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)
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);
306 return PTR_ERR(ssock);
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:
346 case SO_SNDTIMEO_OLD:
347 case SO_SNDTIMEO_NEW:
349 case SO_PREFER_BUSY_POLL:
350 case SO_BUSY_POLL_BUDGET:
351 /* No need to copy: only relevant for msk */
352 return sock_setsockopt(sk->sk_socket, SOL_SOCKET, optname, optval, optlen);
363 case SO_SELECT_ERR_QUEUE:
367 /* SO_OOBINLINE is not supported, let's avoid the related mess
368 * SO_ATTACH_FILTER, SO_ATTACH_BPF, SO_ATTACH_REUSEPORT_CBPF,
369 * SO_DETACH_REUSEPORT_BPF, SO_DETACH_FILTER, SO_LOCK_FILTER,
370 * we must be careful with subflows
372 * SO_ATTACH_REUSEPORT_EBPF is not supported, at it checks
373 * explicitly the sk_protocol field
375 * SO_PEEK_OFF is unsupported, as it is for plain TCP
376 * SO_MAX_PACING_RATE is unsupported, we must be careful with subflows
377 * SO_CNX_ADVICE is currently unsupported, could possibly be relevant,
378 * but likely needs careful design
380 * SO_ZEROCOPY is currently unsupported, TODO in sndmsg
381 * SO_TXTIME is currently unsupported
387 static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,
388 sockptr_t optval, unsigned int optlen)
390 struct sock *sk = (struct sock *)msk;
391 int ret = -EOPNOTSUPP;
392 struct socket *ssock;
396 case IPV6_TRANSPARENT:
399 ssock = __mptcp_nmpc_socket(msk);
402 return PTR_ERR(ssock);
405 ret = tcp_setsockopt(ssock->sk, SOL_IPV6, optname, optval, optlen);
411 sockopt_seq_inc(msk);
415 sk->sk_ipv6only = ssock->sk->sk_ipv6only;
417 case IPV6_TRANSPARENT:
418 inet_sk(sk)->transparent = inet_sk(ssock->sk)->transparent;
421 inet_sk(sk)->freebind = inet_sk(ssock->sk)->freebind;
432 static bool mptcp_supported_sockopt(int level, int optname)
434 if (level == SOL_IP) {
436 /* should work fine */
440 /* the following are control cmsg related */
447 case IP_RECVORIGDSTADDR:
449 case IP_RECVFRAGSIZE:
451 /* common stuff that need some love */
454 case IP_BIND_ADDRESS_NO_PORT:
455 case IP_MTU_DISCOVER:
458 /* possibly less common may deserve some love */
461 /* the following is apparently a no-op for plain TCP */
462 case IP_RECVERR_RFC4884:
466 /* IP_OPTIONS is not supported, needs subflow care */
467 /* IP_HDRINCL, IP_NODEFRAG are not supported, RAW specific */
468 /* IP_MULTICAST_TTL, IP_MULTICAST_LOOP, IP_UNICAST_IF,
469 * IP_ADD_MEMBERSHIP, IP_ADD_SOURCE_MEMBERSHIP, IP_DROP_MEMBERSHIP,
470 * IP_DROP_SOURCE_MEMBERSHIP, IP_BLOCK_SOURCE, IP_UNBLOCK_SOURCE,
471 * MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP MCAST_JOIN_SOURCE_GROUP,
472 * MCAST_LEAVE_SOURCE_GROUP, MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE,
473 * MCAST_MSFILTER, IP_MULTICAST_ALL are not supported, better not deal
476 /* IP_IPSEC_POLICY, IP_XFRM_POLICY are nut supported, unrelated here */
479 if (level == SOL_IPV6) {
483 /* the following are control cmsg related */
484 case IPV6_RECVPKTINFO:
485 case IPV6_2292PKTINFO:
486 case IPV6_RECVHOPLIMIT:
487 case IPV6_2292HOPLIMIT:
490 case IPV6_RECVHOPOPTS:
491 case IPV6_2292HOPOPTS:
492 case IPV6_RECVDSTOPTS:
493 case IPV6_2292DSTOPTS:
494 case IPV6_RECVTCLASS:
496 case IPV6_RECVPATHMTU:
497 case IPV6_RECVORIGDSTADDR:
498 case IPV6_RECVFRAGSIZE:
500 /* the following ones need some love but are quite common */
502 case IPV6_TRANSPARENT:
505 case IPV6_2292PKTOPTIONS:
506 case IPV6_UNICAST_HOPS:
507 case IPV6_MTU_DISCOVER:
510 case IPV6_FLOWINFO_SEND:
511 case IPV6_FLOWLABEL_MGR:
512 case IPV6_MINHOPCOUNT:
514 case IPV6_AUTOFLOWLABEL:
516 /* the following one is a no-op for plain TCP */
517 case IPV6_RECVERR_RFC4884:
521 /* IPV6_HOPOPTS, IPV6_RTHDRDSTOPTS, IPV6_RTHDR, IPV6_DSTOPTS are
524 /* IPV6_MULTICAST_HOPS, IPV6_MULTICAST_LOOP, IPV6_UNICAST_IF,
525 * IPV6_MULTICAST_IF, IPV6_ADDRFORM,
526 * IPV6_ADD_MEMBERSHIP, IPV6_DROP_MEMBERSHIP, IPV6_JOIN_ANYCAST,
527 * IPV6_LEAVE_ANYCAST, IPV6_MULTICAST_ALL, MCAST_JOIN_GROUP, MCAST_LEAVE_GROUP,
528 * MCAST_JOIN_SOURCE_GROUP, MCAST_LEAVE_SOURCE_GROUP,
529 * MCAST_BLOCK_SOURCE, MCAST_UNBLOCK_SOURCE, MCAST_MSFILTER
530 * are not supported better not deal with mcast
532 /* IPV6_ROUTER_ALERT, IPV6_ROUTER_ALERT_ISOLATE are not supported, since are evil */
534 /* IPV6_IPSEC_POLICY, IPV6_XFRM_POLICY are not supported */
535 /* IPV6_ADDR_PREFERENCES is not supported, we must be careful with subflows */
538 if (level == SOL_TCP) {
540 /* the following are no-op or should work just fine */
541 case TCP_THIN_DUPACK:
542 case TCP_DEFER_ACCEPT:
544 /* the following need some love */
547 case TCP_THIN_LINEAR_TIMEOUTS:
556 case TCP_WINDOW_CLAMP:
558 case TCP_USER_TIMEOUT:
560 case TCP_NOTSENT_LOWAT:
564 case TCP_FASTOPEN_CONNECT:
565 case TCP_FASTOPEN_KEY:
566 case TCP_FASTOPEN_NO_COOKIE:
570 /* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
572 /* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
573 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
579 static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
582 struct mptcp_subflow_context *subflow;
583 struct sock *sk = (struct sock *)msk;
584 char name[TCP_CA_NAME_MAX];
591 ret = strncpy_from_sockptr(name, optval,
592 min_t(long, TCP_CA_NAME_MAX - 1, optlen));
598 cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
602 sockopt_seq_inc(msk);
603 mptcp_for_each_subflow(msk, subflow) {
604 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
608 err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
609 if (err < 0 && ret == 0)
611 subflow->setsockopt_seq = msk->setsockopt_seq;
616 strcpy(msk->ca_name, name);
622 static int mptcp_setsockopt_sol_tcp_cork(struct mptcp_sock *msk, sockptr_t optval,
625 struct mptcp_subflow_context *subflow;
626 struct sock *sk = (struct sock *)msk;
629 if (optlen < sizeof(int))
632 if (copy_from_sockptr(&val, optval, sizeof(val)))
636 sockopt_seq_inc(msk);
638 mptcp_for_each_subflow(msk, subflow) {
639 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
642 __tcp_sock_set_cork(ssk, !!val);
646 mptcp_check_and_set_pending(sk);
652 static int mptcp_setsockopt_sol_tcp_nodelay(struct mptcp_sock *msk, sockptr_t optval,
655 struct mptcp_subflow_context *subflow;
656 struct sock *sk = (struct sock *)msk;
659 if (optlen < sizeof(int))
662 if (copy_from_sockptr(&val, optval, sizeof(val)))
666 sockopt_seq_inc(msk);
667 msk->nodelay = !!val;
668 mptcp_for_each_subflow(msk, subflow) {
669 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
672 __tcp_sock_set_nodelay(ssk, !!val);
676 mptcp_check_and_set_pending(sk);
682 static int mptcp_setsockopt_sol_ip_set_transparent(struct mptcp_sock *msk, int optname,
683 sockptr_t optval, unsigned int optlen)
685 struct sock *sk = (struct sock *)msk;
686 struct inet_sock *issk;
687 struct socket *ssock;
690 err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
696 ssock = __mptcp_nmpc_socket(msk);
699 return PTR_ERR(ssock);
702 issk = inet_sk(ssock->sk);
706 issk->freebind = inet_sk(sk)->freebind;
709 issk->transparent = inet_sk(sk)->transparent;
717 sockopt_seq_inc(msk);
722 static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,
723 sockptr_t optval, unsigned int optlen)
725 struct mptcp_subflow_context *subflow;
726 struct sock *sk = (struct sock *)msk;
729 err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
735 sockopt_seq_inc(msk);
736 val = inet_sk(sk)->tos;
737 mptcp_for_each_subflow(msk, subflow) {
738 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
740 __ip_sock_set_tos(ssk, val);
747 static int mptcp_setsockopt_v4(struct mptcp_sock *msk, int optname,
748 sockptr_t optval, unsigned int optlen)
753 return mptcp_setsockopt_sol_ip_set_transparent(msk, optname, optval, optlen);
755 return mptcp_setsockopt_v4_set_tos(msk, optname, optval, optlen);
761 static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
762 sockptr_t optval, unsigned int optlen)
764 struct sock *sk = (struct sock *)msk;
768 /* Limit to first subflow, before the connection establishment */
770 sock = __mptcp_nmpc_socket(msk);
776 ret = tcp_setsockopt(sock->sk, level, optname, optval, optlen);
783 static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
784 sockptr_t optval, unsigned int optlen)
786 struct sock *sk = (void *)msk;
791 ret = mptcp_get_int_option(msk, optval, optlen, &val);
794 if (val < 0 || val > 1)
798 msk->recvmsg_inq = !!val;
804 return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
806 return mptcp_setsockopt_sol_tcp_cork(msk, optval, optlen);
808 return mptcp_setsockopt_sol_tcp_nodelay(msk, optval, optlen);
809 case TCP_DEFER_ACCEPT:
810 /* See tcp.c: TCP_DEFER_ACCEPT does not fail */
811 mptcp_setsockopt_first_sf_only(msk, SOL_TCP, optname, optval, optlen);
814 case TCP_FASTOPEN_CONNECT:
815 case TCP_FASTOPEN_KEY:
816 case TCP_FASTOPEN_NO_COOKIE:
817 return mptcp_setsockopt_first_sf_only(msk, SOL_TCP, optname,
824 int mptcp_setsockopt(struct sock *sk, int level, int optname,
825 sockptr_t optval, unsigned int optlen)
827 struct mptcp_sock *msk = mptcp_sk(sk);
830 pr_debug("msk=%p", msk);
832 if (level == SOL_SOCKET)
833 return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
835 if (!mptcp_supported_sockopt(level, optname))
838 /* @@ the meaning of setsockopt() when the socket is connected and
839 * there are multiple subflows is not yet defined. It is up to the
840 * MPTCP-level socket to configure the subflows until the subflow
841 * is in TCP fallback, when TCP socket options are passed through
842 * to the one remaining subflow.
845 ssk = __mptcp_tcp_fallback(msk);
848 return tcp_setsockopt(ssk, level, optname, optval, optlen);
851 return mptcp_setsockopt_v4(msk, optname, optval, optlen);
853 if (level == SOL_IPV6)
854 return mptcp_setsockopt_v6(msk, optname, optval, optlen);
856 if (level == SOL_TCP)
857 return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
862 static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
863 char __user *optval, int __user *optlen)
865 struct sock *sk = (struct sock *)msk;
866 struct socket *ssock;
873 ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
877 ssock = __mptcp_nmpc_socket(msk);
879 ret = PTR_ERR(ssock);
883 ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);
890 void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
894 memset(info, 0, sizeof(*info));
896 info->mptcpi_subflows = READ_ONCE(msk->pm.subflows);
897 info->mptcpi_add_addr_signal = READ_ONCE(msk->pm.add_addr_signaled);
898 info->mptcpi_add_addr_accepted = READ_ONCE(msk->pm.add_addr_accepted);
899 info->mptcpi_local_addr_used = READ_ONCE(msk->pm.local_addr_used);
901 /* The following limits only make sense for the in-kernel PM */
902 if (mptcp_pm_is_kernel(msk)) {
903 info->mptcpi_subflows_max =
904 mptcp_pm_get_subflows_max(msk);
905 info->mptcpi_add_addr_signal_max =
906 mptcp_pm_get_add_addr_signal_max(msk);
907 info->mptcpi_add_addr_accepted_max =
908 mptcp_pm_get_add_addr_accept_max(msk);
909 info->mptcpi_local_addr_max =
910 mptcp_pm_get_local_addr_max(msk);
913 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags))
914 flags |= MPTCP_INFO_FLAG_FALLBACK;
915 if (READ_ONCE(msk->can_ack))
916 flags |= MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED;
917 info->mptcpi_flags = flags;
918 info->mptcpi_token = READ_ONCE(msk->token);
919 info->mptcpi_write_seq = READ_ONCE(msk->write_seq);
920 info->mptcpi_snd_una = READ_ONCE(msk->snd_una);
921 info->mptcpi_rcv_nxt = READ_ONCE(msk->ack_seq);
922 info->mptcpi_csum_enabled = READ_ONCE(msk->csum_enabled);
924 EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
926 static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
928 struct mptcp_info m_info;
931 if (get_user(len, optlen))
934 len = min_t(unsigned int, len, sizeof(struct mptcp_info));
936 mptcp_diag_fill_info(msk, &m_info);
938 if (put_user(len, optlen))
941 if (copy_to_user(optval, &m_info, len))
947 static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
952 u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
955 copied += sfd->size_subflow_data;
959 if (put_user(copied, optlen))
962 if (copy_to_user(optval, sfd, copylen))
968 static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
969 char __user *optval, int __user *optlen)
973 if (get_user(len, optlen))
976 /* if mptcp_subflow_data size is changed, need to adjust
977 * this function to deal with programs using old version.
979 BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
981 if (len < MIN_INFO_OPTLEN_SIZE)
984 memset(sfd, 0, sizeof(*sfd));
986 copylen = min_t(unsigned int, len, sizeof(*sfd));
987 if (copy_from_user(sfd, optval, copylen))
990 /* size_subflow_data is u32, but len is signed */
991 if (sfd->size_subflow_data > INT_MAX ||
992 sfd->size_user > INT_MAX)
995 if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
996 sfd->size_subflow_data > len)
999 if (sfd->num_subflows || sfd->size_kernel)
1002 return len - sfd->size_subflow_data;
1005 static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
1008 struct mptcp_subflow_context *subflow;
1009 struct sock *sk = (struct sock *)msk;
1010 unsigned int sfcount = 0, copied = 0;
1011 struct mptcp_subflow_data sfd;
1012 char __user *infoptr;
1015 len = mptcp_get_subflow_data(&sfd, optval, optlen);
1019 sfd.size_kernel = sizeof(struct tcp_info);
1020 sfd.size_user = min_t(unsigned int, sfd.size_user,
1021 sizeof(struct tcp_info));
1023 infoptr = optval + sfd.size_subflow_data;
1027 mptcp_for_each_subflow(msk, subflow) {
1028 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1032 if (len && len >= sfd.size_user) {
1033 struct tcp_info info;
1035 tcp_get_info(ssk, &info);
1037 if (copy_to_user(infoptr, &info, sfd.size_user)) {
1042 infoptr += sfd.size_user;
1043 copied += sfd.size_user;
1044 len -= sfd.size_user;
1050 sfd.num_subflows = sfcount;
1052 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1058 static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
1060 const struct inet_sock *inet = inet_sk(sk);
1062 memset(a, 0, sizeof(*a));
1064 if (sk->sk_family == AF_INET) {
1065 a->sin_local.sin_family = AF_INET;
1066 a->sin_local.sin_port = inet->inet_sport;
1067 a->sin_local.sin_addr.s_addr = inet->inet_rcv_saddr;
1069 if (!a->sin_local.sin_addr.s_addr)
1070 a->sin_local.sin_addr.s_addr = inet->inet_saddr;
1072 a->sin_remote.sin_family = AF_INET;
1073 a->sin_remote.sin_port = inet->inet_dport;
1074 a->sin_remote.sin_addr.s_addr = inet->inet_daddr;
1075 #if IS_ENABLED(CONFIG_IPV6)
1076 } else if (sk->sk_family == AF_INET6) {
1077 const struct ipv6_pinfo *np = inet6_sk(sk);
1079 if (WARN_ON_ONCE(!np))
1082 a->sin6_local.sin6_family = AF_INET6;
1083 a->sin6_local.sin6_port = inet->inet_sport;
1085 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
1086 a->sin6_local.sin6_addr = np->saddr;
1088 a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
1090 a->sin6_remote.sin6_family = AF_INET6;
1091 a->sin6_remote.sin6_port = inet->inet_dport;
1092 a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
1097 static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
1100 struct mptcp_subflow_context *subflow;
1101 struct sock *sk = (struct sock *)msk;
1102 unsigned int sfcount = 0, copied = 0;
1103 struct mptcp_subflow_data sfd;
1104 char __user *addrptr;
1107 len = mptcp_get_subflow_data(&sfd, optval, optlen);
1111 sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
1112 sfd.size_user = min_t(unsigned int, sfd.size_user,
1113 sizeof(struct mptcp_subflow_addrs));
1115 addrptr = optval + sfd.size_subflow_data;
1119 mptcp_for_each_subflow(msk, subflow) {
1120 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1124 if (len && len >= sfd.size_user) {
1125 struct mptcp_subflow_addrs a;
1127 mptcp_get_sub_addrs(ssk, &a);
1129 if (copy_to_user(addrptr, &a, sfd.size_user)) {
1134 addrptr += sfd.size_user;
1135 copied += sfd.size_user;
1136 len -= sfd.size_user;
1142 sfd.num_subflows = sfcount;
1144 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1150 static int mptcp_put_int_option(struct mptcp_sock *msk, char __user *optval,
1151 int __user *optlen, int val)
1155 if (get_user(len, optlen))
1160 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1161 unsigned char ucval = (unsigned char)val;
1164 if (put_user(len, optlen))
1166 if (copy_to_user(optval, &ucval, 1))
1169 len = min_t(unsigned int, len, sizeof(int));
1170 if (put_user(len, optlen))
1172 if (copy_to_user(optval, &val, len))
1179 static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
1180 char __user *optval, int __user *optlen)
1184 case TCP_CONGESTION:
1187 case TCP_DEFER_ACCEPT:
1189 case TCP_FASTOPEN_CONNECT:
1190 case TCP_FASTOPEN_KEY:
1191 case TCP_FASTOPEN_NO_COOKIE:
1192 return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
1195 return mptcp_put_int_option(msk, optval, optlen, msk->recvmsg_inq);
1197 return mptcp_put_int_option(msk, optval, optlen, msk->cork);
1199 return mptcp_put_int_option(msk, optval, optlen, msk->nodelay);
1204 static int mptcp_getsockopt_v4(struct mptcp_sock *msk, int optname,
1205 char __user *optval, int __user *optlen)
1207 struct sock *sk = (void *)msk;
1211 return mptcp_put_int_option(msk, optval, optlen, inet_sk(sk)->tos);
1217 static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
1218 char __user *optval, int __user *optlen)
1222 return mptcp_getsockopt_info(msk, optval, optlen);
1224 return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
1225 case MPTCP_SUBFLOW_ADDRS:
1226 return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
1232 int mptcp_getsockopt(struct sock *sk, int level, int optname,
1233 char __user *optval, int __user *option)
1235 struct mptcp_sock *msk = mptcp_sk(sk);
1238 pr_debug("msk=%p", msk);
1240 /* @@ the meaning of setsockopt() when the socket is connected and
1241 * there are multiple subflows is not yet defined. It is up to the
1242 * MPTCP-level socket to configure the subflows until the subflow
1243 * is in TCP fallback, when socket options are passed through
1244 * to the one remaining subflow.
1247 ssk = __mptcp_tcp_fallback(msk);
1250 return tcp_getsockopt(ssk, level, optname, optval, option);
1252 if (level == SOL_IP)
1253 return mptcp_getsockopt_v4(msk, optname, optval, option);
1254 if (level == SOL_TCP)
1255 return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
1256 if (level == SOL_MPTCP)
1257 return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
1261 static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
1263 static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
1264 struct sock *sk = (struct sock *)msk;
1266 if (ssk->sk_prot->keepalive) {
1267 if (sock_flag(sk, SOCK_KEEPOPEN))
1268 ssk->sk_prot->keepalive(ssk, 1);
1270 ssk->sk_prot->keepalive(ssk, 0);
1273 ssk->sk_priority = sk->sk_priority;
1274 ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1275 ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
1276 ssk->sk_ipv6only = sk->sk_ipv6only;
1277 __ip_sock_set_tos(ssk, inet_sk(sk)->tos);
1279 if (sk->sk_userlocks & tx_rx_locks) {
1280 ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1281 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
1282 WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1283 if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1284 WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1287 if (sock_flag(sk, SOCK_LINGER)) {
1288 ssk->sk_lingertime = sk->sk_lingertime;
1289 sock_set_flag(ssk, SOCK_LINGER);
1291 sock_reset_flag(ssk, SOCK_LINGER);
1294 if (sk->sk_mark != ssk->sk_mark) {
1295 ssk->sk_mark = sk->sk_mark;
1299 sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1301 if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
1302 tcp_set_congestion_control(ssk, msk->ca_name, false, true);
1303 __tcp_sock_set_cork(ssk, !!msk->cork);
1304 __tcp_sock_set_nodelay(ssk, !!msk->nodelay);
1306 inet_sk(ssk)->transparent = inet_sk(sk)->transparent;
1307 inet_sk(ssk)->freebind = inet_sk(sk)->freebind;
1310 static void __mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1312 bool slow = lock_sock_fast(ssk);
1314 sync_socket_options(msk, ssk);
1316 unlock_sock_fast(ssk, slow);
1319 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1321 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1323 msk_owned_by_me(msk);
1325 if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1326 __mptcp_sockopt_sync(msk, ssk);
1328 subflow->setsockopt_seq = msk->setsockopt_seq;
1332 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk)
1334 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1336 msk_owned_by_me(msk);
1338 if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1339 sync_socket_options(msk, ssk);
1341 subflow->setsockopt_seq = msk->setsockopt_seq;