projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'dpaa2-irq-coalescing'
[linux-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:
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
d9e4c129
PA		 411static bool mptcp_supported_sockopt(int level, int optname)
412{
d9e4c129
PA		 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_ULP:
529 case TCP_CORK:
530 case TCP_KEEPIDLE:
531 case TCP_KEEPINTVL:
532 case TCP_KEEPCNT:
533 case TCP_SYNCNT:
534 case TCP_SAVE_SYN:
535 case TCP_LINGER2:
536 case TCP_WINDOW_CLAMP:
537 case TCP_QUICKACK:
538 case TCP_USER_TIMEOUT:
539 case TCP_TIMESTAMP:
540 case TCP_NOTSENT_LOWAT:
541 case TCP_TX_DELAY:
542 return true;
543 }
544
545 /* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
546
547 /* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
548 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
549 */
550 /* TCP_FASTOPEN_KEY, TCP_FASTOPEN TCP_FASTOPEN_CONNECT, TCP_FASTOPEN_NO_COOKIE,
551 * are not supported fastopen is currently unsupported
552 */
553 /* TCP_INQ is currently unsupported, needs some recvmsg work */
554 }
555 return false;
556}
557
aa1fbd94
FW		 558static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
559 unsigned int optlen)
560{
561 struct mptcp_subflow_context *subflow;
562 struct sock *sk = (struct sock *)msk;
563 char name[TCP_CA_NAME_MAX];
564 bool cap_net_admin;
565 int ret;
566
567 if (optlen < 1)
568 return -EINVAL;
569
570 ret = strncpy_from_sockptr(name, optval,
571 min_t(long, TCP_CA_NAME_MAX - 1, optlen));
572 if (ret < 0)
573 return -EFAULT;
574
575 name[ret] = 0;
576
577 cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
578
579 ret = 0;
580 lock_sock(sk);
581 sockopt_seq_inc(msk);
582 mptcp_for_each_subflow(msk, subflow) {
583 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
584 int err;
585
586 lock_sock(ssk);
587 err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
588 if (err < 0 && ret == 0)
589 ret = err;
590 subflow->setsockopt_seq = msk->setsockopt_seq;
591 release_sock(ssk);
592 }
593
594 if (ret == 0)
20b5759f 595 strcpy(msk->ca_name, name);
aa1fbd94
FW		 596
597 release_sock(sk);
598 return ret;
599}
600
601static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
602 sockptr_t optval, unsigned int optlen)
603{
604 switch (optname) {
605 case TCP_ULP:
606 return -EOPNOTSUPP;
607 case TCP_CONGESTION:
608 return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
609 }
610
611 return -EOPNOTSUPP;
612}
613
0abdde82
PA		 614int mptcp_setsockopt(struct sock *sk, int level, int optname,
615 sockptr_t optval, unsigned int optlen)
616{
617 struct mptcp_sock *msk = mptcp_sk(sk);
618 struct sock *ssk;
619
620 pr_debug("msk=%p", msk);
621
622 if (level == SOL_SOCKET)
623 return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
624
7a009a70
FW		 625 if (!mptcp_supported_sockopt(level, optname))
626 return -ENOPROTOOPT;
627
0abdde82
PA		 628 /* @@ the meaning of setsockopt() when the socket is connected and
629 * there are multiple subflows is not yet defined. It is up to the
630 * MPTCP-level socket to configure the subflows until the subflow
631 * is in TCP fallback, when TCP socket options are passed through
632 * to the one remaining subflow.
633 */
634 lock_sock(sk);
635 ssk = __mptcp_tcp_fallback(msk);
636 release_sock(sk);
637 if (ssk)
638 return tcp_setsockopt(ssk, level, optname, optval, optlen);
639
640 if (level == SOL_IPV6)
641 return mptcp_setsockopt_v6(msk, optname, optval, optlen);
642
aa1fbd94
FW		 643 if (level == SOL_TCP)
644 return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
645
646 return -EOPNOTSUPP;
647}
648
649static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
650 char __user *optval, int __user *optlen)
651{
652 struct sock *sk = (struct sock *)msk;
653 struct socket *ssock;
654 int ret = -EINVAL;
655 struct sock *ssk;
656
657 lock_sock(sk);
658 ssk = msk->first;
659 if (ssk) {
660 ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
661 goto out;
662 }
663
664 ssock = __mptcp_nmpc_socket(msk);
665 if (!ssock)
666 goto out;
667
668 ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);
669
670out:
671 release_sock(sk);
672 return ret;
673}
674
61bc6e82
FW		 675void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
676{
677 struct sock *sk = &msk->sk.icsk_inet.sk;
61bc6e82 678 u32 flags = 0;
55c42fa7 679 bool slow;
61bc6e82
FW		 680 u8 val;
681
55c42fa7
FW		 682 memset(info, 0, sizeof(*info));
683
684 slow = lock_sock_fast(sk);
685
61bc6e82
FW		 686 info->mptcpi_subflows = READ_ONCE(msk->pm.subflows);
687 info->mptcpi_add_addr_signal = READ_ONCE(msk->pm.add_addr_signaled);
688 info->mptcpi_add_addr_accepted = READ_ONCE(msk->pm.add_addr_accepted);
689 info->mptcpi_local_addr_used = READ_ONCE(msk->pm.local_addr_used);
690 info->mptcpi_subflows_max = mptcp_pm_get_subflows_max(msk);
691 val = mptcp_pm_get_add_addr_signal_max(msk);
692 info->mptcpi_add_addr_signal_max = val;
693 val = mptcp_pm_get_add_addr_accept_max(msk);
694 info->mptcpi_add_addr_accepted_max = val;
695 info->mptcpi_local_addr_max = mptcp_pm_get_local_addr_max(msk);
696 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags))
697 flags |= MPTCP_INFO_FLAG_FALLBACK;
698 if (READ_ONCE(msk->can_ack))
699 flags |= MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED;
700 info->mptcpi_flags = flags;
701 info->mptcpi_token = READ_ONCE(msk->token);
702 info->mptcpi_write_seq = READ_ONCE(msk->write_seq);
703 info->mptcpi_snd_una = READ_ONCE(msk->snd_una);
704 info->mptcpi_rcv_nxt = READ_ONCE(msk->ack_seq);
705 info->mptcpi_csum_enabled = READ_ONCE(msk->csum_enabled);
706
707 unlock_sock_fast(sk, slow);
708}
709EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
710
55c42fa7
FW		 711static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
712{
713 struct mptcp_info m_info;
714 int len;
715
716 if (get_user(len, optlen))
717 return -EFAULT;
718
719 len = min_t(unsigned int, len, sizeof(struct mptcp_info));
720
721 mptcp_diag_fill_info(msk, &m_info);
722
723 if (put_user(len, optlen))
724 return -EFAULT;
725
726 if (copy_to_user(optval, &m_info, len))
727 return -EFAULT;
728
729 return 0;
730}
731
06f15cee
FW		 732static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
733 char __user *optval,
734 u32 copied,
735 int __user *optlen)
736{
737 u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
738
739 if (copied)
740 copied += sfd->size_subflow_data;
741 else
742 copied = copylen;
743
744 if (put_user(copied, optlen))
745 return -EFAULT;
746
747 if (copy_to_user(optval, sfd, copylen))
748 return -EFAULT;
749
750 return 0;
751}
752
753static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
754 char __user *optval, int __user *optlen)
755{
756 int len, copylen;
757
758 if (get_user(len, optlen))
759 return -EFAULT;
760
761 /* if mptcp_subflow_data size is changed, need to adjust
762 * this function to deal with programs using old version.
763 */
764 BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
765
766 if (len < MIN_INFO_OPTLEN_SIZE)
767 return -EINVAL;
768
769 memset(sfd, 0, sizeof(*sfd));
770
771 copylen = min_t(unsigned int, len, sizeof(*sfd));
772 if (copy_from_user(sfd, optval, copylen))
773 return -EFAULT;
774
775 /* size_subflow_data is u32, but len is signed */
776 if (sfd->size_subflow_data > INT_MAX ||
777 sfd->size_user > INT_MAX)
778 return -EINVAL;
779
780 if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
781 sfd->size_subflow_data > len)
782 return -EINVAL;
783
784 if (sfd->num_subflows || sfd->size_kernel)
785 return -EINVAL;
786
787 return len - sfd->size_subflow_data;
788}
789
790static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
791 int __user *optlen)
792{
793 struct mptcp_subflow_context *subflow;
794 struct sock *sk = &msk->sk.icsk_inet.sk;
795 unsigned int sfcount = 0, copied = 0;
796 struct mptcp_subflow_data sfd;
797 char __user *infoptr;
798 int len;
799
800 len = mptcp_get_subflow_data(&sfd, optval, optlen);
801 if (len < 0)
802 return len;
803
804 sfd.size_kernel = sizeof(struct tcp_info);
805 sfd.size_user = min_t(unsigned int, sfd.size_user,
806 sizeof(struct tcp_info));
807
808 infoptr = optval + sfd.size_subflow_data;
809
810 lock_sock(sk);
811
812 mptcp_for_each_subflow(msk, subflow) {
813 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
814
815 ++sfcount;
816
817 if (len && len >= sfd.size_user) {
818 struct tcp_info info;
819
820 tcp_get_info(ssk, &info);
821
822 if (copy_to_user(infoptr, &info, sfd.size_user)) {
823 release_sock(sk);
824 return -EFAULT;
825 }
826
827 infoptr += sfd.size_user;
828 copied += sfd.size_user;
829 len -= sfd.size_user;
830 }
831 }
832
833 release_sock(sk);
834
835 sfd.num_subflows = sfcount;
836
837 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
838 return -EFAULT;
839
840 return 0;
841}
842
c11c5906
FW		 843static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
844{
845 struct inet_sock *inet = inet_sk(sk);
846
847 memset(a, 0, sizeof(*a));
848
849 if (sk->sk_family == AF_INET) {
850 a->sin_local.sin_family = AF_INET;
851 a->sin_local.sin_port = inet->inet_sport;
852 a->sin_local.sin_addr.s_addr = inet->inet_rcv_saddr;
853
854 if (!a->sin_local.sin_addr.s_addr)
855 a->sin_local.sin_addr.s_addr = inet->inet_saddr;
856
857 a->sin_remote.sin_family = AF_INET;
858 a->sin_remote.sin_port = inet->inet_dport;
859 a->sin_remote.sin_addr.s_addr = inet->inet_daddr;
860#if IS_ENABLED(CONFIG_IPV6)
861 } else if (sk->sk_family == AF_INET6) {
862 const struct ipv6_pinfo *np = inet6_sk(sk);
863
864 a->sin6_local.sin6_family = AF_INET6;
865 a->sin6_local.sin6_port = inet->inet_sport;
866
867 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
868 a->sin6_local.sin6_addr = np->saddr;
869 else
870 a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
871
872 a->sin6_remote.sin6_family = AF_INET6;
873 a->sin6_remote.sin6_port = inet->inet_dport;
874 a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
875#endif
876 }
877}
878
879static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
880 int __user *optlen)
881{
882 struct sock *sk = &msk->sk.icsk_inet.sk;
883 struct mptcp_subflow_context *subflow;
884 unsigned int sfcount = 0, copied = 0;
885 struct mptcp_subflow_data sfd;
886 char __user *addrptr;
887 int len;
888
889 len = mptcp_get_subflow_data(&sfd, optval, optlen);
890 if (len < 0)
891 return len;
892
893 sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
894 sfd.size_user = min_t(unsigned int, sfd.size_user,
895 sizeof(struct mptcp_subflow_addrs));
896
897 addrptr = optval + sfd.size_subflow_data;
898
899 lock_sock(sk);
900
901 mptcp_for_each_subflow(msk, subflow) {
902 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
903
904 ++sfcount;
905
906 if (len && len >= sfd.size_user) {
907 struct mptcp_subflow_addrs a;
908
909 mptcp_get_sub_addrs(ssk, &a);
910
911 if (copy_to_user(addrptr, &a, sfd.size_user)) {
912 release_sock(sk);
913 return -EFAULT;
914 }
915
916 addrptr += sfd.size_user;
917 copied += sfd.size_user;
918 len -= sfd.size_user;
919 }
920 }
921
922 release_sock(sk);
923
924 sfd.num_subflows = sfcount;
925
926 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
927 return -EFAULT;
928
929 return 0;
930}
931
aa1fbd94
FW		 932static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
933 char __user *optval, int __user *optlen)
934{
935 switch (optname) {
936 case TCP_ULP:
937 case TCP_CONGESTION:
938 case TCP_INFO:
939 case TCP_CC_INFO:
940 return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
941 optval, optlen);
942 }
0abdde82
PA		 943 return -EOPNOTSUPP;
944}
945
55c42fa7
FW		 946static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
947 char __user *optval, int __user *optlen)
948{
949 switch (optname) {
950 case MPTCP_INFO:
951 return mptcp_getsockopt_info(msk, optval, optlen);
06f15cee
FW		 952 case MPTCP_TCPINFO:
953 return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
c11c5906
FW		 954 case MPTCP_SUBFLOW_ADDRS:
955 return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
55c42fa7
FW		 956 }
957
958 return -EOPNOTSUPP;
959}
960
0abdde82
PA		 961int mptcp_getsockopt(struct sock *sk, int level, int optname,
962 char __user *optval, int __user *option)
963{
964 struct mptcp_sock *msk = mptcp_sk(sk);
965 struct sock *ssk;
966
967 pr_debug("msk=%p", msk);
968
969 /* @@ the meaning of setsockopt() when the socket is connected and
970 * there are multiple subflows is not yet defined. It is up to the
971 * MPTCP-level socket to configure the subflows until the subflow
972 * is in TCP fallback, when socket options are passed through
973 * to the one remaining subflow.
974 */
975 lock_sock(sk);
976 ssk = __mptcp_tcp_fallback(msk);
977 release_sock(sk);
978 if (ssk)
979 return tcp_getsockopt(ssk, level, optname, optval, option);
980
aa1fbd94
FW		 981 if (level == SOL_TCP)
982 return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
55c42fa7
FW		 983 if (level == SOL_MPTCP)
984 return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
0abdde82
PA		 985 return -EOPNOTSUPP;
986}
987
1b3e7ede
FW		 988static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
989{
5d0a6bc8 990 static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
1b3e7ede
FW		 991 struct sock *sk = (struct sock *)msk;
992
993 if (ssk->sk_prot->keepalive) {
994 if (sock_flag(sk, SOCK_KEEPOPEN))
995 ssk->sk_prot->keepalive(ssk, 1);
996 else
997 ssk->sk_prot->keepalive(ssk, 0);
998 }
999
1000 ssk->sk_priority = sk->sk_priority;
5d0a6bc8
FW		 1001 ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1002 ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
1003
1004 if (sk->sk_userlocks & tx_rx_locks) {
1005 ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1006 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
1007 WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1008 if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1009 WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1010 }
1011
1012 if (sock_flag(sk, SOCK_LINGER)) {
1013 ssk->sk_lingertime = sk->sk_lingertime;
1014 sock_set_flag(ssk, SOCK_LINGER);
1015 } else {
1016 sock_reset_flag(ssk, SOCK_LINGER);
1017 }
1018
1019 if (sk->sk_mark != ssk->sk_mark) {
1020 ssk->sk_mark = sk->sk_mark;
1021 sk_dst_reset(ssk);
1022 }
1023
1024 sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1025
1026 if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
20b5759f 1027 tcp_set_congestion_control(ssk, msk->ca_name, false, true);
1b3e7ede
FW		 1028}
1029
df00b087
FW		 1030static void __mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1031{
1b3e7ede
FW		 1032 bool slow = lock_sock_fast(ssk);
1033
1034 sync_socket_options(msk, ssk);
1035
1036 unlock_sock_fast(ssk, slow);
df00b087
FW		 1037}
1038
78962489
FW		 1039void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1040{
df00b087
FW		 1041 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1042
78962489 1043 msk_owned_by_me(msk);
df00b087
FW		 1044
1045 if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1046 __mptcp_sockopt_sync(msk, ssk);
1047
1048 subflow->setsockopt_seq = msk->setsockopt_seq;
1049 }
78962489
FW		 1050}
1051
1052void mptcp_sockopt_sync_all(struct mptcp_sock *msk)
1053{
1054 struct mptcp_subflow_context *subflow;
df00b087
FW		 1055 struct sock *sk = (struct sock *)msk;
1056 u32 seq;
78962489 1057
df00b087 1058 seq = sockopt_seq_reset(sk);
78962489
FW		 1059
1060 mptcp_for_each_subflow(msk, subflow) {
1061 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
df00b087 1062 u32 sseq = READ_ONCE(subflow->setsockopt_seq);
78962489 1063
df00b087
FW		 1064 if (sseq != msk->setsockopt_seq) {
1065 __mptcp_sockopt_sync(msk, ssk);
1066 WRITE_ONCE(subflow->setsockopt_seq, seq);
1067 } else if (sseq != seq) {
1068 WRITE_ONCE(subflow->setsockopt_seq, seq);
1069 }
78962489
FW		 1070
1071 cond_resched();
1072 }
df00b087
FW		 1073
1074 msk->setsockopt_seq = seq;
78962489 1075}
Linux 6.x block layer and io_uring tree(s)