1 // SPDX-License-Identifier: GPL-2.0
4 * Copyright (c) 2021, Red Hat.
7 #define pr_fmt(fmt) "MPTCP: " fmt
9 #include <linux/kernel.h>
10 #include <linux/module.h>
12 #include <net/protocol.h>
14 #include <net/mptcp.h>
17 #define MIN_INFO_OPTLEN_SIZE 16
19 static struct sock *__mptcp_tcp_fallback(struct mptcp_sock *msk)
21 sock_owned_by_me((const struct sock *)msk);
23 if (likely(!__mptcp_check_fallback(msk)))
29 static u32 sockopt_seq_reset(const struct sock *sk)
33 /* Highbits contain state. Allows to distinguish sockopt_seq
34 * of listener and established:
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
43 * Set High order bits to sk_state so ssk->seq == msk->seq test
47 return (u32)sk->sk_state << 24u;
50 static void sockopt_seq_inc(struct mptcp_sock *msk)
52 u32 seq = (msk->setsockopt_seq + 1) & 0x00ffffff;
54 msk->setsockopt_seq = sockopt_seq_reset((struct sock *)msk) + seq;
57 static int mptcp_get_int_option(struct mptcp_sock *msk, sockptr_t optval,
58 unsigned int optlen, int *val)
60 if (optlen < sizeof(int))
63 if (copy_from_sockptr(val, optval, sizeof(*val)))
69 static void mptcp_sol_socket_sync_intval(struct mptcp_sock *msk, int optname, int val)
71 struct mptcp_subflow_context *subflow;
72 struct sock *sk = (struct sock *)msk;
77 mptcp_for_each_subflow(msk, subflow) {
78 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
79 bool slow = lock_sock_fast(ssk);
83 sock_valbool_flag(ssk, SOCK_DBG, !!val);
86 if (ssk->sk_prot->keepalive)
87 ssk->sk_prot->keepalive(ssk, !!val);
88 sock_valbool_flag(ssk, SOCK_KEEPOPEN, !!val);
91 ssk->sk_priority = val;
95 ssk->sk_userlocks |= SOCK_SNDBUF_LOCK;
96 WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
100 ssk->sk_userlocks |= SOCK_RCVBUF_LOCK;
101 WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
104 if (READ_ONCE(ssk->sk_mark) != sk->sk_mark) {
105 ssk->sk_mark = sk->sk_mark;
109 case SO_INCOMING_CPU:
110 WRITE_ONCE(ssk->sk_incoming_cpu, val);
114 subflow->setsockopt_seq = msk->setsockopt_seq;
115 unlock_sock_fast(ssk, slow);
121 static int mptcp_sol_socket_intval(struct mptcp_sock *msk, int optname, int val)
123 sockptr_t optval = KERNEL_SOCKPTR(&val);
124 struct sock *sk = (struct sock *)msk;
127 ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
128 optval, sizeof(val));
132 mptcp_sol_socket_sync_intval(msk, optname, val);
136 static void mptcp_so_incoming_cpu(struct mptcp_sock *msk, int val)
138 struct sock *sk = (struct sock *)msk;
140 WRITE_ONCE(sk->sk_incoming_cpu, val);
142 mptcp_sol_socket_sync_intval(msk, SO_INCOMING_CPU, val);
145 static int mptcp_setsockopt_sol_socket_tstamp(struct mptcp_sock *msk, int optname, int val)
147 sockptr_t optval = KERNEL_SOCKPTR(&val);
148 struct mptcp_subflow_context *subflow;
149 struct sock *sk = (struct sock *)msk;
152 ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
153 optval, sizeof(val));
158 mptcp_for_each_subflow(msk, subflow) {
159 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
160 bool slow = lock_sock_fast(ssk);
162 sock_set_timestamp(sk, optname, !!val);
163 unlock_sock_fast(ssk, slow);
170 static int mptcp_setsockopt_sol_socket_int(struct mptcp_sock *msk, int optname,
176 ret = mptcp_get_int_option(msk, optval, optlen, &val);
182 mptcp_sol_socket_sync_intval(msk, optname, val);
191 return mptcp_sol_socket_intval(msk, optname, val);
192 case SO_INCOMING_CPU:
193 mptcp_so_incoming_cpu(msk, val);
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);
205 static int mptcp_setsockopt_sol_socket_timestamping(struct mptcp_sock *msk,
210 struct mptcp_subflow_context *subflow;
211 struct sock *sk = (struct sock *)msk;
212 struct so_timestamping timestamping;
215 if (optlen == sizeof(timestamping)) {
216 if (copy_from_sockptr(×tamping, optval,
217 sizeof(timestamping)))
219 } else if (optlen == sizeof(int)) {
220 memset(×tamping, 0, sizeof(timestamping));
222 if (copy_from_sockptr(×tamping.flags, optval, sizeof(int)))
228 ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname,
229 KERNEL_SOCKPTR(×tamping),
230 sizeof(timestamping));
236 mptcp_for_each_subflow(msk, subflow) {
237 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
238 bool slow = lock_sock_fast(ssk);
240 sock_set_timestamping(sk, optname, timestamping);
241 unlock_sock_fast(ssk, slow);
249 static int mptcp_setsockopt_sol_socket_linger(struct mptcp_sock *msk, sockptr_t optval,
252 struct mptcp_subflow_context *subflow;
253 struct sock *sk = (struct sock *)msk;
258 if (optlen < sizeof(ling))
261 if (copy_from_sockptr(&ling, optval, sizeof(ling)))
264 kopt = KERNEL_SOCKPTR(&ling);
265 ret = sock_setsockopt(sk->sk_socket, SOL_SOCKET, SO_LINGER, kopt, sizeof(ling));
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);
276 sock_reset_flag(ssk, SOCK_LINGER);
278 ssk->sk_lingertime = sk->sk_lingertime;
279 sock_set_flag(ssk, SOCK_LINGER);
282 subflow->setsockopt_seq = msk->setsockopt_seq;
283 unlock_sock_fast(ssk, slow);
290 static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,
291 sockptr_t optval, unsigned int optlen)
293 struct sock *sk = (struct sock *)msk;
294 struct socket *ssock;
300 case SO_BINDTODEVICE:
301 case SO_BINDTOIFINDEX:
303 ssock = __mptcp_nmpc_socket(msk);
309 ret = sock_setsockopt(ssock, SOL_SOCKET, optname, optval, optlen);
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;
329 case SO_INCOMING_CPU:
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,
337 case SO_TIMESTAMPING_OLD:
338 case SO_TIMESTAMPING_NEW:
339 return mptcp_setsockopt_sol_socket_timestamping(msk, optname,
342 return mptcp_setsockopt_sol_socket_linger(msk, optval, optlen);
344 case SO_RCVTIMEO_OLD:
345 case SO_RCVTIMEO_NEW:
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);
360 case SO_SELECT_ERR_QUEUE:
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
369 * SO_ATTACH_REUSEPORT_EBPF is not supported, at it checks
370 * explicitly the sk_protocol field
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
377 * SO_ZEROCOPY is currently unsupported, TODO in sndmsg
378 * SO_TXTIME is currently unsupported
384 static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,
385 sockptr_t optval, unsigned int optlen)
387 struct sock *sk = (struct sock *)msk;
388 int ret = -EOPNOTSUPP;
389 struct socket *ssock;
394 ssock = __mptcp_nmpc_socket(msk);
400 ret = tcp_setsockopt(ssock->sk, SOL_IPV6, optname, optval, optlen);
402 sk->sk_ipv6only = ssock->sk->sk_ipv6only;
411 static bool mptcp_supported_sockopt(int level, int optname)
413 if (level == SOL_IP) {
415 /* should work fine */
419 /* the following are control cmsg related */
426 case IP_RECVORIGDSTADDR:
428 case IP_RECVFRAGSIZE:
430 /* common stuff that need some love */
433 case IP_BIND_ADDRESS_NO_PORT:
434 case IP_MTU_DISCOVER:
437 /* possibly less common may deserve some love */
440 /* the following is apparently a no-op for plain TCP */
441 case IP_RECVERR_RFC4884:
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
455 /* IP_IPSEC_POLICY, IP_XFRM_POLICY are nut supported, unrelated here */
458 if (level == SOL_IPV6) {
462 /* the following are control cmsg related */
463 case IPV6_RECVPKTINFO:
464 case IPV6_2292PKTINFO:
465 case IPV6_RECVHOPLIMIT:
466 case IPV6_2292HOPLIMIT:
469 case IPV6_RECVHOPOPTS:
470 case IPV6_2292HOPOPTS:
471 case IPV6_RECVDSTOPTS:
472 case IPV6_2292DSTOPTS:
473 case IPV6_RECVTCLASS:
475 case IPV6_RECVPATHMTU:
476 case IPV6_RECVORIGDSTADDR:
477 case IPV6_RECVFRAGSIZE:
479 /* the following ones need some love but are quite common */
481 case IPV6_TRANSPARENT:
484 case IPV6_2292PKTOPTIONS:
485 case IPV6_UNICAST_HOPS:
486 case IPV6_MTU_DISCOVER:
489 case IPV6_FLOWINFO_SEND:
490 case IPV6_FLOWLABEL_MGR:
491 case IPV6_MINHOPCOUNT:
493 case IPV6_AUTOFLOWLABEL:
495 /* the following one is a no-op for plain TCP */
496 case IPV6_RECVERR_RFC4884:
500 /* IPV6_HOPOPTS, IPV6_RTHDRDSTOPTS, IPV6_RTHDR, IPV6_DSTOPTS are
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
511 /* IPV6_ROUTER_ALERT, IPV6_ROUTER_ALERT_ISOLATE are not supported, since are evil */
513 /* IPV6_IPSEC_POLICY, IPV6_XFRM_POLICY are not supported */
514 /* IPV6_ADDR_PREFERENCES is not supported, we must be careful with subflows */
517 if (level == SOL_TCP) {
519 /* the following are no-op or should work just fine */
520 case TCP_THIN_DUPACK:
521 case TCP_DEFER_ACCEPT:
523 /* the following need some love */
526 case TCP_THIN_LINEAR_TIMEOUTS:
536 case TCP_WINDOW_CLAMP:
538 case TCP_USER_TIMEOUT:
540 case TCP_NOTSENT_LOWAT:
545 /* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
547 /* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
548 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
550 /* TCP_FASTOPEN_KEY, TCP_FASTOPEN TCP_FASTOPEN_CONNECT, TCP_FASTOPEN_NO_COOKIE,
551 * are not supported fastopen is currently unsupported
553 /* TCP_INQ is currently unsupported, needs some recvmsg work */
558 static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
561 struct mptcp_subflow_context *subflow;
562 struct sock *sk = (struct sock *)msk;
563 char name[TCP_CA_NAME_MAX];
570 ret = strncpy_from_sockptr(name, optval,
571 min_t(long, TCP_CA_NAME_MAX - 1, optlen));
577 cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
581 sockopt_seq_inc(msk);
582 mptcp_for_each_subflow(msk, subflow) {
583 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
587 err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
588 if (err < 0 && ret == 0)
590 subflow->setsockopt_seq = msk->setsockopt_seq;
595 strcpy(msk->ca_name, name);
601 static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
602 sockptr_t optval, unsigned int optlen)
608 return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
614 int mptcp_setsockopt(struct sock *sk, int level, int optname,
615 sockptr_t optval, unsigned int optlen)
617 struct mptcp_sock *msk = mptcp_sk(sk);
620 pr_debug("msk=%p", msk);
622 if (level == SOL_SOCKET)
623 return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
625 if (!mptcp_supported_sockopt(level, optname))
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.
635 ssk = __mptcp_tcp_fallback(msk);
638 return tcp_setsockopt(ssk, level, optname, optval, optlen);
640 if (level == SOL_IPV6)
641 return mptcp_setsockopt_v6(msk, optname, optval, optlen);
643 if (level == SOL_TCP)
644 return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
649 static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
650 char __user *optval, int __user *optlen)
652 struct sock *sk = (struct sock *)msk;
653 struct socket *ssock;
660 ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
664 ssock = __mptcp_nmpc_socket(msk);
668 ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);
675 void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
677 struct sock *sk = &msk->sk.icsk_inet.sk;
682 memset(info, 0, sizeof(*info));
684 slow = lock_sock_fast(sk);
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);
707 unlock_sock_fast(sk, slow);
709 EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
711 static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
713 struct mptcp_info m_info;
716 if (get_user(len, optlen))
719 len = min_t(unsigned int, len, sizeof(struct mptcp_info));
721 mptcp_diag_fill_info(msk, &m_info);
723 if (put_user(len, optlen))
726 if (copy_to_user(optval, &m_info, len))
732 static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
737 u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
740 copied += sfd->size_subflow_data;
744 if (put_user(copied, optlen))
747 if (copy_to_user(optval, sfd, copylen))
753 static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
754 char __user *optval, int __user *optlen)
758 if (get_user(len, optlen))
761 /* if mptcp_subflow_data size is changed, need to adjust
762 * this function to deal with programs using old version.
764 BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
766 if (len < MIN_INFO_OPTLEN_SIZE)
769 memset(sfd, 0, sizeof(*sfd));
771 copylen = min_t(unsigned int, len, sizeof(*sfd));
772 if (copy_from_user(sfd, optval, copylen))
775 /* size_subflow_data is u32, but len is signed */
776 if (sfd->size_subflow_data > INT_MAX ||
777 sfd->size_user > INT_MAX)
780 if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
781 sfd->size_subflow_data > len)
784 if (sfd->num_subflows || sfd->size_kernel)
787 return len - sfd->size_subflow_data;
790 static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
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;
800 len = mptcp_get_subflow_data(&sfd, optval, optlen);
804 sfd.size_kernel = sizeof(struct tcp_info);
805 sfd.size_user = min_t(unsigned int, sfd.size_user,
806 sizeof(struct tcp_info));
808 infoptr = optval + sfd.size_subflow_data;
812 mptcp_for_each_subflow(msk, subflow) {
813 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
817 if (len && len >= sfd.size_user) {
818 struct tcp_info info;
820 tcp_get_info(ssk, &info);
822 if (copy_to_user(infoptr, &info, sfd.size_user)) {
827 infoptr += sfd.size_user;
828 copied += sfd.size_user;
829 len -= sfd.size_user;
835 sfd.num_subflows = sfcount;
837 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
843 static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
845 struct inet_sock *inet = inet_sk(sk);
847 memset(a, 0, sizeof(*a));
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;
854 if (!a->sin_local.sin_addr.s_addr)
855 a->sin_local.sin_addr.s_addr = inet->inet_saddr;
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);
864 if (WARN_ON_ONCE(!np))
867 a->sin6_local.sin6_family = AF_INET6;
868 a->sin6_local.sin6_port = inet->inet_sport;
870 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
871 a->sin6_local.sin6_addr = np->saddr;
873 a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
875 a->sin6_remote.sin6_family = AF_INET6;
876 a->sin6_remote.sin6_port = inet->inet_dport;
877 a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
882 static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
885 struct sock *sk = &msk->sk.icsk_inet.sk;
886 struct mptcp_subflow_context *subflow;
887 unsigned int sfcount = 0, copied = 0;
888 struct mptcp_subflow_data sfd;
889 char __user *addrptr;
892 len = mptcp_get_subflow_data(&sfd, optval, optlen);
896 sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
897 sfd.size_user = min_t(unsigned int, sfd.size_user,
898 sizeof(struct mptcp_subflow_addrs));
900 addrptr = optval + sfd.size_subflow_data;
904 mptcp_for_each_subflow(msk, subflow) {
905 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
909 if (len && len >= sfd.size_user) {
910 struct mptcp_subflow_addrs a;
912 mptcp_get_sub_addrs(ssk, &a);
914 if (copy_to_user(addrptr, &a, sfd.size_user)) {
919 addrptr += sfd.size_user;
920 copied += sfd.size_user;
921 len -= sfd.size_user;
927 sfd.num_subflows = sfcount;
929 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
935 static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
936 char __user *optval, int __user *optlen)
943 return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
949 static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
950 char __user *optval, int __user *optlen)
954 return mptcp_getsockopt_info(msk, optval, optlen);
956 return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
957 case MPTCP_SUBFLOW_ADDRS:
958 return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
964 int mptcp_getsockopt(struct sock *sk, int level, int optname,
965 char __user *optval, int __user *option)
967 struct mptcp_sock *msk = mptcp_sk(sk);
970 pr_debug("msk=%p", msk);
972 /* @@ the meaning of setsockopt() when the socket is connected and
973 * there are multiple subflows is not yet defined. It is up to the
974 * MPTCP-level socket to configure the subflows until the subflow
975 * is in TCP fallback, when socket options are passed through
976 * to the one remaining subflow.
979 ssk = __mptcp_tcp_fallback(msk);
982 return tcp_getsockopt(ssk, level, optname, optval, option);
984 if (level == SOL_TCP)
985 return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
986 if (level == SOL_MPTCP)
987 return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
991 static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
993 static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
994 struct sock *sk = (struct sock *)msk;
996 if (ssk->sk_prot->keepalive) {
997 if (sock_flag(sk, SOCK_KEEPOPEN))
998 ssk->sk_prot->keepalive(ssk, 1);
1000 ssk->sk_prot->keepalive(ssk, 0);
1003 ssk->sk_priority = sk->sk_priority;
1004 ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1005 ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
1007 if (sk->sk_userlocks & tx_rx_locks) {
1008 ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1009 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
1010 WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1011 if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1012 WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1015 if (sock_flag(sk, SOCK_LINGER)) {
1016 ssk->sk_lingertime = sk->sk_lingertime;
1017 sock_set_flag(ssk, SOCK_LINGER);
1019 sock_reset_flag(ssk, SOCK_LINGER);
1022 if (sk->sk_mark != ssk->sk_mark) {
1023 ssk->sk_mark = sk->sk_mark;
1027 sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1029 if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
1030 tcp_set_congestion_control(ssk, msk->ca_name, false, true);
1033 static void __mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1035 bool slow = lock_sock_fast(ssk);
1037 sync_socket_options(msk, ssk);
1039 unlock_sock_fast(ssk, slow);
1042 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1044 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1046 msk_owned_by_me(msk);
1048 if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1049 __mptcp_sockopt_sync(msk, ssk);
1051 subflow->setsockopt_seq = msk->setsockopt_seq;
1055 void mptcp_sockopt_sync_all(struct mptcp_sock *msk)
1057 struct mptcp_subflow_context *subflow;
1058 struct sock *sk = (struct sock *)msk;
1061 seq = sockopt_seq_reset(sk);
1063 mptcp_for_each_subflow(msk, subflow) {
1064 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1065 u32 sseq = READ_ONCE(subflow->setsockopt_seq);
1067 if (sseq != msk->setsockopt_seq) {
1068 __mptcp_sockopt_sync(msk, ssk);
1069 WRITE_ONCE(subflow->setsockopt_seq, seq);
1070 } else if (sseq != seq) {
1071 WRITE_ONCE(subflow->setsockopt_seq, seq);
1077 msk->setsockopt_seq = seq;