xref: /linux/net/mptcp/sockopt.c (revision 662f11d55ffd02933e1bd275d732b97eddccf870)
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 
19 static struct sock *__mptcp_tcp_fallback(struct mptcp_sock *msk)
20 {
21 	sock_owned_by_me((const struct sock *)msk);
22 
23 	if (likely(!__mptcp_check_fallback(msk)))
24 		return NULL;
25 
26 	return msk->first;
27 }
28 
29 static u32 sockopt_seq_reset(const struct sock *sk)
30 {
31 	sock_owned_by_me(sk);
32 
33 	/* Highbits contain state.  Allows to distinguish sockopt_seq
34 	 * of listener and established:
35 	 * s0 = new_listener()
36 	 * sockopt(s0) - seq is 1
37 	 * s1 = accept(s0) - s1 inherits seq 1 if listener sk (s0)
38 	 * sockopt(s0) - seq increments to 2 on s0
39 	 * sockopt(s1) // seq increments to 2 on s1 (different option)
40 	 * new ssk completes join, inherits options from s0 // seq 2
41 	 * Needs sync from mptcp join logic, but ssk->seq == msk->seq
42 	 *
43 	 * Set High order bits to sk_state so ssk->seq == msk->seq test
44 	 * will fail.
45 	 */
46 
47 	return (u32)sk->sk_state << 24u;
48 }
49 
50 static void sockopt_seq_inc(struct mptcp_sock *msk)
51 {
52 	u32 seq = (msk->setsockopt_seq + 1) & 0x00ffffff;
53 
54 	msk->setsockopt_seq = sockopt_seq_reset((struct sock *)msk) + seq;
55 }
56 
57 static int mptcp_get_int_option(struct mptcp_sock *msk, sockptr_t optval,
58 				unsigned int optlen, int *val)
59 {
60 	if (optlen < sizeof(int))
61 		return -EINVAL;
62 
63 	if (copy_from_sockptr(val, optval, sizeof(*val)))
64 		return -EFAULT;
65 
66 	return 0;
67 }
68 
69 static void mptcp_sol_socket_sync_intval(struct mptcp_sock *msk, int optname, int val)
70 {
71 	struct mptcp_subflow_context *subflow;
72 	struct sock *sk = (struct sock *)msk;
73 
74 	lock_sock(sk);
75 	sockopt_seq_inc(msk);
76 
77 	mptcp_for_each_subflow(msk, subflow) {
78 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
79 		bool slow = lock_sock_fast(ssk);
80 
81 		switch (optname) {
82 		case SO_DEBUG:
83 			sock_valbool_flag(ssk, SOCK_DBG, !!val);
84 			break;
85 		case SO_KEEPALIVE:
86 			if (ssk->sk_prot->keepalive)
87 				ssk->sk_prot->keepalive(ssk, !!val);
88 			sock_valbool_flag(ssk, SOCK_KEEPOPEN, !!val);
89 			break;
90 		case SO_PRIORITY:
91 			ssk->sk_priority = val;
92 			break;
93 		case SO_SNDBUF:
94 		case SO_SNDBUFFORCE:
95 			ssk->sk_userlocks |= SOCK_SNDBUF_LOCK;
96 			WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
97 			break;
98 		case SO_RCVBUF:
99 		case SO_RCVBUFFORCE:
100 			ssk->sk_userlocks |= SOCK_RCVBUF_LOCK;
101 			WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
102 			break;
103 		case SO_MARK:
104 			if (READ_ONCE(ssk->sk_mark) != sk->sk_mark) {
105 				ssk->sk_mark = sk->sk_mark;
106 				sk_dst_reset(ssk);
107 			}
108 			break;
109 		case SO_INCOMING_CPU:
110 			WRITE_ONCE(ssk->sk_incoming_cpu, val);
111 			break;
112 		}
113 
114 		subflow->setsockopt_seq = msk->setsockopt_seq;
115 		unlock_sock_fast(ssk, slow);
116 	}
117 
118 	release_sock(sk);
119 }
120 
121 static int mptcp_sol_socket_intval(struct mptcp_sock *msk, int optname, int val)
122 {
123 	sockptr_t optval = KERNEL_SOCKPTR(&val);
124 	struct sock *sk = (struct sock *)msk;
125 	int ret;
126 
127 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
128 			      optval, sizeof(val));
129 	if (ret)
130 		return ret;
131 
132 	mptcp_sol_socket_sync_intval(msk, optname, val);
133 	return 0;
134 }
135 
136 static void mptcp_so_incoming_cpu(struct mptcp_sock *msk, int val)
137 {
138 	struct sock *sk = (struct sock *)msk;
139 
140 	WRITE_ONCE(sk->sk_incoming_cpu, val);
141 
142 	mptcp_sol_socket_sync_intval(msk, SO_INCOMING_CPU, val);
143 }
144 
145 static int mptcp_setsockopt_sol_socket_tstamp(struct mptcp_sock *msk, int optname, int val)
146 {
147 	sockptr_t optval = KERNEL_SOCKPTR(&val);
148 	struct mptcp_subflow_context *subflow;
149 	struct sock *sk = (struct sock *)msk;
150 	int ret;
151 
152 	ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
153 			      optval, sizeof(val));
154 	if (ret)
155 		return ret;
156 
157 	lock_sock(sk);
158 	mptcp_for_each_subflow(msk, subflow) {
159 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
160 		bool slow = lock_sock_fast(ssk);
161 
162 		sock_set_timestamp(sk, optname, !!val);
163 		unlock_sock_fast(ssk, slow);
164 	}
165 
166 	release_sock(sk);
167 	return 0;
168 }
169 
170 static int mptcp_setsockopt_sol_socket_int(struct mptcp_sock *msk, int optname,
171 					   sockptr_t optval,
172 					   unsigned int optlen)
173 {
174 	int val, ret;
175 
176 	ret = mptcp_get_int_option(msk, optval, optlen, &val);
177 	if (ret)
178 		return ret;
179 
180 	switch (optname) {
181 	case SO_KEEPALIVE:
182 		mptcp_sol_socket_sync_intval(msk, optname, val);
183 		return 0;
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 socket *ssock;
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 		ssock = __mptcp_nmpc_socket(msk);
304 		if (!ssock) {
305 			release_sock(sk);
306 			return -EINVAL;
307 		}
308 
309 		ret = sock_setsockopt(ssock, SOL_SOCKET, optname, optval, optlen);
310 		if (ret == 0) {
311 			if (optname == SO_REUSEPORT)
312 				sk->sk_reuseport = ssock->sk->sk_reuseport;
313 			else if (optname == SO_REUSEADDR)
314 				sk->sk_reuse = ssock->sk->sk_reuse;
315 			else if (optname == SO_BINDTODEVICE)
316 				sk->sk_bound_dev_if = ssock->sk->sk_bound_dev_if;
317 			else if (optname == SO_BINDTOIFINDEX)
318 				sk->sk_bound_dev_if = ssock->sk->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_BUSY_POLL:
347 	case SO_PREFER_BUSY_POLL:
348 	case SO_BUSY_POLL_BUDGET:
349 		/* No need to copy: only relevant for msk */
350 		return sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname, optval, optlen);
351 	case SO_NO_CHECK:
352 	case SO_DONTROUTE:
353 	case SO_BROADCAST:
354 	case SO_BSDCOMPAT:
355 	case SO_PASSCRED:
356 	case SO_PASSSEC:
357 	case SO_RXQ_OVFL:
358 	case SO_WIFI_STATUS:
359 	case SO_NOFCS:
360 	case SO_SELECT_ERR_QUEUE:
361 		return 0;
362 	}
363 
364 	/* SO_OOBINLINE is not supported, let's avoid the related mess
365 	 * SO_ATTACH_FILTER, SO_ATTACH_BPF, SO_ATTACH_REUSEPORT_CBPF,
366 	 * SO_DETACH_REUSEPORT_BPF, SO_DETACH_FILTER, SO_LOCK_FILTER,
367 	 * we must be careful with subflows
368 	 *
369 	 * SO_ATTACH_REUSEPORT_EBPF is not supported, at it checks
370 	 * explicitly the sk_protocol field
371 	 *
372 	 * SO_PEEK_OFF is unsupported, as it is for plain TCP
373 	 * SO_MAX_PACING_RATE is unsupported, we must be careful with subflows
374 	 * SO_CNX_ADVICE is currently unsupported, could possibly be relevant,
375 	 * but likely needs careful design
376 	 *
377 	 * SO_ZEROCOPY is currently unsupported, TODO in sndmsg
378 	 * SO_TXTIME is currently unsupported
379 	 */
380 
381 	return -EOPNOTSUPP;
382 }
383 
384 static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,
385 			       sockptr_t optval, unsigned int optlen)
386 {
387 	struct sock *sk = (struct sock *)msk;
388 	int ret = -EOPNOTSUPP;
389 	struct socket *ssock;
390 
391 	switch (optname) {
392 	case IPV6_V6ONLY:
393 		lock_sock(sk);
394 		ssock = __mptcp_nmpc_socket(msk);
395 		if (!ssock) {
396 			release_sock(sk);
397 			return -EINVAL;
398 		}
399 
400 		ret = tcp_setsockopt(ssock->sk, SOL_IPV6, optname, optval, optlen);
401 		if (ret == 0)
402 			sk->sk_ipv6only = ssock->sk->sk_ipv6only;
403 
404 		release_sock(sk);
405 		break;
406 	}
407 
408 	return ret;
409 }
410 
411 static bool mptcp_supported_sockopt(int level, int optname)
412 {
413 	if (level == SOL_IP) {
414 		switch (optname) {
415 		/* should work fine */
416 		case IP_FREEBIND:
417 		case IP_TRANSPARENT:
418 
419 		/* the following are control cmsg related */
420 		case IP_PKTINFO:
421 		case IP_RECVTTL:
422 		case IP_RECVTOS:
423 		case IP_RECVOPTS:
424 		case IP_RETOPTS:
425 		case IP_PASSSEC:
426 		case IP_RECVORIGDSTADDR:
427 		case IP_CHECKSUM:
428 		case IP_RECVFRAGSIZE:
429 
430 		/* common stuff that need some love */
431 		case IP_TOS:
432 		case IP_TTL:
433 		case IP_BIND_ADDRESS_NO_PORT:
434 		case IP_MTU_DISCOVER:
435 		case IP_RECVERR:
436 
437 		/* possibly less common may deserve some love */
438 		case IP_MINTTL:
439 
440 		/* the following is apparently a no-op for plain TCP */
441 		case IP_RECVERR_RFC4884:
442 			return true;
443 		}
444 
445 		/* IP_OPTIONS is not supported, needs subflow care */
446 		/* IP_HDRINCL, IP_NODEFRAG are not supported, RAW specific */
447 		/* IP_MULTICAST_TTL, IP_MULTICAST_LOOP, IP_UNICAST_IF,
448 		 * IP_ADD_MEMBERSHIP, IP_ADD_SOURCE_MEMBERSHIP, IP_DROP_MEMBERSHIP,
449 		 * IP_DROP_SOURCE_MEMBERSHIP, IP_BLOCK_SOURCE, IP_UNBLOCK_SOURCE,
450 		 * MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP MCAST_JOIN_SOURCE_GROUP,
451 		 * MCAST_LEAVE_SOURCE_GROUP, MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE,
452 		 * MCAST_MSFILTER, IP_MULTICAST_ALL are not supported, better not deal
453 		 * with mcast stuff
454 		 */
455 		/* IP_IPSEC_POLICY, IP_XFRM_POLICY are nut supported, unrelated here */
456 		return false;
457 	}
458 	if (level == SOL_IPV6) {
459 		switch (optname) {
460 		case IPV6_V6ONLY:
461 
462 		/* the following are control cmsg related */
463 		case IPV6_RECVPKTINFO:
464 		case IPV6_2292PKTINFO:
465 		case IPV6_RECVHOPLIMIT:
466 		case IPV6_2292HOPLIMIT:
467 		case IPV6_RECVRTHDR:
468 		case IPV6_2292RTHDR:
469 		case IPV6_RECVHOPOPTS:
470 		case IPV6_2292HOPOPTS:
471 		case IPV6_RECVDSTOPTS:
472 		case IPV6_2292DSTOPTS:
473 		case IPV6_RECVTCLASS:
474 		case IPV6_FLOWINFO:
475 		case IPV6_RECVPATHMTU:
476 		case IPV6_RECVORIGDSTADDR:
477 		case IPV6_RECVFRAGSIZE:
478 
479 		/* the following ones need some love but are quite common */
480 		case IPV6_TCLASS:
481 		case IPV6_TRANSPARENT:
482 		case IPV6_FREEBIND:
483 		case IPV6_PKTINFO:
484 		case IPV6_2292PKTOPTIONS:
485 		case IPV6_UNICAST_HOPS:
486 		case IPV6_MTU_DISCOVER:
487 		case IPV6_MTU:
488 		case IPV6_RECVERR:
489 		case IPV6_FLOWINFO_SEND:
490 		case IPV6_FLOWLABEL_MGR:
491 		case IPV6_MINHOPCOUNT:
492 		case IPV6_DONTFRAG:
493 		case IPV6_AUTOFLOWLABEL:
494 
495 		/* the following one is a no-op for plain TCP */
496 		case IPV6_RECVERR_RFC4884:
497 			return true;
498 		}
499 
500 		/* IPV6_HOPOPTS, IPV6_RTHDRDSTOPTS, IPV6_RTHDR, IPV6_DSTOPTS are
501 		 * not supported
502 		 */
503 		/* IPV6_MULTICAST_HOPS, IPV6_MULTICAST_LOOP, IPV6_UNICAST_IF,
504 		 * IPV6_MULTICAST_IF, IPV6_ADDRFORM,
505 		 * IPV6_ADD_MEMBERSHIP, IPV6_DROP_MEMBERSHIP, IPV6_JOIN_ANYCAST,
506 		 * IPV6_LEAVE_ANYCAST, IPV6_MULTICAST_ALL, MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP,
507 		 * MCAST_JOIN_SOURCE_GROUP, MCAST_LEAVE_SOURCE_GROUP,
508 		 * MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE, MCAST_MSFILTER
509 		 * are not supported better not deal with mcast
510 		 */
511 		/* IPV6_ROUTER_ALERT, IPV6_ROUTER_ALERT_ISOLATE are not supported, since are evil */
512 
513 		/* IPV6_IPSEC_POLICY, IPV6_XFRM_POLICY are not supported */
514 		/* IPV6_ADDR_PREFERENCES is not supported, we must be careful with subflows */
515 		return false;
516 	}
517 	if (level == SOL_TCP) {
518 		switch (optname) {
519 		/* the following are no-op or should work just fine */
520 		case TCP_THIN_DUPACK:
521 		case TCP_DEFER_ACCEPT:
522 
523 		/* the following need some love */
524 		case TCP_MAXSEG:
525 		case TCP_NODELAY:
526 		case TCP_THIN_LINEAR_TIMEOUTS:
527 		case TCP_CONGESTION:
528 		case TCP_CORK:
529 		case TCP_KEEPIDLE:
530 		case TCP_KEEPINTVL:
531 		case TCP_KEEPCNT:
532 		case TCP_SYNCNT:
533 		case TCP_SAVE_SYN:
534 		case TCP_LINGER2:
535 		case TCP_WINDOW_CLAMP:
536 		case TCP_QUICKACK:
537 		case TCP_USER_TIMEOUT:
538 		case TCP_TIMESTAMP:
539 		case TCP_NOTSENT_LOWAT:
540 		case TCP_TX_DELAY:
541 			return true;
542 		}
543 
544 		/* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
545 
546 		/* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
547 		 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
548 		 */
549 		/* TCP_FASTOPEN_KEY, TCP_FASTOPEN TCP_FASTOPEN_CONNECT, TCP_FASTOPEN_NO_COOKIE,
550 		 * are not supported fastopen is currently unsupported
551 		 */
552 		/* TCP_INQ is currently unsupported, needs some recvmsg work */
553 	}
554 	return false;
555 }
556 
557 static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
558 					       unsigned int optlen)
559 {
560 	struct mptcp_subflow_context *subflow;
561 	struct sock *sk = (struct sock *)msk;
562 	char name[TCP_CA_NAME_MAX];
563 	bool cap_net_admin;
564 	int ret;
565 
566 	if (optlen < 1)
567 		return -EINVAL;
568 
569 	ret = strncpy_from_sockptr(name, optval,
570 				   min_t(long, TCP_CA_NAME_MAX - 1, optlen));
571 	if (ret < 0)
572 		return -EFAULT;
573 
574 	name[ret] = 0;
575 
576 	cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
577 
578 	ret = 0;
579 	lock_sock(sk);
580 	sockopt_seq_inc(msk);
581 	mptcp_for_each_subflow(msk, subflow) {
582 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
583 		int err;
584 
585 		lock_sock(ssk);
586 		err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
587 		if (err < 0 && ret == 0)
588 			ret = err;
589 		subflow->setsockopt_seq = msk->setsockopt_seq;
590 		release_sock(ssk);
591 	}
592 
593 	if (ret == 0)
594 		strcpy(msk->ca_name, name);
595 
596 	release_sock(sk);
597 	return ret;
598 }
599 
600 static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
601 				    sockptr_t optval, unsigned int optlen)
602 {
603 	switch (optname) {
604 	case TCP_ULP:
605 		return -EOPNOTSUPP;
606 	case TCP_CONGESTION:
607 		return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
608 	}
609 
610 	return -EOPNOTSUPP;
611 }
612 
613 int mptcp_setsockopt(struct sock *sk, int level, int optname,
614 		     sockptr_t optval, unsigned int optlen)
615 {
616 	struct mptcp_sock *msk = mptcp_sk(sk);
617 	struct sock *ssk;
618 
619 	pr_debug("msk=%p", msk);
620 
621 	if (level == SOL_SOCKET)
622 		return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
623 
624 	if (!mptcp_supported_sockopt(level, optname))
625 		return -ENOPROTOOPT;
626 
627 	/* @@ the meaning of setsockopt() when the socket is connected and
628 	 * there are multiple subflows is not yet defined. It is up to the
629 	 * MPTCP-level socket to configure the subflows until the subflow
630 	 * is in TCP fallback, when TCP socket options are passed through
631 	 * to the one remaining subflow.
632 	 */
633 	lock_sock(sk);
634 	ssk = __mptcp_tcp_fallback(msk);
635 	release_sock(sk);
636 	if (ssk)
637 		return tcp_setsockopt(ssk, level, optname, optval, optlen);
638 
639 	if (level == SOL_IPV6)
640 		return mptcp_setsockopt_v6(msk, optname, optval, optlen);
641 
642 	if (level == SOL_TCP)
643 		return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
644 
645 	return -EOPNOTSUPP;
646 }
647 
648 static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
649 					  char __user *optval, int __user *optlen)
650 {
651 	struct sock *sk = (struct sock *)msk;
652 	struct socket *ssock;
653 	int ret = -EINVAL;
654 	struct sock *ssk;
655 
656 	lock_sock(sk);
657 	ssk = msk->first;
658 	if (ssk) {
659 		ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
660 		goto out;
661 	}
662 
663 	ssock = __mptcp_nmpc_socket(msk);
664 	if (!ssock)
665 		goto out;
666 
667 	ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);
668 
669 out:
670 	release_sock(sk);
671 	return ret;
672 }
673 
674 void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
675 {
676 	struct sock *sk = &msk->sk.icsk_inet.sk;
677 	u32 flags = 0;
678 	bool slow;
679 	u8 val;
680 
681 	memset(info, 0, sizeof(*info));
682 
683 	slow = lock_sock_fast(sk);
684 
685 	info->mptcpi_subflows = READ_ONCE(msk->pm.subflows);
686 	info->mptcpi_add_addr_signal = READ_ONCE(msk->pm.add_addr_signaled);
687 	info->mptcpi_add_addr_accepted = READ_ONCE(msk->pm.add_addr_accepted);
688 	info->mptcpi_local_addr_used = READ_ONCE(msk->pm.local_addr_used);
689 	info->mptcpi_subflows_max = mptcp_pm_get_subflows_max(msk);
690 	val = mptcp_pm_get_add_addr_signal_max(msk);
691 	info->mptcpi_add_addr_signal_max = val;
692 	val = mptcp_pm_get_add_addr_accept_max(msk);
693 	info->mptcpi_add_addr_accepted_max = val;
694 	info->mptcpi_local_addr_max = mptcp_pm_get_local_addr_max(msk);
695 	if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags))
696 		flags |= MPTCP_INFO_FLAG_FALLBACK;
697 	if (READ_ONCE(msk->can_ack))
698 		flags |= MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED;
699 	info->mptcpi_flags = flags;
700 	info->mptcpi_token = READ_ONCE(msk->token);
701 	info->mptcpi_write_seq = READ_ONCE(msk->write_seq);
702 	info->mptcpi_snd_una = READ_ONCE(msk->snd_una);
703 	info->mptcpi_rcv_nxt = READ_ONCE(msk->ack_seq);
704 	info->mptcpi_csum_enabled = READ_ONCE(msk->csum_enabled);
705 
706 	unlock_sock_fast(sk, slow);
707 }
708 EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
709 
710 static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
711 {
712 	struct mptcp_info m_info;
713 	int len;
714 
715 	if (get_user(len, optlen))
716 		return -EFAULT;
717 
718 	len = min_t(unsigned int, len, sizeof(struct mptcp_info));
719 
720 	mptcp_diag_fill_info(msk, &m_info);
721 
722 	if (put_user(len, optlen))
723 		return -EFAULT;
724 
725 	if (copy_to_user(optval, &m_info, len))
726 		return -EFAULT;
727 
728 	return 0;
729 }
730 
731 static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
732 				  char __user *optval,
733 				  u32 copied,
734 				  int __user *optlen)
735 {
736 	u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
737 
738 	if (copied)
739 		copied += sfd->size_subflow_data;
740 	else
741 		copied = copylen;
742 
743 	if (put_user(copied, optlen))
744 		return -EFAULT;
745 
746 	if (copy_to_user(optval, sfd, copylen))
747 		return -EFAULT;
748 
749 	return 0;
750 }
751 
752 static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
753 				  char __user *optval, int __user *optlen)
754 {
755 	int len, copylen;
756 
757 	if (get_user(len, optlen))
758 		return -EFAULT;
759 
760 	/* if mptcp_subflow_data size is changed, need to adjust
761 	 * this function to deal with programs using old version.
762 	 */
763 	BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
764 
765 	if (len < MIN_INFO_OPTLEN_SIZE)
766 		return -EINVAL;
767 
768 	memset(sfd, 0, sizeof(*sfd));
769 
770 	copylen = min_t(unsigned int, len, sizeof(*sfd));
771 	if (copy_from_user(sfd, optval, copylen))
772 		return -EFAULT;
773 
774 	/* size_subflow_data is u32, but len is signed */
775 	if (sfd->size_subflow_data > INT_MAX ||
776 	    sfd->size_user > INT_MAX)
777 		return -EINVAL;
778 
779 	if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
780 	    sfd->size_subflow_data > len)
781 		return -EINVAL;
782 
783 	if (sfd->num_subflows || sfd->size_kernel)
784 		return -EINVAL;
785 
786 	return len - sfd->size_subflow_data;
787 }
788 
789 static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
790 				    int __user *optlen)
791 {
792 	struct mptcp_subflow_context *subflow;
793 	struct sock *sk = &msk->sk.icsk_inet.sk;
794 	unsigned int sfcount = 0, copied = 0;
795 	struct mptcp_subflow_data sfd;
796 	char __user *infoptr;
797 	int len;
798 
799 	len = mptcp_get_subflow_data(&sfd, optval, optlen);
800 	if (len < 0)
801 		return len;
802 
803 	sfd.size_kernel = sizeof(struct tcp_info);
804 	sfd.size_user = min_t(unsigned int, sfd.size_user,
805 			      sizeof(struct tcp_info));
806 
807 	infoptr = optval + sfd.size_subflow_data;
808 
809 	lock_sock(sk);
810 
811 	mptcp_for_each_subflow(msk, subflow) {
812 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
813 
814 		++sfcount;
815 
816 		if (len && len >= sfd.size_user) {
817 			struct tcp_info info;
818 
819 			tcp_get_info(ssk, &info);
820 
821 			if (copy_to_user(infoptr, &info, sfd.size_user)) {
822 				release_sock(sk);
823 				return -EFAULT;
824 			}
825 
826 			infoptr += sfd.size_user;
827 			copied += sfd.size_user;
828 			len -= sfd.size_user;
829 		}
830 	}
831 
832 	release_sock(sk);
833 
834 	sfd.num_subflows = sfcount;
835 
836 	if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
837 		return -EFAULT;
838 
839 	return 0;
840 }
841 
842 static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
843 {
844 	struct inet_sock *inet = inet_sk(sk);
845 
846 	memset(a, 0, sizeof(*a));
847 
848 	if (sk->sk_family == AF_INET) {
849 		a->sin_local.sin_family = AF_INET;
850 		a->sin_local.sin_port = inet->inet_sport;
851 		a->sin_local.sin_addr.s_addr = inet->inet_rcv_saddr;
852 
853 		if (!a->sin_local.sin_addr.s_addr)
854 			a->sin_local.sin_addr.s_addr = inet->inet_saddr;
855 
856 		a->sin_remote.sin_family = AF_INET;
857 		a->sin_remote.sin_port = inet->inet_dport;
858 		a->sin_remote.sin_addr.s_addr = inet->inet_daddr;
859 #if IS_ENABLED(CONFIG_IPV6)
860 	} else if (sk->sk_family == AF_INET6) {
861 		const struct ipv6_pinfo *np = inet6_sk(sk);
862 
863 		if (WARN_ON_ONCE(!np))
864 			return;
865 
866 		a->sin6_local.sin6_family = AF_INET6;
867 		a->sin6_local.sin6_port = inet->inet_sport;
868 
869 		if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
870 			a->sin6_local.sin6_addr = np->saddr;
871 		else
872 			a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
873 
874 		a->sin6_remote.sin6_family = AF_INET6;
875 		a->sin6_remote.sin6_port = inet->inet_dport;
876 		a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
877 #endif
878 	}
879 }
880 
881 static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
882 					  int __user *optlen)
883 {
884 	struct sock *sk = &msk->sk.icsk_inet.sk;
885 	struct mptcp_subflow_context *subflow;
886 	unsigned int sfcount = 0, copied = 0;
887 	struct mptcp_subflow_data sfd;
888 	char __user *addrptr;
889 	int len;
890 
891 	len = mptcp_get_subflow_data(&sfd, optval, optlen);
892 	if (len < 0)
893 		return len;
894 
895 	sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
896 	sfd.size_user = min_t(unsigned int, sfd.size_user,
897 			      sizeof(struct mptcp_subflow_addrs));
898 
899 	addrptr = optval + sfd.size_subflow_data;
900 
901 	lock_sock(sk);
902 
903 	mptcp_for_each_subflow(msk, subflow) {
904 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
905 
906 		++sfcount;
907 
908 		if (len && len >= sfd.size_user) {
909 			struct mptcp_subflow_addrs a;
910 
911 			mptcp_get_sub_addrs(ssk, &a);
912 
913 			if (copy_to_user(addrptr, &a, sfd.size_user)) {
914 				release_sock(sk);
915 				return -EFAULT;
916 			}
917 
918 			addrptr += sfd.size_user;
919 			copied += sfd.size_user;
920 			len -= sfd.size_user;
921 		}
922 	}
923 
924 	release_sock(sk);
925 
926 	sfd.num_subflows = sfcount;
927 
928 	if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
929 		return -EFAULT;
930 
931 	return 0;
932 }
933 
934 static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
935 				    char __user *optval, int __user *optlen)
936 {
937 	switch (optname) {
938 	case TCP_ULP:
939 	case TCP_CONGESTION:
940 	case TCP_INFO:
941 	case TCP_CC_INFO:
942 		return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
943 						      optval, optlen);
944 	}
945 	return -EOPNOTSUPP;
946 }
947 
948 static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
949 				      char __user *optval, int __user *optlen)
950 {
951 	switch (optname) {
952 	case MPTCP_INFO:
953 		return mptcp_getsockopt_info(msk, optval, optlen);
954 	case MPTCP_TCPINFO:
955 		return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
956 	case MPTCP_SUBFLOW_ADDRS:
957 		return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
958 	}
959 
960 	return -EOPNOTSUPP;
961 }
962 
963 int mptcp_getsockopt(struct sock *sk, int level, int optname,
964 		     char __user *optval, int __user *option)
965 {
966 	struct mptcp_sock *msk = mptcp_sk(sk);
967 	struct sock *ssk;
968 
969 	pr_debug("msk=%p", msk);
970 
971 	/* @@ the meaning of setsockopt() when the socket is connected and
972 	 * there are multiple subflows is not yet defined. It is up to the
973 	 * MPTCP-level socket to configure the subflows until the subflow
974 	 * is in TCP fallback, when socket options are passed through
975 	 * to the one remaining subflow.
976 	 */
977 	lock_sock(sk);
978 	ssk = __mptcp_tcp_fallback(msk);
979 	release_sock(sk);
980 	if (ssk)
981 		return tcp_getsockopt(ssk, level, optname, optval, option);
982 
983 	if (level == SOL_TCP)
984 		return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
985 	if (level == SOL_MPTCP)
986 		return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
987 	return -EOPNOTSUPP;
988 }
989 
990 static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
991 {
992 	static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
993 	struct sock *sk = (struct sock *)msk;
994 
995 	if (ssk->sk_prot->keepalive) {
996 		if (sock_flag(sk, SOCK_KEEPOPEN))
997 			ssk->sk_prot->keepalive(ssk, 1);
998 		else
999 			ssk->sk_prot->keepalive(ssk, 0);
1000 	}
1001 
1002 	ssk->sk_priority = sk->sk_priority;
1003 	ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1004 	ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
1005 
1006 	if (sk->sk_userlocks & tx_rx_locks) {
1007 		ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1008 		if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
1009 			WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1010 		if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1011 			WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1012 	}
1013 
1014 	if (sock_flag(sk, SOCK_LINGER)) {
1015 		ssk->sk_lingertime = sk->sk_lingertime;
1016 		sock_set_flag(ssk, SOCK_LINGER);
1017 	} else {
1018 		sock_reset_flag(ssk, SOCK_LINGER);
1019 	}
1020 
1021 	if (sk->sk_mark != ssk->sk_mark) {
1022 		ssk->sk_mark = sk->sk_mark;
1023 		sk_dst_reset(ssk);
1024 	}
1025 
1026 	sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1027 
1028 	if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
1029 		tcp_set_congestion_control(ssk, msk->ca_name, false, true);
1030 }
1031 
1032 static void __mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1033 {
1034 	bool slow = lock_sock_fast(ssk);
1035 
1036 	sync_socket_options(msk, ssk);
1037 
1038 	unlock_sock_fast(ssk, slow);
1039 }
1040 
1041 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1042 {
1043 	struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1044 
1045 	msk_owned_by_me(msk);
1046 
1047 	if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1048 		__mptcp_sockopt_sync(msk, ssk);
1049 
1050 		subflow->setsockopt_seq = msk->setsockopt_seq;
1051 	}
1052 }
1053 
1054 void mptcp_sockopt_sync_all(struct mptcp_sock *msk)
1055 {
1056 	struct mptcp_subflow_context *subflow;
1057 	struct sock *sk = (struct sock *)msk;
1058 	u32 seq;
1059 
1060 	seq = sockopt_seq_reset(sk);
1061 
1062 	mptcp_for_each_subflow(msk, subflow) {
1063 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1064 		u32 sseq = READ_ONCE(subflow->setsockopt_seq);
1065 
1066 		if (sseq != msk->setsockopt_seq) {
1067 			__mptcp_sockopt_sync(msk, ssk);
1068 			WRITE_ONCE(subflow->setsockopt_seq, seq);
1069 		} else if (sseq != seq) {
1070 			WRITE_ONCE(subflow->setsockopt_seq, seq);
1071 		}
1072 
1073 		cond_resched();
1074 	}
1075 
1076 	msk->setsockopt_seq = seq;
1077 }
1078