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;
393 case IPV6_TRANSPARENT:
396 ssock = __mptcp_nmpc_socket(msk);
402 ret = tcp_setsockopt(ssock->sk, SOL_IPV6, optname, optval, optlen);
408 sockopt_seq_inc(msk);
412 sk->sk_ipv6only = ssock->sk->sk_ipv6only;
414 case IPV6_TRANSPARENT:
415 inet_sk(sk)->transparent = inet_sk(ssock->sk)->transparent;
418 inet_sk(sk)->freebind = inet_sk(ssock->sk)->freebind;
429 static bool mptcp_supported_sockopt(int level, int optname)
431 if (level == SOL_IP) {
433 /* should work fine */
437 /* the following are control cmsg related */
444 case IP_RECVORIGDSTADDR:
446 case IP_RECVFRAGSIZE:
448 /* common stuff that need some love */
451 case IP_BIND_ADDRESS_NO_PORT:
452 case IP_MTU_DISCOVER:
455 /* possibly less common may deserve some love */
458 /* the following is apparently a no-op for plain TCP */
459 case IP_RECVERR_RFC4884:
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
473 /* IP_IPSEC_POLICY, IP_XFRM_POLICY are nut supported, unrelated here */
476 if (level == SOL_IPV6) {
480 /* the following are control cmsg related */
481 case IPV6_RECVPKTINFO:
482 case IPV6_2292PKTINFO:
483 case IPV6_RECVHOPLIMIT:
484 case IPV6_2292HOPLIMIT:
487 case IPV6_RECVHOPOPTS:
488 case IPV6_2292HOPOPTS:
489 case IPV6_RECVDSTOPTS:
490 case IPV6_2292DSTOPTS:
491 case IPV6_RECVTCLASS:
493 case IPV6_RECVPATHMTU:
494 case IPV6_RECVORIGDSTADDR:
495 case IPV6_RECVFRAGSIZE:
497 /* the following ones need some love but are quite common */
499 case IPV6_TRANSPARENT:
502 case IPV6_2292PKTOPTIONS:
503 case IPV6_UNICAST_HOPS:
504 case IPV6_MTU_DISCOVER:
507 case IPV6_FLOWINFO_SEND:
508 case IPV6_FLOWLABEL_MGR:
509 case IPV6_MINHOPCOUNT:
511 case IPV6_AUTOFLOWLABEL:
513 /* the following one is a no-op for plain TCP */
514 case IPV6_RECVERR_RFC4884:
518 /* IPV6_HOPOPTS, IPV6_RTHDRDSTOPTS, IPV6_RTHDR, IPV6_DSTOPTS are
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
529 /* IPV6_ROUTER_ALERT, IPV6_ROUTER_ALERT_ISOLATE are not supported, since are evil */
531 /* IPV6_IPSEC_POLICY, IPV6_XFRM_POLICY are not supported */
532 /* IPV6_ADDR_PREFERENCES is not supported, we must be careful with subflows */
535 if (level == SOL_TCP) {
537 /* the following are no-op or should work just fine */
538 case TCP_THIN_DUPACK:
539 case TCP_DEFER_ACCEPT:
541 /* the following need some love */
544 case TCP_THIN_LINEAR_TIMEOUTS:
553 case TCP_WINDOW_CLAMP:
555 case TCP_USER_TIMEOUT:
557 case TCP_NOTSENT_LOWAT:
563 /* TCP_MD5SIG, TCP_MD5SIG_EXT are not supported, MD5 is not compatible with MPTCP */
565 /* TCP_REPAIR, TCP_REPAIR_QUEUE, TCP_QUEUE_SEQ, TCP_REPAIR_OPTIONS,
566 * TCP_REPAIR_WINDOW are not supported, better avoid this mess
568 /* TCP_FASTOPEN_KEY, TCP_FASTOPEN TCP_FASTOPEN_CONNECT, TCP_FASTOPEN_NO_COOKIE,
569 * are not supported fastopen is currently unsupported
575 static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t optval,
578 struct mptcp_subflow_context *subflow;
579 struct sock *sk = (struct sock *)msk;
580 char name[TCP_CA_NAME_MAX];
587 ret = strncpy_from_sockptr(name, optval,
588 min_t(long, TCP_CA_NAME_MAX - 1, optlen));
594 cap_net_admin = ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN);
598 sockopt_seq_inc(msk);
599 mptcp_for_each_subflow(msk, subflow) {
600 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
604 err = tcp_set_congestion_control(ssk, name, true, cap_net_admin);
605 if (err < 0 && ret == 0)
607 subflow->setsockopt_seq = msk->setsockopt_seq;
612 strcpy(msk->ca_name, name);
618 static int mptcp_setsockopt_sol_tcp_cork(struct mptcp_sock *msk, sockptr_t optval,
621 struct mptcp_subflow_context *subflow;
622 struct sock *sk = (struct sock *)msk;
625 if (optlen < sizeof(int))
628 if (copy_from_sockptr(&val, optval, sizeof(val)))
632 sockopt_seq_inc(msk);
634 mptcp_for_each_subflow(msk, subflow) {
635 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
638 __tcp_sock_set_cork(ssk, !!val);
642 mptcp_check_and_set_pending(sk);
648 static int mptcp_setsockopt_sol_tcp_nodelay(struct mptcp_sock *msk, sockptr_t optval,
651 struct mptcp_subflow_context *subflow;
652 struct sock *sk = (struct sock *)msk;
655 if (optlen < sizeof(int))
658 if (copy_from_sockptr(&val, optval, sizeof(val)))
662 sockopt_seq_inc(msk);
663 msk->nodelay = !!val;
664 mptcp_for_each_subflow(msk, subflow) {
665 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
668 __tcp_sock_set_nodelay(ssk, !!val);
672 mptcp_check_and_set_pending(sk);
678 static int mptcp_setsockopt_sol_ip_set_transparent(struct mptcp_sock *msk, int optname,
679 sockptr_t optval, unsigned int optlen)
681 struct sock *sk = (struct sock *)msk;
682 struct inet_sock *issk;
683 struct socket *ssock;
686 err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
692 ssock = __mptcp_nmpc_socket(msk);
698 issk = inet_sk(ssock->sk);
702 issk->freebind = inet_sk(sk)->freebind;
705 issk->transparent = inet_sk(sk)->transparent;
713 sockopt_seq_inc(msk);
718 static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,
719 sockptr_t optval, unsigned int optlen)
721 struct mptcp_subflow_context *subflow;
722 struct sock *sk = (struct sock *)msk;
725 err = ip_setsockopt(sk, SOL_IP, optname, optval, optlen);
731 sockopt_seq_inc(msk);
732 val = inet_sk(sk)->tos;
733 mptcp_for_each_subflow(msk, subflow) {
734 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
736 __ip_sock_set_tos(ssk, val);
743 static int mptcp_setsockopt_v4(struct mptcp_sock *msk, int optname,
744 sockptr_t optval, unsigned int optlen)
749 return mptcp_setsockopt_sol_ip_set_transparent(msk, optname, optval, optlen);
751 return mptcp_setsockopt_v4_set_tos(msk, optname, optval, optlen);
757 static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
758 sockptr_t optval, unsigned int optlen)
760 struct sock *sk = (void *)msk;
765 ret = mptcp_get_int_option(msk, optval, optlen, &val);
768 if (val < 0 || val > 1)
772 msk->recvmsg_inq = !!val;
778 return mptcp_setsockopt_sol_tcp_congestion(msk, optval, optlen);
780 return mptcp_setsockopt_sol_tcp_cork(msk, optval, optlen);
782 return mptcp_setsockopt_sol_tcp_nodelay(msk, optval, optlen);
788 int mptcp_setsockopt(struct sock *sk, int level, int optname,
789 sockptr_t optval, unsigned int optlen)
791 struct mptcp_sock *msk = mptcp_sk(sk);
794 pr_debug("msk=%p", msk);
796 if (level == SOL_SOCKET)
797 return mptcp_setsockopt_sol_socket(msk, optname, optval, optlen);
799 if (!mptcp_supported_sockopt(level, optname))
802 /* @@ the meaning of setsockopt() when the socket is connected and
803 * there are multiple subflows is not yet defined. It is up to the
804 * MPTCP-level socket to configure the subflows until the subflow
805 * is in TCP fallback, when TCP socket options are passed through
806 * to the one remaining subflow.
809 ssk = __mptcp_tcp_fallback(msk);
812 return tcp_setsockopt(ssk, level, optname, optval, optlen);
815 return mptcp_setsockopt_v4(msk, optname, optval, optlen);
817 if (level == SOL_IPV6)
818 return mptcp_setsockopt_v6(msk, optname, optval, optlen);
820 if (level == SOL_TCP)
821 return mptcp_setsockopt_sol_tcp(msk, optname, optval, optlen);
826 static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int optname,
827 char __user *optval, int __user *optlen)
829 struct sock *sk = (struct sock *)msk;
830 struct socket *ssock;
837 ret = tcp_getsockopt(ssk, level, optname, optval, optlen);
841 ssock = __mptcp_nmpc_socket(msk);
845 ret = tcp_getsockopt(ssock->sk, level, optname, optval, optlen);
852 void mptcp_diag_fill_info(struct mptcp_sock *msk, struct mptcp_info *info)
854 struct sock *sk = &msk->sk.icsk_inet.sk;
859 memset(info, 0, sizeof(*info));
861 slow = lock_sock_fast(sk);
863 info->mptcpi_subflows = READ_ONCE(msk->pm.subflows);
864 info->mptcpi_add_addr_signal = READ_ONCE(msk->pm.add_addr_signaled);
865 info->mptcpi_add_addr_accepted = READ_ONCE(msk->pm.add_addr_accepted);
866 info->mptcpi_local_addr_used = READ_ONCE(msk->pm.local_addr_used);
867 info->mptcpi_subflows_max = mptcp_pm_get_subflows_max(msk);
868 val = mptcp_pm_get_add_addr_signal_max(msk);
869 info->mptcpi_add_addr_signal_max = val;
870 val = mptcp_pm_get_add_addr_accept_max(msk);
871 info->mptcpi_add_addr_accepted_max = val;
872 info->mptcpi_local_addr_max = mptcp_pm_get_local_addr_max(msk);
873 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags))
874 flags |= MPTCP_INFO_FLAG_FALLBACK;
875 if (READ_ONCE(msk->can_ack))
876 flags |= MPTCP_INFO_FLAG_REMOTE_KEY_RECEIVED;
877 info->mptcpi_flags = flags;
878 info->mptcpi_token = READ_ONCE(msk->token);
879 info->mptcpi_write_seq = READ_ONCE(msk->write_seq);
880 info->mptcpi_snd_una = READ_ONCE(msk->snd_una);
881 info->mptcpi_rcv_nxt = READ_ONCE(msk->ack_seq);
882 info->mptcpi_csum_enabled = READ_ONCE(msk->csum_enabled);
884 unlock_sock_fast(sk, slow);
886 EXPORT_SYMBOL_GPL(mptcp_diag_fill_info);
888 static int mptcp_getsockopt_info(struct mptcp_sock *msk, char __user *optval, int __user *optlen)
890 struct mptcp_info m_info;
893 if (get_user(len, optlen))
896 len = min_t(unsigned int, len, sizeof(struct mptcp_info));
898 mptcp_diag_fill_info(msk, &m_info);
900 if (put_user(len, optlen))
903 if (copy_to_user(optval, &m_info, len))
909 static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd,
914 u32 copylen = min_t(u32, sfd->size_subflow_data, sizeof(*sfd));
917 copied += sfd->size_subflow_data;
921 if (put_user(copied, optlen))
924 if (copy_to_user(optval, sfd, copylen))
930 static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd,
931 char __user *optval, int __user *optlen)
935 if (get_user(len, optlen))
938 /* if mptcp_subflow_data size is changed, need to adjust
939 * this function to deal with programs using old version.
941 BUILD_BUG_ON(sizeof(*sfd) != MIN_INFO_OPTLEN_SIZE);
943 if (len < MIN_INFO_OPTLEN_SIZE)
946 memset(sfd, 0, sizeof(*sfd));
948 copylen = min_t(unsigned int, len, sizeof(*sfd));
949 if (copy_from_user(sfd, optval, copylen))
952 /* size_subflow_data is u32, but len is signed */
953 if (sfd->size_subflow_data > INT_MAX ||
954 sfd->size_user > INT_MAX)
957 if (sfd->size_subflow_data < MIN_INFO_OPTLEN_SIZE ||
958 sfd->size_subflow_data > len)
961 if (sfd->num_subflows || sfd->size_kernel)
964 return len - sfd->size_subflow_data;
967 static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,
970 struct mptcp_subflow_context *subflow;
971 struct sock *sk = &msk->sk.icsk_inet.sk;
972 unsigned int sfcount = 0, copied = 0;
973 struct mptcp_subflow_data sfd;
974 char __user *infoptr;
977 len = mptcp_get_subflow_data(&sfd, optval, optlen);
981 sfd.size_kernel = sizeof(struct tcp_info);
982 sfd.size_user = min_t(unsigned int, sfd.size_user,
983 sizeof(struct tcp_info));
985 infoptr = optval + sfd.size_subflow_data;
989 mptcp_for_each_subflow(msk, subflow) {
990 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
994 if (len && len >= sfd.size_user) {
995 struct tcp_info info;
997 tcp_get_info(ssk, &info);
999 if (copy_to_user(infoptr, &info, sfd.size_user)) {
1004 infoptr += sfd.size_user;
1005 copied += sfd.size_user;
1006 len -= sfd.size_user;
1012 sfd.num_subflows = sfcount;
1014 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1020 static void mptcp_get_sub_addrs(const struct sock *sk, struct mptcp_subflow_addrs *a)
1022 struct inet_sock *inet = inet_sk(sk);
1024 memset(a, 0, sizeof(*a));
1026 if (sk->sk_family == AF_INET) {
1027 a->sin_local.sin_family = AF_INET;
1028 a->sin_local.sin_port = inet->inet_sport;
1029 a->sin_local.sin_addr.s_addr = inet->inet_rcv_saddr;
1031 if (!a->sin_local.sin_addr.s_addr)
1032 a->sin_local.sin_addr.s_addr = inet->inet_saddr;
1034 a->sin_remote.sin_family = AF_INET;
1035 a->sin_remote.sin_port = inet->inet_dport;
1036 a->sin_remote.sin_addr.s_addr = inet->inet_daddr;
1037 #if IS_ENABLED(CONFIG_IPV6)
1038 } else if (sk->sk_family == AF_INET6) {
1039 const struct ipv6_pinfo *np = inet6_sk(sk);
1041 if (WARN_ON_ONCE(!np))
1044 a->sin6_local.sin6_family = AF_INET6;
1045 a->sin6_local.sin6_port = inet->inet_sport;
1047 if (ipv6_addr_any(&sk->sk_v6_rcv_saddr))
1048 a->sin6_local.sin6_addr = np->saddr;
1050 a->sin6_local.sin6_addr = sk->sk_v6_rcv_saddr;
1052 a->sin6_remote.sin6_family = AF_INET6;
1053 a->sin6_remote.sin6_port = inet->inet_dport;
1054 a->sin6_remote.sin6_addr = sk->sk_v6_daddr;
1059 static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *optval,
1062 struct sock *sk = &msk->sk.icsk_inet.sk;
1063 struct mptcp_subflow_context *subflow;
1064 unsigned int sfcount = 0, copied = 0;
1065 struct mptcp_subflow_data sfd;
1066 char __user *addrptr;
1069 len = mptcp_get_subflow_data(&sfd, optval, optlen);
1073 sfd.size_kernel = sizeof(struct mptcp_subflow_addrs);
1074 sfd.size_user = min_t(unsigned int, sfd.size_user,
1075 sizeof(struct mptcp_subflow_addrs));
1077 addrptr = optval + sfd.size_subflow_data;
1081 mptcp_for_each_subflow(msk, subflow) {
1082 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
1086 if (len && len >= sfd.size_user) {
1087 struct mptcp_subflow_addrs a;
1089 mptcp_get_sub_addrs(ssk, &a);
1091 if (copy_to_user(addrptr, &a, sfd.size_user)) {
1096 addrptr += sfd.size_user;
1097 copied += sfd.size_user;
1098 len -= sfd.size_user;
1104 sfd.num_subflows = sfcount;
1106 if (mptcp_put_subflow_data(&sfd, optval, copied, optlen))
1112 static int mptcp_put_int_option(struct mptcp_sock *msk, char __user *optval,
1113 int __user *optlen, int val)
1117 if (get_user(len, optlen))
1122 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) {
1123 unsigned char ucval = (unsigned char)val;
1126 if (put_user(len, optlen))
1128 if (copy_to_user(optval, &ucval, 1))
1131 len = min_t(unsigned int, len, sizeof(int));
1132 if (put_user(len, optlen))
1134 if (copy_to_user(optval, &val, len))
1141 static int mptcp_getsockopt_sol_tcp(struct mptcp_sock *msk, int optname,
1142 char __user *optval, int __user *optlen)
1146 case TCP_CONGESTION:
1149 return mptcp_getsockopt_first_sf_only(msk, SOL_TCP, optname,
1152 return mptcp_put_int_option(msk, optval, optlen, msk->recvmsg_inq);
1154 return mptcp_put_int_option(msk, optval, optlen, msk->cork);
1156 return mptcp_put_int_option(msk, optval, optlen, msk->nodelay);
1161 static int mptcp_getsockopt_v4(struct mptcp_sock *msk, int optname,
1162 char __user *optval, int __user *optlen)
1164 struct sock *sk = (void *)msk;
1168 return mptcp_put_int_option(msk, optval, optlen, inet_sk(sk)->tos);
1174 static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname,
1175 char __user *optval, int __user *optlen)
1179 return mptcp_getsockopt_info(msk, optval, optlen);
1181 return mptcp_getsockopt_tcpinfo(msk, optval, optlen);
1182 case MPTCP_SUBFLOW_ADDRS:
1183 return mptcp_getsockopt_subflow_addrs(msk, optval, optlen);
1189 int mptcp_getsockopt(struct sock *sk, int level, int optname,
1190 char __user *optval, int __user *option)
1192 struct mptcp_sock *msk = mptcp_sk(sk);
1195 pr_debug("msk=%p", msk);
1197 /* @@ the meaning of setsockopt() when the socket is connected and
1198 * there are multiple subflows is not yet defined. It is up to the
1199 * MPTCP-level socket to configure the subflows until the subflow
1200 * is in TCP fallback, when socket options are passed through
1201 * to the one remaining subflow.
1204 ssk = __mptcp_tcp_fallback(msk);
1207 return tcp_getsockopt(ssk, level, optname, optval, option);
1209 if (level == SOL_IP)
1210 return mptcp_getsockopt_v4(msk, optname, optval, option);
1211 if (level == SOL_TCP)
1212 return mptcp_getsockopt_sol_tcp(msk, optname, optval, option);
1213 if (level == SOL_MPTCP)
1214 return mptcp_getsockopt_sol_mptcp(msk, optname, optval, option);
1218 static void sync_socket_options(struct mptcp_sock *msk, struct sock *ssk)
1220 static const unsigned int tx_rx_locks = SOCK_RCVBUF_LOCK | SOCK_SNDBUF_LOCK;
1221 struct sock *sk = (struct sock *)msk;
1223 if (ssk->sk_prot->keepalive) {
1224 if (sock_flag(sk, SOCK_KEEPOPEN))
1225 ssk->sk_prot->keepalive(ssk, 1);
1227 ssk->sk_prot->keepalive(ssk, 0);
1230 ssk->sk_priority = sk->sk_priority;
1231 ssk->sk_bound_dev_if = sk->sk_bound_dev_if;
1232 ssk->sk_incoming_cpu = sk->sk_incoming_cpu;
1233 __ip_sock_set_tos(ssk, inet_sk(sk)->tos);
1235 if (sk->sk_userlocks & tx_rx_locks) {
1236 ssk->sk_userlocks |= sk->sk_userlocks & tx_rx_locks;
1237 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
1238 WRITE_ONCE(ssk->sk_sndbuf, sk->sk_sndbuf);
1239 if (sk->sk_userlocks & SOCK_RCVBUF_LOCK)
1240 WRITE_ONCE(ssk->sk_rcvbuf, sk->sk_rcvbuf);
1243 if (sock_flag(sk, SOCK_LINGER)) {
1244 ssk->sk_lingertime = sk->sk_lingertime;
1245 sock_set_flag(ssk, SOCK_LINGER);
1247 sock_reset_flag(ssk, SOCK_LINGER);
1250 if (sk->sk_mark != ssk->sk_mark) {
1251 ssk->sk_mark = sk->sk_mark;
1255 sock_valbool_flag(ssk, SOCK_DBG, sock_flag(sk, SOCK_DBG));
1257 if (inet_csk(sk)->icsk_ca_ops != inet_csk(ssk)->icsk_ca_ops)
1258 tcp_set_congestion_control(ssk, msk->ca_name, false, true);
1259 __tcp_sock_set_cork(ssk, !!msk->cork);
1260 __tcp_sock_set_nodelay(ssk, !!msk->nodelay);
1262 inet_sk(ssk)->transparent = inet_sk(sk)->transparent;
1263 inet_sk(ssk)->freebind = inet_sk(sk)->freebind;
1266 static void __mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1268 bool slow = lock_sock_fast(ssk);
1270 sync_socket_options(msk, ssk);
1272 unlock_sock_fast(ssk, slow);
1275 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk)
1277 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1279 msk_owned_by_me(msk);
1281 if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1282 __mptcp_sockopt_sync(msk, ssk);
1284 subflow->setsockopt_seq = msk->setsockopt_seq;
1288 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk)
1290 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1292 msk_owned_by_me(msk);
1294 if (READ_ONCE(subflow->setsockopt_seq) != msk->setsockopt_seq) {
1295 sync_socket_options(msk, ssk);
1297 subflow->setsockopt_seq = msk->setsockopt_seq;