk->sk_rcvlowat = 0;
1460 
1461 	/* subflows must ignore any latency-related settings: will not affect
1462 	 * the user-space - only the msk is relevant - but will foul the
1463 	 * mptcp scheduler
1464 	 */
1465 	tcp_sk(ssk)->notsent_lowat = UINT_MAX;
1466 
1467 	if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1468 		sync_socket_options(msk, ssk);
1469 
1470 		subflow->setsockopt_seq = msk->setsockopt_seq;
1471 	}
1472 }
1473 
1474 /* unfortunately this is different enough from the tcp version so
1475  * that we can't factor it out
1476  */
1477 int mptcp_set_rcvlowat(struct sock *sk, int val)
1478 {
1479 	struct mptcp_subflow_context *subflow;
1480 	int space, cap;
1481 
1482 	if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1483 		cap = sk->sk_rcvbuf >> 1;
1484 	else
1485 		cap = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_rmem[2]) >> 1;
1486 	val = min(val, cap);
1487 	WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);
1488 
1489 	/* Check if we need to signal EPOLLIN right now */
1490 	if (mptcp_epollin_ready(sk))
1491 		sk->sk_data_ready(sk);
1492 
1493 	if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1494 		return 0;
1495 
1496 	space = __tcp_space_from_win(mptcp_sk(sk)->scaling_ratio, val);
1497 	if (space <= sk->sk_rcvbuf)
1498 		return 0;
1499 
1500 	/* propagate the rcvbuf changes to all the subflows */
1501 	WRITE_ONCE(sk->sk_rcvbuf, space);
1502 	mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
1503 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1504 		bool slow;
1505 
1506 		slow = lock_sock_fast(ssk);
1507 		WRITE_ONCE(ssk->sk_rcvbuf, space);
1508 		tcp_sk(ssk)->window_clamp = val;
1509 		unlock_sock_fast(ssk, slow);
1510 	}
1511 	return 0;
1512 }
1513