1 /* SPDX-License-Identifier: GPL-2.0 */
4 * Copyright (c) 2017 - 2019, Intel Corporation.
7 #ifndef __MPTCP_PROTOCOL_H
8 #define __MPTCP_PROTOCOL_H
10 #include <linux/random.h>
12 #include <net/inet_connection_sock.h>
13 #include <uapi/linux/mptcp.h>
14 #include <net/genetlink.h>
16 #define MPTCP_SUPPORTED_VERSION 1
18 /* MPTCP option bits */
19 #define OPTION_MPTCP_MPC_SYN BIT(0)
20 #define OPTION_MPTCP_MPC_SYNACK BIT(1)
21 #define OPTION_MPTCP_MPC_ACK BIT(2)
22 #define OPTION_MPTCP_MPJ_SYN BIT(3)
23 #define OPTION_MPTCP_MPJ_SYNACK BIT(4)
24 #define OPTION_MPTCP_MPJ_ACK BIT(5)
25 #define OPTION_MPTCP_ADD_ADDR BIT(6)
26 #define OPTION_MPTCP_RM_ADDR BIT(7)
27 #define OPTION_MPTCP_FASTCLOSE BIT(8)
28 #define OPTION_MPTCP_PRIO BIT(9)
29 #define OPTION_MPTCP_RST BIT(10)
30 #define OPTION_MPTCP_DSS BIT(11)
31 #define OPTION_MPTCP_FAIL BIT(12)
33 #define OPTION_MPTCP_CSUMREQD BIT(13)
35 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
37 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
40 /* MPTCP option subtypes */
41 #define MPTCPOPT_MP_CAPABLE 0
42 #define MPTCPOPT_MP_JOIN 1
43 #define MPTCPOPT_DSS 2
44 #define MPTCPOPT_ADD_ADDR 3
45 #define MPTCPOPT_RM_ADDR 4
46 #define MPTCPOPT_MP_PRIO 5
47 #define MPTCPOPT_MP_FAIL 6
48 #define MPTCPOPT_MP_FASTCLOSE 7
49 #define MPTCPOPT_RST 8
51 /* MPTCP suboption lengths */
52 #define TCPOLEN_MPTCP_MPC_SYN 4
53 #define TCPOLEN_MPTCP_MPC_SYNACK 12
54 #define TCPOLEN_MPTCP_MPC_ACK 20
55 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22
56 #define TCPOLEN_MPTCP_MPJ_SYN 12
57 #define TCPOLEN_MPTCP_MPJ_SYNACK 16
58 #define TCPOLEN_MPTCP_MPJ_ACK 24
59 #define TCPOLEN_MPTCP_DSS_BASE 4
60 #define TCPOLEN_MPTCP_DSS_ACK32 4
61 #define TCPOLEN_MPTCP_DSS_ACK64 8
62 #define TCPOLEN_MPTCP_DSS_MAP32 10
63 #define TCPOLEN_MPTCP_DSS_MAP64 14
64 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2
65 #define TCPOLEN_MPTCP_ADD_ADDR 16
66 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18
67 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8
68 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10
69 #define TCPOLEN_MPTCP_ADD_ADDR6 28
70 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30
71 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20
72 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22
73 #define TCPOLEN_MPTCP_PORT_LEN 2
74 #define TCPOLEN_MPTCP_PORT_ALIGN 2
75 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3
76 #define TCPOLEN_MPTCP_PRIO 3
77 #define TCPOLEN_MPTCP_PRIO_ALIGN 4
78 #define TCPOLEN_MPTCP_FASTCLOSE 12
79 #define TCPOLEN_MPTCP_RST 4
80 #define TCPOLEN_MPTCP_FAIL 12
82 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
84 /* MPTCP MP_JOIN flags */
85 #define MPTCPOPT_BACKUP BIT(0)
86 #define MPTCPOPT_THMAC_LEN 8
88 /* MPTCP MP_CAPABLE flags */
89 #define MPTCP_VERSION_MASK (0x0F)
90 #define MPTCP_CAP_CHECKSUM_REQD BIT(7)
91 #define MPTCP_CAP_EXTENSIBILITY BIT(6)
92 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5)
93 #define MPTCP_CAP_HMAC_SHA256 BIT(0)
94 #define MPTCP_CAP_FLAG_MASK (0x1F)
97 #define MPTCP_DSS_DATA_FIN BIT(4)
98 #define MPTCP_DSS_DSN64 BIT(3)
99 #define MPTCP_DSS_HAS_MAP BIT(2)
100 #define MPTCP_DSS_ACK64 BIT(1)
101 #define MPTCP_DSS_HAS_ACK BIT(0)
102 #define MPTCP_DSS_FLAG_MASK (0x1F)
104 /* MPTCP ADD_ADDR flags */
105 #define MPTCP_ADDR_ECHO BIT(0)
107 /* MPTCP MP_PRIO flags */
108 #define MPTCP_PRIO_BKUP BIT(0)
110 /* MPTCP TCPRST flags */
111 #define MPTCP_RST_TRANSIENT BIT(0)
113 /* MPTCP socket atomic flags */
114 #define MPTCP_NOSPACE 1
115 #define MPTCP_WORK_RTX 2
116 #define MPTCP_FALLBACK_DONE 4
117 #define MPTCP_WORK_CLOSE_SUBFLOW 5
119 /* MPTCP socket release cb flags */
120 #define MPTCP_PUSH_PENDING 1
121 #define MPTCP_CLEAN_UNA 2
122 #define MPTCP_ERROR_REPORT 3
123 #define MPTCP_RETRANSMIT 4
124 #define MPTCP_FLUSH_JOIN_LIST 5
125 #define MPTCP_SYNC_STATE 6
126 #define MPTCP_SYNC_SNDBUF 7
128 struct mptcp_skb_cb {
135 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0]))
137 static inline bool before64(__u64 seq1, __u64 seq2)
139 return (__s64)(seq1 - seq2) < 0;
142 #define after64(seq2, seq1) before64(seq1, seq2)
144 struct mptcp_options_received {
169 u8 hmac[MPTCPOPT_HMAC_LEN];
170 struct mptcp_addr_info addr;
171 struct mptcp_rm_list rm_list;
176 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
178 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
179 ((nib & 0xF) << 8) | field);
182 enum mptcp_pm_status {
183 MPTCP_PM_ADD_ADDR_RECEIVED,
184 MPTCP_PM_ADD_ADDR_SEND_ACK,
185 MPTCP_PM_RM_ADDR_RECEIVED,
186 MPTCP_PM_ESTABLISHED,
187 MPTCP_PM_SUBFLOW_ESTABLISHED,
188 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
189 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
190 * accounted int id_avail_bitmap
195 MPTCP_PM_TYPE_KERNEL = 0,
196 MPTCP_PM_TYPE_USERSPACE,
199 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
202 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
203 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
205 enum mptcp_addr_signal_status {
206 MPTCP_ADD_ADDR_SIGNAL,
208 MPTCP_RM_ADDR_SIGNAL,
211 /* max value of mptcp_addr_info.id */
212 #define MPTCP_PM_MAX_ADDR_ID U8_MAX
214 struct mptcp_pm_data {
215 struct mptcp_addr_info local;
216 struct mptcp_addr_info remote;
217 struct list_head anno_list;
218 struct list_head userspace_pm_local_addr_list;
220 spinlock_t lock; /*protects the whole PM data */
227 bool remote_deny_join_id0;
228 u8 add_addr_signaled;
229 u8 add_addr_accepted;
234 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
235 struct mptcp_rm_list rm_list_tx;
236 struct mptcp_rm_list rm_list_rx;
239 struct mptcp_pm_addr_entry {
240 struct list_head list;
241 struct mptcp_addr_info addr;
247 struct mptcp_data_frag {
248 struct list_head list;
257 /* MPTCP connection sock */
259 /* inet_connection_sock must be the first member */
260 struct inet_connection_sock sk;
268 atomic64_t rcv_wnd_sent;
269 u64 rcv_data_fin_seq;
274 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
275 * recovery related fields are under data_lock
281 unsigned long timer_ival;
285 unsigned long cb_flags;
286 bool recovery; /* closing subflow write queue reinjected */
288 bool fully_established;
290 bool snd_data_fin_enable;
292 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
294 bool allow_infinite_fallback;
295 u8 pending_state; /* A subflow asked to set this sk_state,
296 * protected by the msk data lock
306 struct work_struct work;
307 struct sk_buff *ooo_last_skb;
308 struct rb_root out_of_order_queue;
309 struct sk_buff_head receive_queue;
310 struct list_head conn_list;
311 struct list_head rtx_queue;
312 struct mptcp_data_frag *first_pending;
313 struct list_head join_list;
314 struct sock *first; /* The mptcp ops can safely dereference, using suitable
315 * ONCE annotation, the subflow outside the socket
316 * lock as such sock is freed after close().
318 struct mptcp_pm_data pm;
319 struct mptcp_sched_ops *sched;
321 u32 space; /* bytes copied in last measurement window */
322 u32 copied; /* bytes copied in this measurement window */
323 u64 time; /* start time of measurement window */
324 u64 rtt_us; /* last maximum rtt of subflows */
330 char ca_name[TCP_CA_NAME_MAX];
333 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
334 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
336 #define mptcp_for_each_subflow(__msk, __subflow) \
337 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
338 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
339 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
341 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
343 sock_owned_by_me((const struct sock *)msk);
346 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
348 /* the msk socket don't use the backlog, also account for the bulk
351 static inline int __mptcp_rmem(const struct sock *sk)
353 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
356 static inline int mptcp_win_from_space(const struct sock *sk, int space)
358 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
361 static inline int __mptcp_space(const struct sock *sk)
363 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
366 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
368 const struct mptcp_sock *msk = mptcp_sk(sk);
370 return READ_ONCE(msk->first_pending);
373 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
375 struct mptcp_sock *msk = mptcp_sk(sk);
376 struct mptcp_data_frag *cur;
378 cur = msk->first_pending;
379 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
380 list_next_entry(cur, list);
383 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
385 const struct mptcp_sock *msk = mptcp_sk(sk);
387 if (!msk->first_pending)
390 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
393 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
396 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
398 struct mptcp_sock *msk = mptcp_sk(sk);
400 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
403 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
406 struct csum_pseudo_header {
413 struct mptcp_subflow_request_sock {
414 struct tcp_request_sock sk;
429 struct mptcp_sock *msk;
430 struct hlist_nulls_node token_node;
433 static inline struct mptcp_subflow_request_sock *
434 mptcp_subflow_rsk(const struct request_sock *rsk)
436 return (struct mptcp_subflow_request_sock *)rsk;
439 enum mptcp_data_avail {
440 MPTCP_SUBFLOW_NODATA,
441 MPTCP_SUBFLOW_DATA_AVAIL,
444 struct mptcp_delegated_action {
445 struct napi_struct napi;
446 struct list_head head;
449 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
451 #define MPTCP_DELEGATE_SCHEDULED 0
452 #define MPTCP_DELEGATE_SEND 1
453 #define MPTCP_DELEGATE_ACK 2
454 #define MPTCP_DELEGATE_SNDBUF 3
456 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
457 /* MPTCP subflow context */
458 struct mptcp_subflow_context {
459 struct list_head node;/* conn_list of subflows */
463 unsigned long avg_pacing_rate; /* protected by msk socket lock */
474 __wsum map_data_csum;
476 u32 request_mptcp : 1, /* send MP_CAPABLE */
477 request_join : 1, /* send MP_JOIN */
479 mp_capable : 1, /* remote is MPTCP capable */
480 mp_join : 1, /* remote is JOINing */
481 fully_established : 1, /* path validated */
482 pm_notified : 1, /* PM hook called for established status */
492 send_infinite_map : 1,
493 remote_key_valid : 1, /* received the peer key from */
494 disposable : 1, /* ctx can be free at ulp release time */
495 stale : 1, /* unable to snd/rcv data, do not use for xmit */
496 valid_csum_seen : 1, /* at least one csum validated */
497 is_mptfo : 1, /* subflow is doing TFO */
499 enum mptcp_data_avail data_avail;
506 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
507 u64 iasn; /* initial ack sequence number, MPC subflows only */
509 s16 local_id; /* if negative not initialized yet */
512 u8 reset_transient:1;
518 long delegated_status;
519 unsigned long fail_tout;
523 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
526 u32 stale_rcv_tstamp;
527 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
528 * protected by the msk socket lock
531 struct sock *tcp_sock; /* tcp sk backpointer */
532 struct sock *conn; /* parent mptcp_sock */
533 const struct inet_connection_sock_af_ops *icsk_af_ops;
534 void (*tcp_state_change)(struct sock *sk);
535 void (*tcp_error_report)(struct sock *sk);
540 static inline struct mptcp_subflow_context *
541 mptcp_subflow_ctx(const struct sock *sk)
543 struct inet_connection_sock *icsk = inet_csk(sk);
545 /* Use RCU on icsk_ulp_data only for sock diag code */
546 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
549 static inline struct sock *
550 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
552 return subflow->tcp_sock;
556 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
558 memset(&subflow->reset, 0, sizeof(subflow->reset));
559 subflow->request_mptcp = 1;
560 WRITE_ONCE(subflow->local_id, -1);
564 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
566 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
567 subflow->ssn_offset -
568 subflow->map_subflow_seq;
572 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
574 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
577 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
579 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
581 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
582 struct mptcp_delegated_action *delegated;
585 /* the caller held the subflow bh socket lock */
586 lockdep_assert_in_softirq();
588 /* The implied barrier pairs with tcp_release_cb_override()
589 * mptcp_napi_poll(), and ensures the below list check sees list
590 * updates done prior to delegated status bits changes
592 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
593 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
594 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
597 delegated = this_cpu_ptr(&mptcp_delegated_actions);
598 schedule = list_empty(&delegated->head);
599 list_add_tail(&subflow->delegated_node, &delegated->head);
600 sock_hold(mptcp_subflow_tcp_sock(subflow));
602 napi_schedule(&delegated->napi);
606 static inline struct mptcp_subflow_context *
607 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
609 struct mptcp_subflow_context *ret;
611 if (list_empty(&delegated->head))
614 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
615 list_del_init(&ret->delegated_node);
619 int mptcp_is_enabled(const struct net *net);
620 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
621 int mptcp_is_checksum_enabled(const struct net *net);
622 int mptcp_allow_join_id0(const struct net *net);
623 unsigned int mptcp_stale_loss_cnt(const struct net *net);
624 int mptcp_get_pm_type(const struct net *net);
625 const char *mptcp_get_scheduler(const struct net *net);
626 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
627 struct mptcp_subflow_context *subflow,
628 const struct mptcp_options_received *mp_opt);
629 bool __mptcp_retransmit_pending_data(struct sock *sk);
630 void mptcp_check_and_set_pending(struct sock *sk);
631 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
632 bool mptcp_subflow_data_available(struct sock *sk);
633 void __init mptcp_subflow_init(void);
634 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
635 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
636 struct mptcp_subflow_context *subflow);
637 void __mptcp_subflow_send_ack(struct sock *ssk);
638 void mptcp_subflow_reset(struct sock *ssk);
639 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
640 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
641 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
642 bool __mptcp_close(struct sock *sk, long timeout);
643 void mptcp_cancel_work(struct sock *sk);
644 void __mptcp_unaccepted_force_close(struct sock *sk);
645 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
646 void mptcp_set_state(struct sock *sk, int state);
648 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
649 const struct mptcp_addr_info *b, bool use_port);
650 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
652 /* called with sk socket lock held */
653 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
654 const struct mptcp_addr_info *remote);
655 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
656 struct socket **new_sock);
657 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
658 struct sockaddr_storage *addr,
659 unsigned short family);
660 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
661 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
662 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
663 void mptcp_sched_init(void);
664 int mptcp_init_sched(struct mptcp_sock *msk,
665 struct mptcp_sched_ops *sched);
666 void mptcp_release_sched(struct mptcp_sock *msk);
667 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
669 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
670 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
671 int mptcp_sched_get_send(struct mptcp_sock *msk);
672 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
674 static inline bool __tcp_can_send(const struct sock *ssk)
676 /* only send if our side has not closed yet */
677 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
680 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
682 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
683 if (subflow->request_join && !subflow->fully_established)
686 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
689 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
691 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
693 void mptcp_subflow_drop_ctx(struct sock *ssk);
695 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
696 struct mptcp_subflow_context *ctx)
698 sk->sk_data_ready = sock_def_readable;
699 sk->sk_state_change = ctx->tcp_state_change;
700 sk->sk_write_space = sk_stream_write_space;
701 sk->sk_error_report = ctx->tcp_error_report;
703 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
706 void __init mptcp_proto_init(void);
707 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
708 int __init mptcp_proto_v6_init(void);
711 struct sock *mptcp_sk_clone_init(const struct sock *sk,
712 const struct mptcp_options_received *mp_opt,
714 struct request_sock *req);
715 void mptcp_get_options(const struct sk_buff *skb,
716 struct mptcp_options_received *mp_opt);
718 void mptcp_finish_connect(struct sock *sk);
719 void __mptcp_sync_state(struct sock *sk, int state);
720 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
722 static inline void mptcp_stop_tout_timer(struct sock *sk)
724 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
727 sk_stop_timer(sk, &sk->sk_timer);
728 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
731 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
733 /* avoid 0 timestamp, as that means no close timeout */
734 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
737 static inline void mptcp_start_tout_timer(struct sock *sk)
739 mptcp_set_close_tout(sk, tcp_jiffies32);
740 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
743 static inline bool mptcp_is_fully_established(struct sock *sk)
745 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
746 READ_ONCE(mptcp_sk(sk)->fully_established);
748 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
749 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
750 bool mptcp_finish_join(struct sock *sk);
751 bool mptcp_schedule_work(struct sock *sk);
752 int mptcp_setsockopt(struct sock *sk, int level, int optname,
753 sockptr_t optval, unsigned int optlen);
754 int mptcp_getsockopt(struct sock *sk, int level, int optname,
755 char __user *optval, int __user *option);
757 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
758 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
763 return __mptcp_expand_seq(old_seq, cur_seq);
765 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
766 void __mptcp_data_acked(struct sock *sk);
767 void __mptcp_error_report(struct sock *sk);
768 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
769 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
771 return READ_ONCE(msk->snd_data_fin_enable) &&
772 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
775 static inline void mptcp_write_space(struct sock *sk)
777 if (sk_stream_is_writeable(sk)) {
778 /* pairs with memory barrier in mptcp_poll */
780 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
781 sk_stream_write_space(sk);
785 static inline void __mptcp_sync_sndbuf(struct sock *sk)
787 struct mptcp_subflow_context *subflow;
788 int ssk_sndbuf, new_sndbuf;
790 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
793 new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
794 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
795 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
797 subflow->cached_sndbuf = ssk_sndbuf;
798 new_sndbuf += ssk_sndbuf;
801 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
802 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
803 mptcp_write_space(sk);
806 /* The called held both the msk socket and the subflow socket locks,
809 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
811 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
813 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
814 __mptcp_sync_sndbuf(sk);
817 /* the caller held only the subflow socket lock, either in process or
818 * BH context. Additionally this can be called under the msk data lock,
819 * so we can't acquire such lock here: let the delegate action acquires
820 * the needed locks in suitable order.
822 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
824 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
826 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
830 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
834 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
836 #define MPTCP_TOKEN_MAX_RETRIES 4
838 void __init mptcp_token_init(void);
839 static inline void mptcp_token_init_request(struct request_sock *req)
841 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
844 int mptcp_token_new_request(struct request_sock *req);
845 void mptcp_token_destroy_request(struct request_sock *req);
846 int mptcp_token_new_connect(struct sock *ssk);
847 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
848 struct mptcp_sock *msk);
849 bool mptcp_token_exists(u32 token);
850 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
851 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
853 void mptcp_token_destroy(struct mptcp_sock *msk);
855 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
857 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
858 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
860 void __init mptcp_pm_init(void);
861 void mptcp_pm_data_init(struct mptcp_sock *msk);
862 void mptcp_pm_data_reset(struct mptcp_sock *msk);
863 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
864 struct mptcp_addr_info *addr);
865 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
867 struct mptcp_pm_addr_entry *entry);
868 bool mptcp_pm_addr_families_match(const struct sock *sk,
869 const struct mptcp_addr_info *loc,
870 const struct mptcp_addr_info *rem);
871 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
872 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
873 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
874 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
875 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
876 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
877 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
878 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
879 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk, const struct sock *ssk,
880 const struct mptcp_subflow_context *subflow);
881 void mptcp_pm_add_addr_received(const struct sock *ssk,
882 const struct mptcp_addr_info *addr);
883 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
884 const struct mptcp_addr_info *addr);
885 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
886 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
887 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
888 const struct mptcp_rm_list *rm_list);
889 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
890 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
891 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
892 struct mptcp_addr_info *addr,
893 struct mptcp_addr_info *rem,
895 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
896 const struct mptcp_addr_info *addr);
897 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
898 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
899 struct mptcp_pm_add_entry *
900 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
901 const struct mptcp_addr_info *addr, bool check_id);
902 struct mptcp_pm_add_entry *
903 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
904 const struct mptcp_addr_info *addr);
905 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
907 u8 *flags, int *ifindex);
908 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
909 u8 *flags, int *ifindex);
910 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
912 u8 *flags, int *ifindex);
913 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
914 struct mptcp_pm_addr_entry *loc,
915 struct mptcp_pm_addr_entry *rem, u8 bkup);
916 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
917 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
918 struct mptcp_pm_addr_entry *loc,
919 struct mptcp_pm_addr_entry *rem, u8 bkup);
920 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
921 const struct mptcp_addr_info *addr,
923 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
924 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
925 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
926 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
927 struct list_head *rm_list);
929 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
930 int mptcp_nl_cmd_announce(struct sk_buff *skb, struct genl_info *info);
931 int mptcp_nl_cmd_remove(struct sk_buff *skb, struct genl_info *info);
932 int mptcp_nl_cmd_sf_create(struct sk_buff *skb, struct genl_info *info);
933 int mptcp_nl_cmd_sf_destroy(struct sk_buff *skb, struct genl_info *info);
935 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
936 const struct sock *ssk, gfp_t gfp);
937 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
938 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
939 void mptcp_event_pm_listener(const struct sock *ssk,
940 enum mptcp_event_type event);
941 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
943 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
944 const struct mptcp_options_received *mp_opt);
945 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
946 struct request_sock *req);
948 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
950 return READ_ONCE(msk->pm.addr_signal) &
951 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
954 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
956 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
959 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
961 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
964 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
966 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
969 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
971 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
974 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
976 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
979 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
981 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
983 if (family == AF_INET6)
984 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
986 len += MPTCPOPT_THMAC_LEN;
987 /* account for 2 trailing 'nop' options */
989 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
994 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
996 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
999 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1002 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1003 unsigned int opt_size, unsigned int remaining,
1004 struct mptcp_addr_info *addr, bool *echo,
1005 bool *drop_other_suboptions);
1006 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1007 struct mptcp_rm_list *rm_list);
1008 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1009 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1010 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1012 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1014 int local_id = READ_ONCE(subflow->local_id);
1021 void __init mptcp_pm_nl_init(void);
1022 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1023 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1024 const struct mptcp_rm_list *rm_list);
1025 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1026 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1027 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1028 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1030 /* called under PM lock */
1031 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1033 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1034 WRITE_ONCE(msk->pm.accept_subflow, true);
1037 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1039 spin_lock_bh(&msk->pm.lock);
1040 __mptcp_pm_close_subflow(msk);
1041 spin_unlock_bh(&msk->pm.lock);
1044 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1045 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1047 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1049 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1052 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1054 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1056 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1059 static inline bool mptcp_check_fallback(const struct sock *sk)
1061 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1062 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1064 return __mptcp_check_fallback(msk);
1067 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1069 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) {
1070 pr_debug("TCP fallback already done (msk=%p)", msk);
1073 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1076 static inline void mptcp_do_fallback(struct sock *ssk)
1078 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1079 struct sock *sk = subflow->conn;
1080 struct mptcp_sock *msk;
1083 __mptcp_do_fallback(msk);
1084 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1085 gfp_t saved_allocation = ssk->sk_allocation;
1087 /* we are in a atomic (BH) scope, override ssk default for data
1090 ssk->sk_allocation = GFP_ATOMIC;
1091 ssk->sk_shutdown |= SEND_SHUTDOWN;
1092 tcp_shutdown(ssk, SEND_SHUTDOWN);
1093 ssk->sk_allocation = saved_allocation;
1097 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1099 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1101 struct mptcp_ext *mpext;
1103 mpext = skb ? mptcp_get_ext(skb) : NULL;
1104 if (mpext && mpext->infinite_map)
1110 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1112 return (subflow->request_mptcp || subflow->request_join);
1115 static inline bool subflow_simultaneous_connect(struct sock *sk)
1117 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1119 return (1 << sk->sk_state) &
1120 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1121 is_active_ssk(subflow) &&
1122 !subflow->conn_finished;
1125 #ifdef CONFIG_SYN_COOKIES
1126 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1127 struct sk_buff *skb);
1128 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1129 struct sk_buff *skb);
1130 void __init mptcp_join_cookie_init(void);
1133 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1134 struct sk_buff *skb) {}
1136 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1137 struct sk_buff *skb)
1142 static inline void mptcp_join_cookie_init(void) {}
1145 #endif /* __MPTCP_PROTOCOL_H */