xref: /openbmc/linux/net/mptcp/sockopt.c (revision febf2aaf)
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 			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 			break;
99 		case SO_RCVBUF:
100 		case SO_RCVBUFFORCE:
101 			ssk->sk_userlocks |= SOCK_RCVBUF_LOCK;
102 			WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
103 			break;
104 		case SO_MARK:
105 			if (READ_ONCE(ssk->sk_mark) != sk->sk_mark) {
106 				ssk->sk_mark = sk->sk_mark;
107 				sk_dst_reset(ssk);
108 			}
109 			break;
110 		case SO_INCOMING_CPU:
111 			WRITE_ONCE(ssk->sk_incoming_cpu, val);
112 			break;
113 		}
114 
115 		subflow->setsockopt_seq = msk->setsockopt_seq;
116 		unlock_sock_fast(ssk, slow);
117 	}
118 
119 	release_sock(sk);
120 }
121 
122 static int mptcp_sol_socket_intval(struct mptcp_sock *msk, int optname, int val)
123 {
124 	sockptr_t optval = KERNEL_SOCKPTR(&val);
125 	struct sock *sk = (struct sock *)msk;
126 	int ret;
127 
128 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
129 			      optval, sizeof(val));
130 	if (ret)
131 		return ret;
132 
133 	mptcp_sol_socket_sync_intval(msk, optname, val);
134 	return 0;
135 }
136 
137 static void mptcp_so_incoming_cpu(struct mptcp_sock *msk, int val)
138 {
139 	struct sock *sk = (struct sock *)msk;
140 
141 	WRITE_ONCE(sk->sk_incoming_cpu, val);
142 
143 	mptcp_sol_socket_sync_intval(msk, SO_INCOMING_CPU, val);
144 }
145 
146 static int mptcp_setsockopt_sol_socket_tstamp(struct mptcp_sock *msk, int optname, int val)
147 {
148 	sockptr_t optval = KERNEL_SOCKPTR(&val);
149 	struct mptcp_subflow_context *subflow;
150 	struct sock *sk = (struct sock *)msk;
151 	int ret;
152 
153 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
154 			      optval, sizeof(val));
155 	if (ret)
156 		return ret;
157 
158 	lock_sock(sk);
159 	mptcp_for_each_subflow(msk, subflow) {
160 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
161 		bool slow = lock_sock_fast(ssk);
162 
163 		sock_set_timestamp(sk, optname, !!val);
164 		unlock_sock_fast(ssk, slow);
165 	}
166 
167 	release_sock(sk);
168 	return 0;
169 }
170 
171 static int mptcp_setsockopt_sol_socket_int(struct mptcp_sock *msk, int optname,
172 					   sockptr_t optval,
173 					   unsigned int optlen)
174 {
175 	int val, ret;
176 
177 	ret = mptcp_get_int_option(msk, optval, optlen, &val);
178 	if (ret)
179 		return ret;
180 
181 	switch (optname) {
182 	case SO_KEEPALIVE:
183 		mptcp_sol_socket_sync_intval(msk, optname, val);
184 		return 0;
185 	case SO_DEBUG:
186 	case SO_MARK:
187 	case SO_PRIORITY:
188 	case SO_SNDBUF:
189 	case SO_SNDBUFFORCE:
190 	case SO_RCVBUF:
191 	case SO_RCVBUFFORCE:
192 		return mptcp_sol_socket_intval(msk, optname, val);
193 	case SO_INCOMING_CPU:
194 		mptcp_so_incoming_cpu(msk, val);
195 		return 0;
196 	case SO_TIMESTAMP_OLD:
197 	case SO_TIMESTAMP_NEW:
198 	case SO_TIMESTAMPNS_OLD:
199 	case SO_TIMESTAMPNS_NEW:
200 		return mptcp_setsockopt_sol_socket_tstamp(msk, optname, val);
201 	}
202 
203 	return -ENOPROTOOPT;
204 }
205 
206 static int mptcp_setsockopt_sol_socket_timestamping(struct mptcp_sock *msk,
207 						    int optname,
208 						    sockptr_t optval,
209 						    unsigned int optlen)
210 {
211 	struct mptcp_subflow_context *subflow;
212 	struct sock *sk = (struct sock *)msk;
213 	struct so_timestamping timestamping;
214 	int ret;
215 
216 	if (optlen == sizeof(timestamping)) {
217 		if (copy_from_sockptr(&timestamping, optval,
218 				      sizeof(timestamping)))
219 			return -EFAULT;
220 	} else if (optlen == sizeof(int)) {
221 		memset(&timestamping, 0, sizeof(timestamping));
222 
223 		if (copy_from_sockptr(&timestamping.flags, optval, sizeof(int)))
224 			return -EFAULT;
225 	} else {
226 		return -EINVAL;
227 	}
228 
229 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
230 			      KERNEL_SOCKPTR(&timestamping),
231 			      sizeof(timestamping));
232 	if (ret)
233 		return ret;
234 
235 	lock_sock(sk);
236 
237 	mptcp_for_each_subflow(msk, subflow) {
238 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
239 		bool slow = lock_sock_fast(ssk);
240 
241 		sock_set_timestamping(sk, optname, timestamping);
242 		unlock_sock_fast(ssk, slow);
243 	}
244 
245 	release_sock(sk);
246 
247 	return 0;
248 }
249 
250 static int mptcp_setsockopt_sol_socket_linger(struct mptcp_sock *msk, sockptr_t optval,
251 					      unsigned int optlen)
252 {
253 	struct mptcp_subflow_context *subflow;
254 	struct sock *sk = (struct sock *)msk;
255 	struct linger ling;
256 	sockptr_t kopt;
257 	int ret;
258 
259 	if (optlen < sizeof(ling))
260 		return -EINVAL;
261 
262 	if (copy_from_sockptr(&ling, optval, sizeof(ling)))
263 		return -EFAULT;
264 
265 	kopt = KERNEL_SOCKPTR(&ling);
266 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, SO_LINGER, kopt, sizeof(ling));
267 	if (ret)
268 		return ret;
269 
270 	lock_sock(sk);
271 	sockopt_seq_inc(msk);
272 	mptcp_for_each_subflow(msk, subflow) {
273 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
274 		bool slow = lock_sock_fast(ssk);
275 
276 		if (!ling.l_onoff) {
277 			sock_reset_flag(ssk, SOCK_LINGER);
278 		} else {
279 			ssk->sk_lingertime = sk->sk_lingertime;
280 			sock_set_flag(ssk, SOCK_LINGER);
281 		}
282 
283 		subflow->setsockopt_seq = msk->setsockopt_seq;
284 		unlock_sock_fast(ssk, slow);
285 	}
286 
287 	release_sock(sk);
288 	return 0;
289 }
290 
291 static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,
292 				       sockptr_t optval, unsigned int optlen)
293 {
294 	struct sock *sk = (struct sock *)msk;
295 	struct socket *ssock;
296 	int ret;
297 
298 	switch (optname) {
299 	case SO_REUSEPORT:
300 	case SO_REUSEADDR:
301 	case SO_BINDTODEVICE:
302 	case SO_BINDTOIFINDEX:
303 		lock_sock(sk);
304 		ssock = __mptcp_nmpc_socket(msk);
305 		if (IS_ERR(ssock)) {
306 			release_sock(sk);
307 			return PTR_ERR(ssock);
308 		}
309 
310 		ret = sock_setsockopt(ssock, SOL_SOCKET, optname, optval, optlen);
311 		if (ret == 0) {
312 			if (optname == SO_REUSEPORT)
313 				sk->sk_reuseport = ssock->sk->sk_reuseport;
314 			else if (optname == SO_REUSEADDR)
315 				sk->sk_reuse = ssock->sk->sk_reuse;
316 			else if (optname == SO_BINDTODEVICE)
317 				sk->sk_bound_dev_if = ssock->sk->sk_bound_dev_if;
318 			else if (optname == SO_BINDTOIFINDEX)
319 				sk->sk_bound_dev_if = ssock->sk->sk_bound_dev_if;
320 		}
321 		release_sock(sk);
322 		return ret;
323 	case SO_KEEPALIVE:
324 	case SO_PRIORITY:
325 	case SO_SNDBUF:
326 	case SO_SNDBUFFORCE:
327 	case SO_RCVBUF:
328 	case SO_RCVBUFFORCE:
329 	case SO_MARK:
330 	case SO_INCOMING_CPU:
331 	case SO_DEBUG:
332 	case SO_TIMESTAMP_OLD:
333 	case SO_TIMESTAMP_NEW:
334 	case SO_TIMESTAMPNS_OLD:
335 	case SO_TIMESTAMPNS_NEW:
336 		return mptcp_setsockopt_sol_socket_int(msk, optname, optval,
337 						       optlen);
338 	case SO_TIMESTAMPING_OLD:
339 	case SO_TIMESTAMPING_NEW:
340 		return mptcp_setsockopt_sol_socket_timestamping(msk, optname,
341 								optval, optlen);
342 	case SO_LINGER:
343 		return mptcp_setsockopt_sol_socket_linger(msk, optval, optlen);
344 	case SO_RCVLOWAT:
345 	case SO_RCVTIMEO_OLD:
346 	case SO_RCVTIMEO_NEW:
347 	case SO_SNDTIMEO_OLD:
348 	case SO_SNDTIMEO_NEW:
349 	case SO_BUSY_POLL:
350 	case SO_PREFER_BUSY_POLL:
351 	case SO_BUSY_POLL_BUDGET:
352 		/* No need to copy: only relevant for msk */
353 		return sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname, optval, optlen);
354 	case SO_NO_CHECK:
355 	case SO_DONTROUTE:
356 	case SO_BROADCAST:
357 	case SO_BSDCOMPAT:
358 	case SO_PASSCRED:
359 	case SO_PASSPIDFD:
360 	case SO_PASSSEC:
361 	case SO_RXQ_OVFL:
362 	case SO_WIFI_STATUS:
363 	case SO_NOFCS:
364 	case SO_SELECT_ERR_QUEUE:
365 		return 0;
366 	}
367 
368 	/* SO_OOBINLINE is not supported, let's avoid the related mess
369 	 * SO_ATTACH_FILTER, SO_ATTACH_BPF, SO_ATTACH_REUSEPORT_CBPF,
370 	 * SO_DETACH_REUSEPORT_BPF, SO_DETACH_FILTER, SO_LOCK_FILTER,
371 	 * we must be careful with subflows
372 	 *
373 	 * SO_ATTACH_REUSEPORT_EBPF is not supported, at it checks
374 	 * explicitly the sk_protocol field
375 	 *
376 	 * SO_PEEK_OFF is unsupported, as it is for plain TCP
377 	 * SO_MAX_PACING_RATE is unsupported, we must be careful with subflows
378 	 * SO_CNX_ADVICE is currently unsupported, could possibly be relevant,
379 	 * but likely needs careful design
380 	 *
381 	 * SO_ZEROCOPY is currently unsupported, TODO in sndmsg
382 	 * SO_TXTIME is currently unsupported
383 	 */
384 
385 	return -EOPNOTSUPP;
386 }
387 
388 static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,
389 			       sockptr_t optval, unsigned int optlen)
390 {
391 	struct sock *sk = (struct sock *)msk;
392 	int ret = -EOPNOTSUPP;
393 	struct socket *ssock;
394 
395 	switch (optname) {
396 	case IPV6_V6ONLY:
397 	case IPV6_TRANSPARENT:
398 	case IPV6_FREEBIND:
399 		lock_sock(sk);
400 		ssock = __mptcp_nmpc_socket(msk);
401 		if (IS_ERR(ssock)) {
402 			release_sock(sk);
403 			return PTR_ERR(ssock);
404 		}
405 
406 		ret = tcp_setsockopt(ssock->sk, SOL_IPV6, optname, optval, optlen);
407 		if (ret != 0) {
408 			release_sock(sk);
409 			return ret;
410 		}
411 
412 		sockopt_seq_inc(msk);
413 
414 		switch (optname) {
415 		case IPV6_V6ONLY:
416 			sk->sk_ipv6only = ssock->sk->sk_ipv6only;
417 			break;
418 		case IPV6_TRANSPARENT:
419 			inet_sk(sk)->transparent = inet_sk(ssock->sk)->transparent;
420 			break;
421 		case IPV6_FREEBIND:
422 			inet_sk(sk)->freebind = inet_sk(ssock->sk)->freebind;
423 			break;
424 		}
425 
426 		release_sock(sk);
427 		break;
428 	}
429 
430 	return ret;
431 }
432 
433 static bool mptcp_supported_sockopt(int level, int optname)
434 {
435 	if (level == SOL_IP) {
436 		switch (optname) {
437 		/* should work fine */
438 		case IP_FREEBIND:
439 		case IP_TRANSPARENT:
440 
441 		/* the following are control cmsg related */
442 		case IP_PKTINFO:
443 		case IP_RECVTTL:
444 		case IP_RECVTOS:
445 		case IP_RECVOPTS:
446 		case IP_RETOPTS:
447 		case IP_PASSSEC:
448 		case IP_RECVORIGDSTADDR:
449 		case IP_CHECKSUM:
450 		case IP_RECVFRAGSIZE:
451 
452 		/* common stuff that need some love */
453 		case IP_TOS:
454 		case IP_TTL:
455 		case IP_BIND_ADDRESS_NO_PORT:
456 		case IP_MTU_DISCOVER:
457 		case IP_RECVERR:
458 
459 		/* possibly less common may deserve some love */
460 		case IP_MINTTL:
461 
462 		/* the following is apparently a no-op for plain TCP */
463 		case IP_RECVERR_RFC4884:
464 			return true;
465 		}
466 
467 		/* IP_OPTIONS is not supported, needs subflow care */
468 		/* IP_HDRINCL, IP_NODEFRAG are not supported, RAW specific */
469 		/* IP_MULTICAST_TTL, IP_MULTICAST_LOOP, IP_UNICAST_IF,
470 		 * IP_ADD_MEMBERSHIP, IP_ADD_SOURCE_MEMBERSHIP, IP_DROP_MEMBERSHIP,
471 		 * IP_DROP_SOURCE_MEMBERSHIP, IP_BLOCK_SOURCE, IP_UNBLOCK_SOURCE,
472 		 * MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP MCAST_JOIN_SOURCE_GROUP,
473 		 * MCAST_LEAVE_SOURCE_GROUP, MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE,
474 		 * MCAST_MSFILTER, IP_MULTICAST_ALL are not supported, better not deal
475 		 * with mcast stuff
476 		 */
477 		/* IP_IPSEC_POLICY, IP_XFRM_POLICY are nut supported, unrelated here */
478 		return false;
479 	}
480 	if (level == SOL_IPV6) {
481 		switch (optname) {
482 		case IPV6_V6ONLY:
483 
484 		/* the following are control cmsg related */
485 		case IPV6_RECVPKTINFO:
486 		case IPV6_2292PKTINFO:
487 		case IPV6_RECVHOPLIMIT:
488 		case IPV6_2292HOPLIMIT:
489 		case IPV6_RECVRTHDR:
490 		case IPV6_2292RTHDR:
491 		case IPV6_RECVHOPOPTS:
492 		case IPV6_2292HOPOPTS:
493 		case IPV6_RECVDSTOPTS:
494 		case IPV6_2292DSTOPTS:
495 		case IPV6_RECVTCLASS:
496 		case IPV6_FLOWINFO:
497 		case IPV6_RECVPATHMTU:
498 		case IPV6_RECVORIGDSTADDR:
499 		case IPV6_RECVFRAGSIZE:
500 
501 		/* the following ones need some love but are quite common */
502 		case IPV6_TCLASS:
503 		case IPV6_TRANSPARENT:
504 		case IPV6_FREEBIND:
505 		case IPV6_PKTINFO:
506 		case IPV6_2292PKTOPTIONS:
507 		case IPV6_UNICAST_HOPS:
508 		case IPV6_MTU_DISCOVER:
509 		case IPV6_MTU:
510 		case IPV6_RECVERR:
511 		case IPV6_FLOWINFO_SEND:
512 		case IPV6_FLOWLABEL_MGR:
513 		case IPV6_MINHOPCOUNT:
514 		case IPV6_DONTFRAG:
515 		case IPV6_AUTOFLOWLABEL:
516 
517 		/* the following one is a no-op for plain TCP */
518 		case IPV6_RECVERR_RFC4884:
519 			return true;
520 		}
521 
522 		/* IPV6_HOPOPTS, IPV6_RTHDRDSTOPTS, IPV6_RTHDR, IPV6_DSTOPTS are
523 		 * not supported
524 		 */
525 		/* IPV6_MULTICAST_HOPS, IPV6_MULTICAST_LOOP, IPV6_UNICAST_IF,
526 		 * IPV6_MULTICAST_IF, IPV6_ADDRFORM,
527 		 * IPV6_ADD_MEMBERSHIP, IPV6_DROP_MEMBERSHIP, IPV6_JOIN_ANYCAST,
528 		 * IPV6_LEAVE_ANYCAST, IPV6_MULTICAST_ALL, MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP,
529 		 * MCAST_JOIN_SOURCE_GROUP, MCAST_LEAVE_SOURCE_GROUP,
530 		 * MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE, MCAST_MSFILTER
531 		 * are not supported better not deal with mcast
532 		 */
533 		/* IPV6_ROUTER_ALERT, IPV6_ROUTER_ALERT_ISOLATE are not supported, since are evil */
534 
535 		/* IPV6_IPSEC_POLICY, IPV6_XFRM_POLICY are not supported */
536 		/* IPV6_ADDR_PREFERENCES is not supported, we must be careful with subflows */
537 		return false;
538 	}
539 	if (level == SOL_TCP) {
540 		switch (optname) {
541 		/* the following are no-op or should work just fine */
542 		case TCP_THIN_DUPACK:
543 		case TCP_DEFER_ACCEPT:
544 
545 		/* the following need some love */
546 		case TCP_MAXSEG:
547 		case TCP_NODELAY:
548 		case TCP_THIN_LINEAR_TIMEOUTS:
549 		case TCP_CONGESTION:
550 		case TCP_CORK:
551 		case TCP_KEEPIDLE:
552 		case TCP_KEEPINTVL:
553 		case TCP_KEEPCNT:
554 		case TCP_SYNCNT:
555 		case TCP_SAVE_SYN:
556 		case TCP_LINGER2:
557 		case TCP_WINDOW_CLAMP:
558 		case TCP_QUICKACK:
559 		case TCP_USER_TIMEOUT:
560 		case TCP_TIMESTAMP:
561 		case TCP_NOTSENT_LOWAT:
562 		case TCP_TX_DELAY:
563 		case TCP_INQ:
564 		case TCP_FASTOPEN:
565 		case TCP_FASTOPEN_CONNECT:
566 		case TCP_FASTOPEN_KEY:
567 		case TCP_FASTOPEN_NO_COOKIE:
568 			return true;
569 		}
570 
571 		/* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
572 
573 		/* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
574 		 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
575 		 */
576 	}
577 	return false;
578 }
579 
580 static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
581 					       unsigned int optlen)
582 {
583 	struct mptcp_subflow_context *subflow;
584 	struct sock *sk = (struct sock *)msk;
585 	char name[TCP_CA_NAME_MAX];
586 	bool cap_net_admin;
587 	int ret;
588 
589 	if (optlen < 1)
590 		return -EINVAL;
591 
592 	ret = strncpy_from_sockptr(name, optval,
593 				   min_t(long, TCP_CA_NAME_MAX - 1, optlen));
594 	if (ret < 0)
595 		return -EFAULT;
596 
597 	name[ret] = 0;
598 
599 	cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
600 
601 	ret = 0;
602 	lock_sock(sk);
603 	sockopt_seq_inc(msk);
604 	mptcp_for_each_subflow(msk, subflow) {
605 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
606 		int err;
607 
608 		lock_sock(ssk);
609 		err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
610 		if (err < 0 && ret == 0)
611 			ret = err;
612 		subflow->setsockopt_seq = msk->setsockopt_seq;
613 		release_sock(ssk);
614 	}
615 
616 	if (ret == 0)
617 		strcpy(msk->ca_name, name);
618 
619 	release_sock(sk);
620 	return ret;
621 }
622 
623 static int mptcp_setsockopt_sol_tcp_cork(struct mptcp_sock *msk, sockptr_t optval,
624 					 unsigned int optlen)
625 {
626 	struct mptcp_subflow_context *subflow;
627 	struct sock *sk = (struct sock *)msk;
628 	int val;
629 
630 	if (optlen < sizeof(int))
631 		return -EINVAL;
632 
633 	if (copy_from_sockptr(&val, optval, sizeof(val)))
634 		return -EFAULT;
635 
636 	lock_sock(sk);
637 	sockopt_seq_inc(msk);
638 	msk->cork = !!val;
639 	mptcp_for_each_subflow(msk, subflow) {
640 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
641 
642 		lock_sock(ssk);
643 		__tcp_sock_set_cork(ssk, !!val);
644 		release_sock(ssk);
645 	}
646 	if (!val)
647 		mptcp_check_and_set_pending(sk);
648 	release_sock(sk);
649 
650 	return 0;
651 }
652 
653 static int mptcp_setsockopt_sol_tcp_nodelay(struct mptcp_sock *msk, sockptr_t optval,
654 					    unsigned int optlen)
655 {
656 	struct mptcp_subflow_context *subflow;
657 	struct sock *sk = (struct sock *)msk;
658 	int val;
659 
660 	if (optlen < sizeof(int))
661 		return -EINVAL;
662 
663 	if (copy_from_sockptr(&val, optval, sizeof(val)))
664 		return -EFAULT;
665 
666 	lock_sock(sk);
667 	sockopt_seq_inc(msk);
668 	msk->nodelay = !!val;
669 	mptcp_for_each_subflow(msk, subflow) {
670 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
671 
672 		lock_sock(ssk);
673 		__tcp_sock_set_nodelay(ssk, !!val);
674 		release_sock(ssk);
675 	}
676 	if (val)
677 		mptcp_check_and_set_pending(sk);
678 	release_sock(sk);
679 
680 	return 0;
681 }
682 
683 static int mptcp_setsockopt_sol_ip_set_transparent(struct mptcp_sock *msk, int optname,
684 						   sockptr_t optval, unsigned int optlen)
685 {
686 	struct sock *sk = (struct sock *)msk;
687 	struct inet_sock *issk;
688 	struct socket *ssock;
689 	int err;
690 
691 	err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
692 	if (err != 0)
693 		return err;
694 
695 	lock_sock(sk);
696 
697 	ssock = __mptcp_nmpc_socket(msk);
698 	if (IS_ERR(ssock)) {
699 		release_sock(sk);
700 		return PTR_ERR(ssock);
701 	}
702 
703 	issk = inet_sk(ssock->sk);
704 
705 	switch (optname) {
706 	case IP_FREEBIND:
707 		issk->freebind = inet_sk(sk)->freebind;
708 		break;
709 	case IP_TRANSPARENT:
710 		issk->transparent = inet_sk(sk)->transparent;
711 		break;
712 	default:
713 		release_sock(sk);
714 		WARN_ON_ONCE(1);
715 		return -EOPNOTSUPP;
716 	}
717 
718 	sockopt_seq_inc(msk);
719 	release_sock(sk);
720 	return 0;
721 }
722 
723 static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,
724 				       sockptr_t optval, unsigned int optlen)
725 {
726 	struct mptcp_subflow_context *subflow;
727 	struct sock *sk = (struct sock *)msk;
728 	int err, val;
729 
730 	err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
731 
732 	if (err != 0)
733 		return err;
734 
735 	lock_sock(sk);
736 	sockopt_seq_inc(msk);
737 	val = inet_sk(sk)->tos;
738 	mptcp_for_each_subflow(msk, subflow) {
739 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
740 
741 		__ip_sock_set_tos(ssk, val);
742 	}
743 	release_sock(sk);
744 
745 	return 0;
746 }
747 
748 static int mptcp_setsockopt_v4(struct mptcp_sock *msk, int optname,
749 			       sockptr_t optval, unsigned int optlen)
750 {
751 	switch (optname) {
752 	case IP_FREEBIND:
753 	case IP_TRANSPARENT:
754 		return mptcp_setsockopt_sol_ip_set_transparent(msk, optname, optval, optlen);
755 	case IP_TOS:
756 		return mptcp_setsockopt_v4_set_tos(msk, optname, optval, optlen);
757 	}
758 
759 	return -EOPNOTSUPP;
760 }
761 
762 static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
763 					  sockptr_t optval, unsigned int optlen)
764 {
765 	struct sock *sk = (struct sock *)msk;
766 	struct socket *sock;
767 	int ret;
768 
769 	/* Limit to first subflow, before the connection establishment */
770 	lock_sock(sk);
771 	sock = __mptcp_nmpc_socket(msk);
772 	if (IS_ERR(sock)) {
773 		ret = PTR_ERR(sock);
774 		goto unlock;
775 	}
776 
777 	ret = tcp_setsockopt(sock->sk, level, optname, optval, optlen);
778 
779 unlock:
780 	release_sock(sk);
781 	return ret;
782 }
783 
784 static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
785 				    sockptr_t optval, unsigned int optlen)
786 {
787 	struct sock *sk = (void *)msk;
788 	int ret, val;
789 
790 	switch (optname) {
791 	case TCP_INQ:
792 		ret = mptcp_get_int_option(msk, optval, optlen, &val);
793 		if (ret)
794 			return ret;
795 		if (val < 0 || val > 1)
796 			return -EINVAL;
797 
798 		lock_sock(sk);
799 		msk->recvmsg_inq = !!val;
800 		release_sock(sk);
801 		return 0;
802 	case TCP_ULP:
803 		return -EOPNOTSUPP;
804 	case TCP_CONGESTION:
805 		return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
806 	case TCP_CORK:
807 		return mptcp_setsockopt_sol_tcp_cork(msk, optval, optlen);
808 	case TCP_NODELAY:
809 		return mptcp_setsockopt_sol_tcp_nodelay(msk, optval, optlen);
810 	case TCP_DEFER_ACCEPT:
811 		/* See tcp.c: TCP_DEFER_ACCEPT does not fail */
812 		mptcp_setsockopt_first_sf_only(msk, SOL_TCP, optname, optval, optlen);
813 		return 0;
814 	case TCP_FASTOPEN:
815 	case TCP_FASTOPEN_CONNECT:
816 	case TCP_FASTOPEN_KEY:
817 	case TCP_FASTOPEN_NO_COOKIE:
818 		return mptcp_setsockopt_first_sf_only(msk, SOL_TCP, optname,
819 						      optval, optlen);
820 	}
821 
822 	return -EOPNOTSUPP;
823 }
824 
825 int mptcp_setsockopt(struct sock *sk, int level, int optname,
826 		     sockptr_t optval, unsigned int optlen)
827 {
828 	struct mptcp_sock *msk = mptcp_sk(sk);
829 	struct sock *ssk;
830 
831 	pr_debug("msk=%p", msk);
832 
833 	if (level == SOL_SOCKET)
834 		return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
835 
836 	if (!mptcp_supported_sockopt(level, optname))
837 		return -ENOPROTOOPT;
838 
839 	/* @@ the meaning of setsockopt() when the socket is connected and
840 	 * there are multiple subflows is not yet defined. It is up to the
841 	 * MPTCP-level socket to configure the subflows until the subflow
842 	 * is in TCP fallback, when TCP socket options are passed through
843 	 * to the one remaining subflow.
844 	 */
845 	lock_sock(sk);
846 	ssk = __mptcp_tcp_fallback(msk);
847 	release_sock(sk);
848 	if (ssk)
849 		return tcp_setsockopt(ssk, level, optname, optval, optlen);
850 
851 	if (level == SOL_IP)
852 		return mptcp_setsockopt_v4(msk, optname, optval, optlen);
853 
854 	if (level == SOL_IPV6)
855 		return mptcp_setsockopt_v6(msk, optname, optval, optlen);
856 
857 	if (level == SOL_TCP)
858 		return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
859 
860 	return -EOPNOTSUPP;
861 }
862 
863 static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
864 					  char __user *optval, int __user *optlen)
865 {
866 	struct sock *sk = (struct sock *)msk;
867 	struct socket *ssock;
868 	int ret;
869 	struct sock *ssk;
870 
871 	lock_sock(sk);
872 	ssk = msk->first;
873 	if (ssk) {
874 		ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
875 		goto out;
876 	}
877 
878 	ssock = __mptcp_nmpc_socket(msk);
879 	if (IS_ERR(ssock)) {
880 		ret = PTR_ERR(ssock);
881 		goto out;
882 	}
883 
884 	ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);
885 
886 out:
887 	release_sock(sk);
888 	return ret;
889 }
890 
891 void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
892 {
893 	struct sock *sk = (struct sock *)msk;
894 	u32 flags = 0;
895 	bool slow;
896 
897 	memset(info, 0, sizeof(*info));
898 
899 	info->mptcpi_subflows = READ_ONCE(msk->pm.subflows);
900 	info->mptcpi_add_addr_signal = READ_ONCE(msk->pm.add_addr_signaled);
901 	info->mptcpi_add_addr_accepted = READ_ONCE(msk->pm.add_addr_accepted);
902 	info->mptcpi_local_addr_used = READ_ONCE(msk->pm.local_addr_used);
903 
904 	if (inet_sk_state_load(sk) == TCP_LISTEN)
905 		return;
906 
907 	/* The following limits only make sense for the in-kernel PM */
908 	if (mptcp_pm_is_kernel(msk)) {
909 		info->mptcpi_subflows_max =
910 			mptcp_pm_get_subflows_max(msk);
911 		info->mptcpi_add_addr_signal_max =
912 			mptcp_pm_get_add_addr_signal_max(msk);
913 		info->mptcpi_add_addr_accepted_max =
914 			mptcp_pm_get_add_addr_accept_max(msk);
915 		info->mptcpi_local_addr_max =
916 			mptcp_pm_get_local_addr_max(msk);
917 	}
918 
919 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags))
920 		flags |= MPTCP_INFO_FLAG_FALLBACK;
921 	if (READ_ONCE(msk->can_ack))
922 		flags |= MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED;
923 	info->mptcpi_flags = flags;
924 	mptcp_data_lock(sk);
925 	info->mptcpi_snd_una = msk->snd_una;
926 	info->mptcpi_rcv_nxt = msk->ack_seq;
927 	info->mptcpi_bytes_acked = msk->bytes_acked;
928 	mptcp_data_unlock(sk);
929 
930 	slow = lock_sock_fast(sk);
931 	info->mptcpi_csum_enabled = msk->csum_enabled;
932 	info->mptcpi_token = msk->token;
933 	info->mptcpi_write_seq = msk->write_seq;
934 	info->mptcpi_retransmits = inet_csk(sk)->icsk_retransmits;
935 	info->mptcpi_bytes_sent = msk->bytes_sent;
936 	info->mptcpi_bytes_received = msk->bytes_received;
937 	info->mptcpi_bytes_retrans = msk->bytes_retrans;
938 	unlock_sock_fast(sk, slow);
939 }
940 EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
941 
942 static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
943 {
944 	struct mptcp_info m_info;
945 	int len;
946 
947 	if (get_user(len, optlen))
948 		return -EFAULT;
949 
950 	len = min_t(unsigned int, len, sizeof(struct mptcp_info));
951 
952 	mptcp_diag_fill_info(msk, &m_info);
953 
954 	if (put_user(len, optlen))
955 		return -EFAULT;
956 
957 	if (copy_to_user(optval, &m_info, len))
958 		return -EFAULT;
959 
960 	return 0;
961 }
962 
963 static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
964 				  char __user *optval,
965 				  u32 copied,
966 				  int __user *optlen)
967 {
968 	u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
969 
970 	if (copied)
971 		copied += sfd->size_subflow_data;
972 	else
973 		copied = copylen;
974 
975 	if (put_user(copied, optlen))
976 		return -EFAULT;
977 
978 	if (copy_to_user(optval, sfd, copylen))
979 		return -EFAULT;
980 
981 	return 0;
982 }
983 
984 static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
985 				  char __user *optval,
986 				  int __user *optlen)
987 {
988 	int len, copylen;
989 
990 	if (get_user(len, optlen))
991 		return -EFAULT;
992 
993 	/* if mptcp_subflow_data size is changed, need to adjust
994 	 * this function to deal with programs using old version.
995 	 */
996 	BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
997 
998 	if (len < MIN_INFO_OPTLEN_SIZE)
999 		return -EINVAL;
1000 
1001 	memset(sfd, 0, sizeof(*sfd));
1002 
1003 	copylen = min_t(unsigned int, len, sizeof(*sfd));
1004 	if (copy_from_user(sfd, optval, copylen))
1005 		return -EFAULT;
1006 
1007 	/* size_subflow_data is u32, but len is signed */
1008 	if (sfd->size_subflow_data > INT_MAX ||
1009 	    sfd->size_user > INT_MAX)
1010 		return -EINVAL;
1011 
1012 	if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
1013 	    sfd->size_subflow_data > len)
1014 		return -EINVAL;
1015 
1016 	if (sfd->num_subflows || sfd->size_kernel)
1017 		return -EINVAL;
1018 
1019 	return len - sfd->size_subflow_data;
1020 }
1021 
1022 static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
1023 				    int __user *optlen)
1024 {
1025 	struct mptcp_subflow_context *subflow;
1026 	struct sock *sk = (struct sock *)msk;
1027 	unsigned int sfcount = 0, copied = 0;
1028 	struct mptcp_subflow_data sfd;
1029 	char __user *infoptr;
1030 	int len;
1031 
1032 	len = mptcp_get_subflow_data(&sfd, optval, optlen);
1033 	if (len < 0)
1034 		return len;
1035 
1036 	sfd.size_kernel = sizeof(struct tcp_info);
1037 	sfd.size_user = min_t(unsigned int, sfd.size_user,
1038 			      sizeof(struct tcp_info));
1039 
1040 	infoptr = optval + sfd.size_subflow_data;
1041 
1042 	lock_sock(sk);
1043 
1044 	mptcp_for_each_subflow(msk, subflow) {
1045 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1046 
1047 		++sfcount;
1048 
1049 		if (len && len >= sfd.size_user) {
1050 			struct tcp_info info;
1051 
1052 			tcp_get_info(ssk, &info);
1053 
1054 			if (copy_to_user(infoptr, &info, sfd.size_user)) {
1055 				release_sock(sk);
1056 				return -EFAULT;
1057 			}
1058 
1059 			infoptr += sfd.size_user;
1060 			copied += sfd.size_user;
1061 			len -= sfd.size_user;
1062 		}
1063 	}
1064 
1065 	release_sock(sk);
1066 
1067 	sfd.num_subflows = sfcount;
1068 
1069 	if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1070 		return -EFAULT;
1071 
1072 	return 0;
1073 }
1074 
1075 static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
1076 {
1077 	const struct inet_sock *inet = inet_sk(sk);
1078 
1079 	memset(a, 0, sizeof(*a));
1080 
1081 	if (sk->sk_family == AF_INET) {
1082 		a->sin_local.sin_family = AF_INET;
1083 		a->sin_local.sin_port = inet->inet_sport;
1084 		a->sin_local.sin_addr.s_addr = inet->inet_rcv_saddr;
1085 
1086 		if (!a->sin_local.sin_addr.s_addr)
1087 			a->sin_local.sin_addr.s_addr = inet->inet_saddr;
1088 
1089 		a->sin_remote.sin_family = AF_INET;
1090 		a->sin_remote.sin_port = inet->inet_dport;
1091 		a->sin_remote.sin_addr.s_addr = inet->inet_daddr;
1092 #if IS_ENABLED(CONFIG_IPV6)
1093 	} else if (sk->sk_family == AF_INET6) {
1094 		const struct ipv6_pinfo *np = inet6_sk(sk);
1095 
1096 		if (WARN_ON_ONCE(!np))
1097 			return;
1098 
1099 		a->sin6_local.sin6_family = AF_INET6;
1100 		a->sin6_local.sin6_port = inet->inet_sport;
1101 
1102 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
1103 			a->sin6_local.sin6_addr = np->saddr;
1104 		else
1105 			a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
1106 
1107 		a->sin6_remote.sin6_family = AF_INET6;
1108 		a->sin6_remote.sin6_port = inet->inet_dport;
1109 		a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
1110 #endif
1111 	}
1112 }
1113 
1114 static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
1115 					  int __user *optlen)
1116 {
1117 	struct mptcp_subflow_context *subflow;
1118 	struct sock *sk = (struct sock *)msk;
1119 	unsigned int sfcount = 0, copied = 0;
1120 	struct mptcp_subflow_data sfd;
1121 	char __user *addrptr;
1122 	int len;
1123 
1124 	len = mptcp_get_subflow_data(&sfd, optval, optlen);
1125 	if (len < 0)
1126 		return len;
1127 
1128 	sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
1129 	sfd.size_user = min_t(unsigned int, sfd.size_user,
1130 			      sizeof(struct mptcp_subflow_addrs));
1131 
1132 	addrptr = optval + sfd.size_subflow_data;
1133 
1134 	lock_sock(sk);
1135 
1136 	mptcp_for_each_subflow(msk, subflow) {
1137 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1138 
1139 		++sfcount;
1140 
1141 		if (len && len >= sfd.size_user) {
1142 			struct mptcp_subflow_addrs a;
1143 
1144 			mptcp_get_sub_addrs(ssk, &a);
1145 
1146 			if (copy_to_user(addrptr, &a, sfd.size_user)) {
1147 				release_sock(sk);
1148 				return -EFAULT;
1149 			}
1150 
1151 			addrptr += sfd.size_user;
1152 			copied += sfd.size_user;
1153 			len -= sfd.size_user;
1154 		}
1155 	}
1156 
1157 	release_sock(sk);
1158 
1159 	sfd.num_subflows = sfcount;
1160 
1161 	if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1162 		return -EFAULT;
1163 
1164 	return 0;
1165 }
1166 
1167 static int mptcp_get_full_info(struct mptcp_full_info *mfi,
1168 			       char __user *optval,
1169 			       int __user *optlen)
1170 {
1171 	int len;
1172 
1173 	BUILD_BUG_ON(offsetof(struct mptcp_full_info, mptcp_info) !=
1174 		     MIN_FULL_INFO_OPTLEN_SIZE);
1175 
1176 	if (get_user(len, optlen))
1177 		return -EFAULT;
1178 
1179 	if (len < MIN_FULL_INFO_OPTLEN_SIZE)
1180 		return -EINVAL;
1181 
1182 	memset(mfi, 0, sizeof(*mfi));
1183 	if (copy_from_user(mfi, optval, MIN_FULL_INFO_OPTLEN_SIZE))
1184 		return -EFAULT;
1185 
1186 	if (mfi->size_tcpinfo_kernel ||
1187 	    mfi->size_sfinfo_kernel ||
1188 	    mfi->num_subflows)
1189 		return -EINVAL;
1190 
1191 	if (mfi->size_sfinfo_user > INT_MAX ||
1192 	    mfi->size_tcpinfo_user > INT_MAX)
1193 		return -EINVAL;
1194 
1195 	return len - MIN_FULL_INFO_OPTLEN_SIZE;
1196 }
1197 
1198 static int mptcp_put_full_info(struct mptcp_full_info *mfi,
1199 			       char __user *optval,
1200 			       u32 copylen,
1201 			       int __user *optlen)
1202 {
1203 	copylen += MIN_FULL_INFO_OPTLEN_SIZE;
1204 	if (put_user(copylen, optlen))
1205 		return -EFAULT;
1206 
1207 	if (copy_to_user(optval, mfi, copylen))
1208 		return -EFAULT;
1209 	return 0;
1210 }
1211 
1212 static int mptcp_getsockopt_full_info(struct mptcp_sock *msk, char __user *optval,
1213 				      int __user *optlen)
1214 {
1215 	unsigned int sfcount = 0, copylen = 0;
1216 	struct mptcp_subflow_context *subflow;
1217 	struct sock *sk = (struct sock *)msk;
1218 	void __user *tcpinfoptr, *sfinfoptr;
1219 	struct mptcp_full_info mfi;
1220 	int len;
1221 
1222 	len = mptcp_get_full_info(&mfi, optval, optlen);
1223 	if (len < 0)
1224 		return len;
1225 
1226 	/* don't bother filling the mptcp info if there is not enough
1227 	 * user-space-provided storage
1228 	 */
1229 	if (len > 0) {
1230 		mptcp_diag_fill_info(msk, &mfi.mptcp_info);
1231 		copylen += min_t(unsigned int, len, sizeof(struct mptcp_info));
1232 	}
1233 
1234 	mfi.size_tcpinfo_kernel = sizeof(struct tcp_info);
1235 	mfi.size_tcpinfo_user = min_t(unsigned int, mfi.size_tcpinfo_user,
1236 				      sizeof(struct tcp_info));
1237 	sfinfoptr = u64_to_user_ptr(mfi.subflow_info);
1238 	mfi.size_sfinfo_kernel = sizeof(struct mptcp_subflow_info);
1239 	mfi.size_sfinfo_user = min_t(unsigned int, mfi.size_sfinfo_user,
1240 				     sizeof(struct mptcp_subflow_info));
1241 	tcpinfoptr = u64_to_user_ptr(mfi.tcp_info);
1242 
1243 	lock_sock(sk);
1244 	mptcp_for_each_subflow(msk, subflow) {
1245 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1246 		struct mptcp_subflow_info sfinfo;
1247 		struct tcp_info tcp_info;
1248 
1249 		if (sfcount++ >= mfi.size_arrays_user)
1250 			continue;
1251 
1252 		/* fetch addr/tcp_info only if the user space buffers
1253 		 * are wide enough
1254 		 */
1255 		memset(&sfinfo, 0, sizeof(sfinfo));
1256 		sfinfo.id = subflow->subflow_id;
1257 		if (mfi.size_sfinfo_user >
1258 		    offsetof(struct mptcp_subflow_info, addrs))
1259 			mptcp_get_sub_addrs(ssk, &sfinfo.addrs);
1260 		if (copy_to_user(sfinfoptr, &sfinfo, mfi.size_sfinfo_user))
1261 			goto fail_release;
1262 
1263 		if (mfi.size_tcpinfo_user) {
1264 			tcp_get_info(ssk, &tcp_info);
1265 			if (copy_to_user(tcpinfoptr, &tcp_info,
1266 					 mfi.size_tcpinfo_user))
1267 				goto fail_release;
1268 		}
1269 
1270 		tcpinfoptr += mfi.size_tcpinfo_user;
1271 		sfinfoptr += mfi.size_sfinfo_user;
1272 	}
1273 	release_sock(sk);
1274 
1275 	mfi.num_subflows = sfcount;
1276 	if (mptcp_put_full_info(&mfi, optval, copylen, optlen))
1277 		return -EFAULT;
1278 
1279 	return 0;
1280 
1281 fail_release:
1282 	release_sock(sk);
1283 	return -EFAULT;
1284 }
1285 
1286 static int mptcp_put_int_option(struct mptcp_sock *msk, char __user *optval,
1287 				int __user *optlen, int val)
1288 {
1289 	int len;
1290 
1291 	if (get_user(len, optlen))
1292 		return -EFAULT;
1293 	if (len < 0)
1294 		return -EINVAL;
1295 
1296 	if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1297 		unsigned char ucval = (unsigned char)val;
1298 
1299 		len = 1;
1300 		if (put_user(len, optlen))
1301 			return -EFAULT;
1302 		if (copy_to_user(optval, &ucval, 1))
1303 			return -EFAULT;
1304 	} else {
1305 		len = min_t(unsigned int, len, sizeof(int));
1306 		if (put_user(len, optlen))
1307 			return -EFAULT;
1308 		if (copy_to_user(optval, &val, len))
1309 			return -EFAULT;
1310 	}
1311 
1312 	return 0;
1313 }
1314 
1315 static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
1316 				    char __user *optval, int __user *optlen)
1317 {
1318 	switch (optname) {
1319 	case TCP_ULP:
1320 	case TCP_CONGESTION:
1321 	case TCP_INFO:
1322 	case TCP_CC_INFO:
1323 	case TCP_DEFER_ACCEPT:
1324 	case TCP_FASTOPEN:
1325 	case TCP_FASTOPEN_CONNECT:
1326 	case TCP_FASTOPEN_KEY:
1327 	case TCP_FASTOPEN_NO_COOKIE:
1328 		return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
1329 						      optval, optlen);
1330 	case TCP_INQ:
1331 		return mptcp_put_int_option(msk, optval, optlen, msk->recvmsg_inq);
1332 	case TCP_CORK:
1333 		return mptcp_put_int_option(msk, optval, optlen, msk->cork);
1334 	case TCP_NODELAY:
1335 		return mptcp_put_int_option(msk, optval, optlen, msk->nodelay);
1336 	}
1337 	return -EOPNOTSUPP;
1338 }
1339 
1340 static int mptcp_getsockopt_v4(struct mptcp_sock *msk, int optname,
1341 			       char __user *optval, int __user *optlen)
1342 {
1343 	struct sock *sk = (void *)msk;
1344 
1345 	switch (optname) {
1346 	case IP_TOS:
1347 		return mptcp_put_int_option(msk, optval, optlen, inet_sk(sk)->tos);
1348 	}
1349 
1350 	return -EOPNOTSUPP;
1351 }
1352 
1353 static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
1354 				      char __user *optval, int __user *optlen)
1355 {
1356 	switch (optname) {
1357 	case MPTCP_INFO:
1358 		return mptcp_getsockopt_info(msk, optval, optlen);
1359 	case MPTCP_FULL_INFO:
1360 		return mptcp_getsockopt_full_info(msk, optval, optlen);
1361 	case MPTCP_TCPINFO:
1362 		return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
1363 	case MPTCP_SUBFLOW_ADDRS:
1364 		return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
1365 	}
1366 
1367 	return -EOPNOTSUPP;
1368 }
1369 
1370 int mptcp_getsockopt(struct sock *sk, int level, int optname,
1371 		     char __user *optval, int __user *option)
1372 {
1373 	struct mptcp_sock *msk = mptcp_sk(sk);
1374 	struct sock *ssk;
1375 
1376 	pr_debug("msk=%p", msk);
1377 
1378 	/* @@ the meaning of setsockopt() when the socket is connected and
1379 	 * there are multiple subflows is not yet defined. It is up to the
1380 	 * MPTCP-level socket to configure the subflows until the subflow
1381 	 * is in TCP fallback, when socket options are passed through
1382 	 * to the one remaining subflow.
1383 	 */
1384 	lock_sock(sk);
1385 	ssk = __mptcp_tcp_fallback(msk);
1386 	release_sock(sk);
1387 	if (ssk)
1388 		return tcp_getsockopt(ssk, level, optname, optval, option);
1389 
1390 	if (level == SOL_IP)
1391 		return mptcp_getsockopt_v4(msk, optname, optval, option);
1392 	if (level == SOL_TCP)
1393 		return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
1394 	if (level == SOL_MPTCP)
1395 		return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
1396 	return -EOPNOTSUPP;
1397 }
1398 
1399 static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
1400 {
1401 	static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
1402 	struct sock *sk = (struct sock *)msk;
1403 
1404 	if (ssk->sk_prot->keepalive) {
1405 		if (sock_flag(sk, SOCK_KEEPOPEN))
1406 			ssk->sk_prot->keepalive(ssk, 1);
1407 		else
1408 			ssk->sk_prot->keepalive(ssk, 0);
1409 	}
1410 
1411 	ssk->sk_priority = sk->sk_priority;
1412 	ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1413 	ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
1414 	ssk->sk_ipv6only = sk->sk_ipv6only;
1415 	__ip_sock_set_tos(ssk, inet_sk(sk)->tos);
1416 
1417 	if (sk->sk_userlocks & tx_rx_locks) {
1418 		ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1419 		if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
1420 			WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1421 		if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1422 			WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1423 	}
1424 
1425 	if (sock_flag(sk, SOCK_LINGER)) {
1426 		ssk->sk_lingertime = sk->sk_lingertime;
1427 		sock_set_flag(ssk, SOCK_LINGER);
1428 	} else {
1429 		sock_reset_flag(ssk, SOCK_LINGER);
1430 	}
1431 
1432 	if (sk->sk_mark != ssk->sk_mark) {
1433 		ssk->sk_mark = sk->sk_mark;
1434 		sk_dst_reset(ssk);
1435 	}
1436 
1437 	sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1438 
1439 	if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
1440 		tcp_set_congestion_control(ssk, msk->ca_name, false, true);
1441 	__tcp_sock_set_cork(ssk, !!msk->cork);
1442 	__tcp_sock_set_nodelay(ssk, !!msk->nodelay);
1443 
1444 	inet_sk(ssk)->transparent = inet_sk(sk)->transparent;
1445 	inet_sk(ssk)->freebind = inet_sk(sk)->freebind;
1446 }
1447 
1448 static void __mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1449 {
1450 	bool slow = lock_sock_fast(ssk);
1451 
1452 	sync_socket_options(msk, ssk);
1453 
1454 	unlock_sock_fast(ssk, slow);
1455 }
1456 
1457 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1458 {
1459 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1460 
1461 	msk_owned_by_me(msk);
1462 
1463 	if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1464 		__mptcp_sockopt_sync(msk, ssk);
1465 
1466 		subflow->setsockopt_seq = msk->setsockopt_seq;
1467 	}
1468 }
1469 
1470 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk)
1471 {
1472 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1473 
1474 	msk_owned_by_me(msk);
1475 
1476 	if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1477 		sync_socket_options(msk, ssk);
1478 
1479 		subflow->setsockopt_seq = msk->setsockopt_seq;
1480 	}
1481 }
1482