projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vrf: use dev_replace_track() for better tracking
[linux-2.6-block.git] / net / mptcp / sockopt.c
CommitLineData
0abdde82
PA		 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
06f15cee
FW		 17#define MIN_INFO_OPTLEN_SIZE 16
18
0abdde82
PA		 19static 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
df00b087
FW		 29static 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
1b3e7ede
FW		 50static 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
57static 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
69static 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) {
a03c99b2
FW		 82 case SO_DEBUG:
83 sock_valbool_flag(ssk, SOCK_DBG, !!val);
84 break;
1b3e7ede
FW		 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;
5d0a6bc8
FW		 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;
36704413
FW		 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;
6f0d7198
FW		 109 case SO_INCOMING_CPU:
110 WRITE_ONCE(ssk->sk_incoming_cpu, val);
111 break;
1b3e7ede
FW		 112 }
113
114 subflow->setsockopt_seq = msk->setsockopt_seq;
115 unlock_sock_fast(ssk, slow);
116 }
117
118 release_sock(sk);
119}
120
121static 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
6f0d7198
FW		 136static 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
9061f24b
FW		 145static 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
6c9a0a0f 162 sock_set_timestamp(sk, optname, !!val);
9061f24b
FW		 163 unlock_sock_fast(ssk, slow);
164 }
165
166 release_sock(sk);
167 return 0;
168}
169
1b3e7ede 170static int mptcp_setsockopt_sol_socket_int(struct mptcp_sock *msk, int optname,
6c9a0a0f
YL		 171 sockptr_t optval,
172 unsigned int optlen)
1b3e7ede
FW		 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;
a03c99b2 184 case SO_DEBUG:
36704413 185 case SO_MARK:
1b3e7ede 186 case SO_PRIORITY:
5d0a6bc8
FW		 187 case SO_SNDBUF:
188 case SO_SNDBUFFORCE:
189 case SO_RCVBUF:
190 case SO_RCVBUFFORCE:
1b3e7ede 191 return mptcp_sol_socket_intval(msk, optname, val);
6f0d7198
FW		 192 case SO_INCOMING_CPU:
193 mptcp_so_incoming_cpu(msk, val);
194 return 0;
9061f24b
FW		 195 case SO_TIMESTAMP_OLD:
196 case SO_TIMESTAMP_NEW:
197 case SO_TIMESTAMPNS_OLD:
198 case SO_TIMESTAMPNS_NEW:
9061f24b 199 return mptcp_setsockopt_sol_socket_tstamp(msk, optname, val);
1b3e7ede
FW		 200 }
201
202 return -ENOPROTOOPT;
203}
204
6c9a0a0f
YL		 205static 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;
d463126e
YL		 212 struct so_timestamping timestamping;
213 int ret;
6c9a0a0f 214
d463126e
YL		 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 }
6c9a0a0f
YL		 227
228 ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
d463126e
YL		 229 KERNEL_SOCKPTR(&timestamping),
230 sizeof(timestamping));
6c9a0a0f
YL		 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
d463126e 240 sock_set_timestamping(sk, optname, timestamping);
6c9a0a0f
YL		 241 unlock_sock_fast(ssk, slow);
242 }
243
244 release_sock(sk);
245
246 return 0;
247}
248
268b1238
FW		 249static 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
0abdde82
PA		 290static 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:
5d0a6bc8
FW		 300 case SO_BINDTODEVICE:
301 case SO_BINDTOIFINDEX:
0abdde82
PA		 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;
5d0a6bc8
FW		 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;
0abdde82
PA		 319 }
320 release_sock(sk);
321 return ret;
1b3e7ede
FW		 322 case SO_KEEPALIVE:
323 case SO_PRIORITY:
5d0a6bc8
FW		 324 case SO_SNDBUF:
325 case SO_SNDBUFFORCE:
326 case SO_RCVBUF:
327 case SO_RCVBUFFORCE:
36704413 328 case SO_MARK:
6f0d7198 329 case SO_INCOMING_CPU:
a03c99b2 330 case SO_DEBUG:
9061f24b
FW		 331 case SO_TIMESTAMP_OLD:
332 case SO_TIMESTAMP_NEW:
333 case SO_TIMESTAMPNS_OLD:
334 case SO_TIMESTAMPNS_NEW:
6c9a0a0f
YL		 335 return mptcp_setsockopt_sol_socket_int(msk, optname, optval,
336 optlen);
9061f24b
FW		 337 case SO_TIMESTAMPING_OLD:
338 case SO_TIMESTAMPING_NEW:
6c9a0a0f
YL		 339 return mptcp_setsockopt_sol_socket_timestamping(msk, optname,
340 optval, optlen);
268b1238
FW		 341 case SO_LINGER:
342 return mptcp_setsockopt_sol_socket_linger(msk, optval, optlen);
7a009a70
FW		 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);
a03c99b2
FW		 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;
0abdde82
PA		 362 }
363
7a009a70
FW		 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;
0abdde82
PA		 382}
383
384static 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:
c9406a23
FW		 393 case IPV6_TRANSPARENT:
394 case IPV6_FREEBIND:
0abdde82
PA		 395 lock_sock(sk);
396 ssock = __mptcp_nmpc_socket(msk);
397 if (!ssock) {
398 release_sock(sk);
399 return -EINVAL;
400 }
401
402 ret = tcp_setsockopt(ssock->sk, SOL_IPV6, optname, optval, optlen);
c9406a23
FW		 403 if (ret != 0) {
404 release_sock(sk);
405 return ret;
406 }
407
408 sockopt_seq_inc(msk);
409
410 switch (optname) {
411 case IPV6_V6ONLY:
0abdde82 412 sk->sk_ipv6only = ssock->sk->sk_ipv6only;
c9406a23
FW		 413 break;
414 case IPV6_TRANSPARENT:
415 inet_sk(sk)->transparent = inet_sk(ssock->sk)->transparent;
416 break;
417 case IPV6_FREEBIND:
418 inet_sk(sk)->freebind = inet_sk(ssock->sk)->freebind;
419 break;
420 }
0abdde82
PA		 421
422 release_sock(sk);
423 break;
424 }
425
426 return ret;
427}
428
d9e4c129
PA		 429static bool mptcp_supported_sockopt(int level, int optname)
430{
d9e4c129
PA		 431 if (level == SOL_IP) {
432 switch (optname) {
433 /* should work fine */
434 case IP_FREEBIND:
435 case IP_TRANSPARENT:
436
437 /* the following are control cmsg related */
438 case IP_PKTINFO:
439 case IP_RECVTTL:
440 case IP_RECVTOS:
441 case IP_RECVOPTS:
442 case IP_RETOPTS:
443 case IP_PASSSEC:
444 case IP_RECVORIGDSTADDR:
445 case IP_CHECKSUM:
446 case IP_RECVFRAGSIZE:
447
448 /* common stuff that need some love */
449 case IP_TOS:
450 case IP_TTL:
451 case IP_BIND_ADDRESS_NO_PORT:
452 case IP_MTU_DISCOVER:
453 case IP_RECVERR:
454
455 /* possibly less common may deserve some love */
456 case IP_MINTTL:
457
458 /* the following is apparently a no-op for plain TCP */
459 case IP_RECVERR_RFC4884:
460 return true;
461 }
462
463 /* IP_OPTIONS is not supported, needs subflow care */
464 /* IP_HDRINCL, IP_NODEFRAG are not supported, RAW specific */
465 /* IP_MULTICAST_TTL, IP_MULTICAST_LOOP, IP_UNICAST_IF,
466 * IP_ADD_MEMBERSHIP, IP_ADD_SOURCE_MEMBERSHIP, IP_DROP_MEMBERSHIP,
467 * IP_DROP_SOURCE_MEMBERSHIP, IP_BLOCK_SOURCE, IP_UNBLOCK_SOURCE,
468 * MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP MCAST_JOIN_SOURCE_GROUP,
469 * MCAST_LEAVE_SOURCE_GROUP, MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE,
470 * MCAST_MSFILTER, IP_MULTICAST_ALL are not supported, better not deal
471 * with mcast stuff
472 */
473 /* IP_IPSEC_POLICY, IP_XFRM_POLICY are nut supported, unrelated here */
474 return false;
475 }
476 if (level == SOL_IPV6) {
477 switch (optname) {
478 case IPV6_V6ONLY:
479
480 /* the following are control cmsg related */
481 case IPV6_RECVPKTINFO:
482 case IPV6_2292PKTINFO:
483 case IPV6_RECVHOPLIMIT:
484 case IPV6_2292HOPLIMIT:
485 case IPV6_RECVRTHDR:
486 case IPV6_2292RTHDR:
487 case IPV6_RECVHOPOPTS:
488 case IPV6_2292HOPOPTS:
489 case IPV6_RECVDSTOPTS:
490 case IPV6_2292DSTOPTS:
491 case IPV6_RECVTCLASS:
492 case IPV6_FLOWINFO:
493 case IPV6_RECVPATHMTU:
494 case IPV6_RECVORIGDSTADDR:
495 case IPV6_RECVFRAGSIZE:
496
497 /* the following ones need some love but are quite common */
498 case IPV6_TCLASS:
499 case IPV6_TRANSPARENT:
500 case IPV6_FREEBIND:
501 case IPV6_PKTINFO:
502 case IPV6_2292PKTOPTIONS:
503 case IPV6_UNICAST_HOPS:
504 case IPV6_MTU_DISCOVER:
505 case IPV6_MTU:
506 case IPV6_RECVERR:
507 case IPV6_FLOWINFO_SEND:
508 case IPV6_FLOWLABEL_MGR:
509 case IPV6_MINHOPCOUNT:
510 case IPV6_DONTFRAG:
511 case IPV6_AUTOFLOWLABEL:
512
513 /* the following one is a no-op for plain TCP */
514 case IPV6_RECVERR_RFC4884:
515 return true;
516 }
517
518 /* IPV6_HOPOPTS, IPV6_RTHDRDSTOPTS, IPV6_RTHDR, IPV6_DSTOPTS are
519 * not supported
520 */
521 /* IPV6_MULTICAST_HOPS, IPV6_MULTICAST_LOOP, IPV6_UNICAST_IF,
522 * IPV6_MULTICAST_IF, IPV6_ADDRFORM,
523 * IPV6_ADD_MEMBERSHIP, IPV6_DROP_MEMBERSHIP, IPV6_JOIN_ANYCAST,
524 * IPV6_LEAVE_ANYCAST, IPV6_MULTICAST_ALL, MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP,
525 * MCAST_JOIN_SOURCE_GROUP, MCAST_LEAVE_SOURCE_GROUP,
526 * MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE, MCAST_MSFILTER
527 * are not supported better not deal with mcast
528 */
529 /* IPV6_ROUTER_ALERT, IPV6_ROUTER_ALERT_ISOLATE are not supported, since are evil */
530
531 /* IPV6_IPSEC_POLICY, IPV6_XFRM_POLICY are not supported */
532 /* IPV6_ADDR_PREFERENCES is not supported, we must be careful with subflows */
533 return false;
534 }
535 if (level == SOL_TCP) {
536 switch (optname) {
537 /* the following are no-op or should work just fine */
538 case TCP_THIN_DUPACK:
539 case TCP_DEFER_ACCEPT:
540
541 /* the following need some love */
542 case TCP_MAXSEG:
543 case TCP_NODELAY:
544 case TCP_THIN_LINEAR_TIMEOUTS:
545 case TCP_CONGESTION:
546 case TCP_ULP:
547 case TCP_CORK:
548 case TCP_KEEPIDLE:
549 case TCP_KEEPINTVL:
550 case TCP_KEEPCNT:
551 case TCP_SYNCNT:
552 case TCP_SAVE_SYN:
553 case TCP_LINGER2:
554 case TCP_WINDOW_CLAMP:
555 case TCP_QUICKACK:
556 case TCP_USER_TIMEOUT:
557 case TCP_TIMESTAMP:
558 case TCP_NOTSENT_LOWAT:
559 case TCP_TX_DELAY:
560 return true;
561 }
562
563 /* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
564
565 /* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
566 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
567 */
568 /* TCP_FASTOPEN_KEY, TCP_FASTOPEN TCP_FASTOPEN_CONNECT, TCP_FASTOPEN_NO_COOKIE,
569 * are not supported fastopen is currently unsupported
570 */
571 /* TCP_INQ is currently unsupported, needs some recvmsg work */
572 }
573 return false;
574}
575
aa1fbd94
FW		 576static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
577 unsigned int optlen)
578{
579 struct mptcp_subflow_context *subflow;
580 struct sock *sk = (struct sock *)msk;
581 char name[TCP_CA_NAME_MAX];
582 bool cap_net_admin;
583 int ret;
584
585 if (optlen < 1)
586 return -EINVAL;
587
588 ret = strncpy_from_sockptr(name, optval,
589 min_t(long, TCP_CA_NAME_MAX - 1, optlen));
590 if (ret < 0)
591 return -EFAULT;
592
593 name[ret] = 0;
594
595 cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
596
597 ret = 0;
598 lock_sock(sk);
599 sockopt_seq_inc(msk);
600 mptcp_for_each_subflow(msk, subflow) {
601 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
602 int err;
603
604 lock_sock(ssk);
605 err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
606 if (err < 0 && ret == 0)
607 ret = err;
608 subflow->setsockopt_seq = msk->setsockopt_seq;
609 release_sock(ssk);
610 }
611
612 if (ret == 0)
20b5759f 613 strcpy(msk->ca_name, name);
aa1fbd94
FW		 614
615 release_sock(sk);
616 return ret;
617}
618
c9406a23
FW		 619static int mptcp_setsockopt_sol_ip_set_transparent(struct mptcp_sock *msk, int optname,
620 sockptr_t optval, unsigned int optlen)
621{
622 struct sock *sk = (struct sock *)msk;
623 struct inet_sock *issk;
624 struct socket *ssock;
625 int err;
626
627 err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
628 if (err != 0)
629 return err;
630
631 lock_sock(sk);
632
633 ssock = __mptcp_nmpc_socket(msk);
634 if (!ssock) {
635 release_sock(sk);
636 return -EINVAL;
637 }
638
639 issk = inet_sk(ssock->sk);
640
641 switch (optname) {
642 case IP_FREEBIND:
643 issk->freebind = inet_sk(sk)->freebind;
644 break;
645 case IP_TRANSPARENT:
646 issk->transparent = inet_sk(sk)->transparent;
647 break;
648 default:
649 release_sock(sk);
650 WARN_ON_ONCE(1);
651 return -EOPNOTSUPP;
652 }
653
654 sockopt_seq_inc(msk);
655 release_sock(sk);
656 return 0;
657}
658
ffcacff8
PS		 659static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,
660 sockptr_t optval, unsigned int optlen)
661{
662 struct mptcp_subflow_context *subflow;
663 struct sock *sk = (struct sock *)msk;
664 int err, val;
665
666 err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
667
668 if (err != 0)
669 return err;
670
671 lock_sock(sk);
672 sockopt_seq_inc(msk);
673 val = inet_sk(sk)->tos;
674 mptcp_for_each_subflow(msk, subflow) {
675 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
676
677 __ip_sock_set_tos(ssk, val);
678 }
679 release_sock(sk);
680
681 return err;
682}
683
684static int mptcp_setsockopt_v4(struct mptcp_sock *msk, int optname,
685 sockptr_t optval, unsigned int optlen)
686{
687 switch (optname) {
c9406a23
FW		 688 case IP_FREEBIND:
689 case IP_TRANSPARENT:
690 return mptcp_setsockopt_sol_ip_set_transparent(msk, optname, optval, optlen);
ffcacff8
PS		 691 case IP_TOS:
692 return mptcp_setsockopt_v4_set_tos(msk, optname, optval, optlen);
693 }
694
695 return -EOPNOTSUPP;
696}
697
aa1fbd94
FW		 698static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
699 sockptr_t optval, unsigned int optlen)
700{
701 switch (optname) {
702 case TCP_ULP:
703 return -EOPNOTSUPP;
704 case TCP_CONGESTION:
705 return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
706 }
707
708 return -EOPNOTSUPP;
709}
710
0abdde82
PA		 711int mptcp_setsockopt(struct sock *sk, int level, int optname,
712 sockptr_t optval, unsigned int optlen)
713{
714 struct mptcp_sock *msk = mptcp_sk(sk);
715 struct sock *ssk;
716
717 pr_debug("msk=%p", msk);
718
719 if (level == SOL_SOCKET)
720 return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
721
7a009a70
FW		 722 if (!mptcp_supported_sockopt(level, optname))
723 return -ENOPROTOOPT;
724
0abdde82
PA		 725 /* @@ the meaning of setsockopt() when the socket is connected and
726 * there are multiple subflows is not yet defined. It is up to the
727 * MPTCP-level socket to configure the subflows until the subflow
728 * is in TCP fallback, when TCP socket options are passed through
729 * to the one remaining subflow.
730 */
731 lock_sock(sk);
732 ssk = __mptcp_tcp_fallback(msk);
733 release_sock(sk);
734 if (ssk)
735 return tcp_setsockopt(ssk, level, optname, optval, optlen);
736
ffcacff8
PS		 737 if (level == SOL_IP)
738 return mptcp_setsockopt_v4(msk, optname, optval, optlen);
739
0abdde82
PA		 740 if (level == SOL_IPV6)
741 return mptcp_setsockopt_v6(msk, optname, optval, optlen);
742
aa1fbd94
FW		 743 if (level == SOL_TCP)
744 return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
745
746 return -EOPNOTSUPP;
747}
748
749static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
750 char __user *optval, int __user *optlen)
751{
752 struct sock *sk = (struct sock *)msk;
753 struct socket *ssock;
754 int ret = -EINVAL;
755 struct sock *ssk;
756
757 lock_sock(sk);
758 ssk = msk->first;
759 if (ssk) {
760 ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
761 goto out;
762 }
763
764 ssock = __mptcp_nmpc_socket(msk);
765 if (!ssock)
766 goto out;
767
768 ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);
769
770out:
771 release_sock(sk);
772 return ret;
773}
774
61bc6e82
FW		 775void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
776{
777 struct sock *sk = &msk->sk.icsk_inet.sk;
61bc6e82 778 u32 flags = 0;
55c42fa7 779 bool slow;
61bc6e82
FW		 780 u8 val;
781
55c42fa7
FW		 782 memset(info, 0, sizeof(*info));
783
784 slow = lock_sock_fast(sk);
785
61bc6e82
FW		 786 info->mptcpi_subflows = READ_ONCE(msk->pm.subflows);
787 info->mptcpi_add_addr_signal = READ_ONCE(msk->pm.add_addr_signaled);
788 info->mptcpi_add_addr_accepted = READ_ONCE(msk->pm.add_addr_accepted);
789 info->mptcpi_local_addr_used = READ_ONCE(msk->pm.local_addr_used);
790 info->mptcpi_subflows_max = mptcp_pm_get_subflows_max(msk);
791 val = mptcp_pm_get_add_addr_signal_max(msk);
792 info->mptcpi_add_addr_signal_max = val;
793 val = mptcp_pm_get_add_addr_accept_max(msk);
794 info->mptcpi_add_addr_accepted_max = val;
795 info->mptcpi_local_addr_max = mptcp_pm_get_local_addr_max(msk);
796 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags))
797 flags |= MPTCP_INFO_FLAG_FALLBACK;
798 if (READ_ONCE(msk->can_ack))
799 flags |= MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED;
800 info->mptcpi_flags = flags;
801 info->mptcpi_token = READ_ONCE(msk->token);
802 info->mptcpi_write_seq = READ_ONCE(msk->write_seq);
803 info->mptcpi_snd_una = READ_ONCE(msk->snd_una);
804 info->mptcpi_rcv_nxt = READ_ONCE(msk->ack_seq);
805 info->mptcpi_csum_enabled = READ_ONCE(msk->csum_enabled);
806
807 unlock_sock_fast(sk, slow);
808}
809EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
810
55c42fa7
FW		 811static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
812{
813 struct mptcp_info m_info;
814 int len;
815
816 if (get_user(len, optlen))
817 return -EFAULT;
818
819 len = min_t(unsigned int, len, sizeof(struct mptcp_info));
820
821 mptcp_diag_fill_info(msk, &m_info);
822
823 if (put_user(len, optlen))
824 return -EFAULT;
825
826 if (copy_to_user(optval, &m_info, len))
827 return -EFAULT;
828
829 return 0;
830}
831
06f15cee
FW		 832static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
833 char __user *optval,
834 u32 copied,
835 int __user *optlen)
836{
837 u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
838
839 if (copied)
840 copied += sfd->size_subflow_data;
841 else
842 copied = copylen;
843
844 if (put_user(copied, optlen))
845 return -EFAULT;
846
847 if (copy_to_user(optval, sfd, copylen))
848 return -EFAULT;
849
850 return 0;
851}
852
853static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
854 char __user *optval, int __user *optlen)
855{
856 int len, copylen;
857
858 if (get_user(len, optlen))
859 return -EFAULT;
860
861 /* if mptcp_subflow_data size is changed, need to adjust
862 * this function to deal with programs using old version.
863 */
864 BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
865
866 if (len < MIN_INFO_OPTLEN_SIZE)
867 return -EINVAL;
868
869 memset(sfd, 0, sizeof(*sfd));
870
871 copylen = min_t(unsigned int, len, sizeof(*sfd));
872 if (copy_from_user(sfd, optval, copylen))
873 return -EFAULT;
874
875 /* size_subflow_data is u32, but len is signed */
876 if (sfd->size_subflow_data > INT_MAX ||
877 sfd->size_user > INT_MAX)
878 return -EINVAL;
879
880 if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
881 sfd->size_subflow_data > len)
882 return -EINVAL;
883
884 if (sfd->num_subflows || sfd->size_kernel)
885 return -EINVAL;
886
887 return len - sfd->size_subflow_data;
888}
889
890static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
891 int __user *optlen)
892{
893 struct mptcp_subflow_context *subflow;
894 struct sock *sk = &msk->sk.icsk_inet.sk;
895 unsigned int sfcount = 0, copied = 0;
896 struct mptcp_subflow_data sfd;
897 char __user *infoptr;
898 int len;
899
900 len = mptcp_get_subflow_data(&sfd, optval, optlen);
901 if (len < 0)
902 return len;
903
904 sfd.size_kernel = sizeof(struct tcp_info);
905 sfd.size_user = min_t(unsigned int, sfd.size_user,
906 sizeof(struct tcp_info));
907
908 infoptr = optval + sfd.size_subflow_data;
909
910 lock_sock(sk);
911
912 mptcp_for_each_subflow(msk, subflow) {
913 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
914
915 ++sfcount;
916
917 if (len && len >= sfd.size_user) {
918 struct tcp_info info;
919
920 tcp_get_info(ssk, &info);
921
922 if (copy_to_user(infoptr, &info, sfd.size_user)) {
923 release_sock(sk);
924 return -EFAULT;
925 }
926
927 infoptr += sfd.size_user;
928 copied += sfd.size_user;
929 len -= sfd.size_user;
930 }
931 }
932
933 release_sock(sk);
934
935 sfd.num_subflows = sfcount;
936
937 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
938 return -EFAULT;
939
940 return 0;
941}
942
c11c5906
FW		 943static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
944{
945 struct inet_sock *inet = inet_sk(sk);
946
947 memset(a, 0, sizeof(*a));
948
949 if (sk->sk_family == AF_INET) {
950 a->sin_local.sin_family = AF_INET;
951 a->sin_local.sin_port = inet->inet_sport;
952 a->sin_local.sin_addr.s_addr = inet->inet_rcv_saddr;
953
954 if (!a->sin_local.sin_addr.s_addr)
955 a->sin_local.sin_addr.s_addr = inet->inet_saddr;
956
957 a->sin_remote.sin_family = AF_INET;
958 a->sin_remote.sin_port = inet->inet_dport;
959 a->sin_remote.sin_addr.s_addr = inet->inet_daddr;
960#if IS_ENABLED(CONFIG_IPV6)
961 } else if (sk->sk_family == AF_INET6) {
962 const struct ipv6_pinfo *np = inet6_sk(sk);
963
29211e7d
TG		 964 if (WARN_ON_ONCE(!np))
965 return;
966
c11c5906
FW		 967 a->sin6_local.sin6_family = AF_INET6;
968 a->sin6_local.sin6_port = inet->inet_sport;
969
970 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
971 a->sin6_local.sin6_addr = np->saddr;
972 else
973 a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
974
975 a->sin6_remote.sin6_family = AF_INET6;
976 a->sin6_remote.sin6_port = inet->inet_dport;
977 a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
978#endif
979 }
980}
981
982static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
983 int __user *optlen)
984{
985 struct sock *sk = &msk->sk.icsk_inet.sk;
986 struct mptcp_subflow_context *subflow;
987 unsigned int sfcount = 0, copied = 0;
988 struct mptcp_subflow_data sfd;
989 char __user *addrptr;
990 int len;
991
992 len = mptcp_get_subflow_data(&sfd, optval, optlen);
993 if (len < 0)
994 return len;
995
996 sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
997 sfd.size_user = min_t(unsigned int, sfd.size_user,
998 sizeof(struct mptcp_subflow_addrs));
999
1000 addrptr = optval + sfd.size_subflow_data;
1001
1002 lock_sock(sk);
1003
1004 mptcp_for_each_subflow(msk, subflow) {
1005 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1006
1007 ++sfcount;
1008
1009 if (len && len >= sfd.size_user) {
1010 struct mptcp_subflow_addrs a;
1011
1012 mptcp_get_sub_addrs(ssk, &a);
1013
1014 if (copy_to_user(addrptr, &a, sfd.size_user)) {
1015 release_sock(sk);
1016 return -EFAULT;
1017 }
1018
1019 addrptr += sfd.size_user;
1020 copied += sfd.size_user;
1021 len -= sfd.size_user;
1022 }
1023 }
1024
1025 release_sock(sk);
1026
1027 sfd.num_subflows = sfcount;
1028
1029 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1030 return -EFAULT;
1031
1032 return 0;
1033}
1034
aa1fbd94
FW		 1035static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
1036 char __user *optval, int __user *optlen)
1037{
1038 switch (optname) {
1039 case TCP_ULP:
1040 case TCP_CONGESTION:
1041 case TCP_INFO:
1042 case TCP_CC_INFO:
1043 return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
1044 optval, optlen);
1045 }
0abdde82
PA		 1046 return -EOPNOTSUPP;
1047}
1048
55c42fa7
FW		 1049static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
1050 char __user *optval, int __user *optlen)
1051{
1052 switch (optname) {
1053 case MPTCP_INFO:
1054 return mptcp_getsockopt_info(msk, optval, optlen);
06f15cee
FW		 1055 case MPTCP_TCPINFO:
1056 return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
c11c5906
FW		 1057 case MPTCP_SUBFLOW_ADDRS:
1058 return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
55c42fa7
FW		 1059 }
1060
1061 return -EOPNOTSUPP;
1062}
1063
0abdde82
PA		 1064int mptcp_getsockopt(struct sock *sk, int level, int optname,
1065 char __user *optval, int __user *option)
1066{
1067 struct mptcp_sock *msk = mptcp_sk(sk);
1068 struct sock *ssk;
1069
1070 pr_debug("msk=%p", msk);
1071
1072 /* @@ the meaning of setsockopt() when the socket is connected and
1073 * there are multiple subflows is not yet defined. It is up to the
1074 * MPTCP-level socket to configure the subflows until the subflow
1075 * is in TCP fallback, when socket options are passed through
1076 * to the one remaining subflow.
1077 */
1078 lock_sock(sk);
1079 ssk = __mptcp_tcp_fallback(msk);
1080 release_sock(sk);
1081 if (ssk)
1082 return tcp_getsockopt(ssk, level, optname, optval, option);
1083
aa1fbd94
FW		 1084 if (level == SOL_TCP)
1085 return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
55c42fa7
FW		 1086 if (level == SOL_MPTCP)
1087 return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
0abdde82
PA		 1088 return -EOPNOTSUPP;
1089}
1090
1b3e7ede
FW		 1091static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
1092{
5d0a6bc8 1093 static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
1b3e7ede
FW		 1094 struct sock *sk = (struct sock *)msk;
1095
1096 if (ssk->sk_prot->keepalive) {
1097 if (sock_flag(sk, SOCK_KEEPOPEN))
1098 ssk->sk_prot->keepalive(ssk, 1);
1099 else
1100 ssk->sk_prot->keepalive(ssk, 0);
1101 }
1102
1103 ssk->sk_priority = sk->sk_priority;
5d0a6bc8
FW		 1104 ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1105 ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
ffcacff8 1106 __ip_sock_set_tos(ssk, inet_sk(sk)->tos);
5d0a6bc8
FW		 1107
1108 if (sk->sk_userlocks & tx_rx_locks) {
1109 ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1110 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
1111 WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1112 if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1113 WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1114 }
1115
1116 if (sock_flag(sk, SOCK_LINGER)) {
1117 ssk->sk_lingertime = sk->sk_lingertime;
1118 sock_set_flag(ssk, SOCK_LINGER);
1119 } else {
1120 sock_reset_flag(ssk, SOCK_LINGER);
1121 }
1122
1123 if (sk->sk_mark != ssk->sk_mark) {
1124 ssk->sk_mark = sk->sk_mark;
1125 sk_dst_reset(ssk);
1126 }
1127
1128 sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1129
1130 if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
20b5759f 1131 tcp_set_congestion_control(ssk, msk->ca_name, false, true);
c9406a23
FW		 1132
1133 inet_sk(ssk)->transparent = inet_sk(sk)->transparent;
1134 inet_sk(ssk)->freebind = inet_sk(sk)->freebind;
1b3e7ede
FW		 1135}
1136
df00b087
FW		 1137static void __mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1138{
1b3e7ede
FW		 1139 bool slow = lock_sock_fast(ssk);
1140
1141 sync_socket_options(msk, ssk);
1142
1143 unlock_sock_fast(ssk, slow);
df00b087
FW		 1144}
1145
78962489
FW		 1146void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1147{
df00b087
FW		 1148 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1149
78962489 1150 msk_owned_by_me(msk);
df00b087
FW		 1151
1152 if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1153 __mptcp_sockopt_sync(msk, ssk);
1154
1155 subflow->setsockopt_seq = msk->setsockopt_seq;
1156 }
78962489
FW		 1157}
1158
1159void mptcp_sockopt_sync_all(struct mptcp_sock *msk)
1160{
1161 struct mptcp_subflow_context *subflow;
df00b087
FW		 1162 struct sock *sk = (struct sock *)msk;
1163 u32 seq;
78962489 1164
df00b087 1165 seq = sockopt_seq_reset(sk);
78962489
FW		 1166
1167 mptcp_for_each_subflow(msk, subflow) {
1168 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
df00b087 1169 u32 sseq = READ_ONCE(subflow->setsockopt_seq);
78962489 1170
df00b087
FW		 1171 if (sseq != msk->setsockopt_seq) {
1172 __mptcp_sockopt_sync(msk, ssk);
1173 WRITE_ONCE(subflow->setsockopt_seq, seq);
1174 } else if (sseq != seq) {
1175 WRITE_ONCE(subflow->setsockopt_seq, seq);
1176 }
78962489
FW		 1177
1178 cond_resched();
1179 }
df00b087
FW		 1180
1181 msk->setsockopt_seq = seq;
78962489 1182}
Linux 6.x block layer and io_uring tree(s)