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 #include "mptcp_pm_gen.h"
18 #define MPTCP_SUPPORTED_VERSION 1
20 /* MPTCP option bits */
21 #define OPTION_MPTCP_MPC_SYN BIT(0)
22 #define OPTION_MPTCP_MPC_SYNACK BIT(1)
23 #define OPTION_MPTCP_MPC_ACK BIT(2)
24 #define OPTION_MPTCP_MPJ_SYN BIT(3)
25 #define OPTION_MPTCP_MPJ_SYNACK BIT(4)
26 #define OPTION_MPTCP_MPJ_ACK BIT(5)
27 #define OPTION_MPTCP_ADD_ADDR BIT(6)
28 #define OPTION_MPTCP_RM_ADDR BIT(7)
29 #define OPTION_MPTCP_FASTCLOSE BIT(8)
30 #define OPTION_MPTCP_PRIO BIT(9)
31 #define OPTION_MPTCP_RST BIT(10)
32 #define OPTION_MPTCP_DSS BIT(11)
33 #define OPTION_MPTCP_FAIL BIT(12)
35 #define OPTION_MPTCP_CSUMREQD BIT(13)
37 #define OPTIONS_MPTCP_MPC (OPTION_MPTCP_MPC_SYN | OPTION_MPTCP_MPC_SYNACK | \
39 #define OPTIONS_MPTCP_MPJ (OPTION_MPTCP_MPJ_SYN | OPTION_MPTCP_MPJ_SYNACK | \
42 /* MPTCP option subtypes */
43 #define MPTCPOPT_MP_CAPABLE 0
44 #define MPTCPOPT_MP_JOIN 1
45 #define MPTCPOPT_DSS 2
46 #define MPTCPOPT_ADD_ADDR 3
47 #define MPTCPOPT_RM_ADDR 4
48 #define MPTCPOPT_MP_PRIO 5
49 #define MPTCPOPT_MP_FAIL 6
50 #define MPTCPOPT_MP_FASTCLOSE 7
51 #define MPTCPOPT_RST 8
53 /* MPTCP suboption lengths */
54 #define TCPOLEN_MPTCP_MPC_SYN 4
55 #define TCPOLEN_MPTCP_MPC_SYNACK 12
56 #define TCPOLEN_MPTCP_MPC_ACK 20
57 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22
58 #define TCPOLEN_MPTCP_MPJ_SYN 12
59 #define TCPOLEN_MPTCP_MPJ_SYNACK 16
60 #define TCPOLEN_MPTCP_MPJ_ACK 24
61 #define TCPOLEN_MPTCP_DSS_BASE 4
62 #define TCPOLEN_MPTCP_DSS_ACK32 4
63 #define TCPOLEN_MPTCP_DSS_ACK64 8
64 #define TCPOLEN_MPTCP_DSS_MAP32 10
65 #define TCPOLEN_MPTCP_DSS_MAP64 14
66 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2
67 #define TCPOLEN_MPTCP_ADD_ADDR 16
68 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18
69 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8
70 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10
71 #define TCPOLEN_MPTCP_ADD_ADDR6 28
72 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30
73 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20
74 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22
75 #define TCPOLEN_MPTCP_PORT_LEN 2
76 #define TCPOLEN_MPTCP_PORT_ALIGN 2
77 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3
78 #define TCPOLEN_MPTCP_PRIO 3
79 #define TCPOLEN_MPTCP_PRIO_ALIGN 4
80 #define TCPOLEN_MPTCP_FASTCLOSE 12
81 #define TCPOLEN_MPTCP_RST 4
82 #define TCPOLEN_MPTCP_FAIL 12
84 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA)
86 /* MPTCP MP_JOIN flags */
87 #define MPTCPOPT_BACKUP BIT(0)
88 #define MPTCPOPT_THMAC_LEN 8
90 /* MPTCP MP_CAPABLE flags */
91 #define MPTCP_VERSION_MASK (0x0F)
92 #define MPTCP_CAP_CHECKSUM_REQD BIT(7)
93 #define MPTCP_CAP_EXTENSIBILITY BIT(6)
94 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5)
95 #define MPTCP_CAP_HMAC_SHA256 BIT(0)
96 #define MPTCP_CAP_FLAG_MASK (0x1F)
99 #define MPTCP_DSS_DATA_FIN BIT(4)
100 #define MPTCP_DSS_DSN64 BIT(3)
101 #define MPTCP_DSS_HAS_MAP BIT(2)
102 #define MPTCP_DSS_ACK64 BIT(1)
103 #define MPTCP_DSS_HAS_ACK BIT(0)
104 #define MPTCP_DSS_FLAG_MASK (0x1F)
106 /* MPTCP ADD_ADDR flags */
107 #define MPTCP_ADDR_ECHO BIT(0)
109 /* MPTCP MP_PRIO flags */
110 #define MPTCP_PRIO_BKUP BIT(0)
112 /* MPTCP TCPRST flags */
113 #define MPTCP_RST_TRANSIENT BIT(0)
115 /* MPTCP socket atomic flags */
116 #define MPTCP_NOSPACE 1
117 #define MPTCP_WORK_RTX 2
118 #define MPTCP_FALLBACK_DONE 4
119 #define MPTCP_WORK_CLOSE_SUBFLOW 5
121 /* MPTCP socket release cb flags */
122 #define MPTCP_PUSH_PENDING 1
123 #define MPTCP_CLEAN_UNA 2
124 #define MPTCP_ERROR_REPORT 3
125 #define MPTCP_RETRANSMIT 4
126 #define MPTCP_FLUSH_JOIN_LIST 5
127 #define MPTCP_SYNC_STATE 6
128 #define MPTCP_SYNC_SNDBUF 7
130 struct mptcp_skb_cb {
137 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0]))
139 static inline bool before64(__u64 seq1, __u64 seq2)
141 return (__s64)(seq1 - seq2) < 0;
144 #define after64(seq2, seq1) before64(seq1, seq2)
146 struct mptcp_options_received {
171 u8 hmac[MPTCPOPT_HMAC_LEN];
172 struct mptcp_addr_info addr;
173 struct mptcp_rm_list rm_list;
178 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
180 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
181 ((nib & 0xF) << 8) | field);
184 enum mptcp_pm_status {
185 MPTCP_PM_ADD_ADDR_RECEIVED,
186 MPTCP_PM_ADD_ADDR_SEND_ACK,
187 MPTCP_PM_RM_ADDR_RECEIVED,
188 MPTCP_PM_ESTABLISHED,
189 MPTCP_PM_SUBFLOW_ESTABLISHED,
190 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
191 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
192 * accounted int id_avail_bitmap
197 MPTCP_PM_TYPE_KERNEL = 0,
198 MPTCP_PM_TYPE_USERSPACE,
201 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
204 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
205 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
207 enum mptcp_addr_signal_status {
208 MPTCP_ADD_ADDR_SIGNAL,
210 MPTCP_RM_ADDR_SIGNAL,
213 /* max value of mptcp_addr_info.id */
214 #define MPTCP_PM_MAX_ADDR_ID U8_MAX
216 struct mptcp_pm_data {
217 struct mptcp_addr_info local;
218 struct mptcp_addr_info remote;
219 struct list_head anno_list;
220 struct list_head userspace_pm_local_addr_list;
222 spinlock_t lock; /*protects the whole PM data */
229 bool remote_deny_join_id0;
230 u8 add_addr_signaled;
231 u8 add_addr_accepted;
236 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
237 struct mptcp_rm_list rm_list_tx;
238 struct mptcp_rm_list rm_list_rx;
241 struct mptcp_pm_addr_entry {
242 struct list_head list;
243 struct mptcp_addr_info addr;
249 struct mptcp_data_frag {
250 struct list_head list;
259 /* MPTCP connection sock */
261 /* inet_connection_sock must be the first member */
262 struct inet_connection_sock sk;
270 atomic64_t rcv_wnd_sent;
271 u64 rcv_data_fin_seq;
277 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
278 * recovery related fields are under data_lock
284 unsigned long timer_ival;
288 unsigned long cb_flags;
289 bool recovery; /* closing subflow write queue reinjected */
291 bool fully_established;
293 bool snd_data_fin_enable;
295 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
297 bool allow_infinite_fallback;
298 u8 pending_state; /* A subflow asked to set this sk_state,
299 * protected by the msk data lock
309 struct work_struct work;
310 struct sk_buff *ooo_last_skb;
311 struct rb_root out_of_order_queue;
312 struct sk_buff_head receive_queue;
313 struct list_head conn_list;
314 struct list_head rtx_queue;
315 struct mptcp_data_frag *first_pending;
316 struct list_head join_list;
317 struct sock *first; /* The mptcp ops can safely dereference, using suitable
318 * ONCE annotation, the subflow outside the socket
319 * lock as such sock is freed after close().
321 struct mptcp_pm_data pm;
322 struct mptcp_sched_ops *sched;
324 u32 space; /* bytes copied in last measurement window */
325 u32 copied; /* bytes copied in this measurement window */
326 u64 time; /* start time of measurement window */
327 u64 rtt_us; /* last maximum rtt of subflows */
333 char ca_name[TCP_CA_NAME_MAX];
336 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
337 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
339 #define mptcp_for_each_subflow(__msk, __subflow) \
340 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
341 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
342 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
344 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
346 sock_owned_by_me((const struct sock *)msk);
349 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
351 /* the msk socket don't use the backlog, also account for the bulk
354 static inline int __mptcp_rmem(const struct sock *sk)
356 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
359 static inline int mptcp_win_from_space(const struct sock *sk, int space)
361 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
364 static inline int __mptcp_space(const struct sock *sk)
366 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
369 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
371 const struct mptcp_sock *msk = mptcp_sk(sk);
373 return READ_ONCE(msk->first_pending);
376 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
378 struct mptcp_sock *msk = mptcp_sk(sk);
379 struct mptcp_data_frag *cur;
381 cur = msk->first_pending;
382 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
383 list_next_entry(cur, list);
386 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
388 const struct mptcp_sock *msk = mptcp_sk(sk);
390 if (!msk->first_pending)
393 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
396 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
399 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
401 struct mptcp_sock *msk = mptcp_sk(sk);
403 if (msk->snd_una == READ_ONCE(msk->snd_nxt))
406 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
409 struct csum_pseudo_header {
416 struct mptcp_subflow_request_sock {
417 struct tcp_request_sock sk;
432 struct mptcp_sock *msk;
433 struct hlist_nulls_node token_node;
436 static inline struct mptcp_subflow_request_sock *
437 mptcp_subflow_rsk(const struct request_sock *rsk)
439 return (struct mptcp_subflow_request_sock *)rsk;
442 struct mptcp_delegated_action {
443 struct napi_struct napi;
444 struct list_head head;
447 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
449 #define MPTCP_DELEGATE_SCHEDULED 0
450 #define MPTCP_DELEGATE_SEND 1
451 #define MPTCP_DELEGATE_ACK 2
452 #define MPTCP_DELEGATE_SNDBUF 3
454 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
455 /* MPTCP subflow context */
456 struct mptcp_subflow_context {
457 struct list_head node;/* conn_list of subflows */
461 unsigned long avg_pacing_rate; /* protected by msk socket lock */
472 __wsum map_data_csum;
474 u32 request_mptcp : 1, /* send MP_CAPABLE */
475 request_join : 1, /* send MP_JOIN */
477 mp_capable : 1, /* remote is MPTCP capable */
478 mp_join : 1, /* remote is JOINing */
479 fully_established : 1, /* path validated */
480 pm_notified : 1, /* PM hook called for established status */
490 send_infinite_map : 1,
491 remote_key_valid : 1, /* received the peer key from */
492 disposable : 1, /* ctx can be free at ulp release time */
493 stale : 1, /* unable to snd/rcv data, do not use for xmit */
494 valid_csum_seen : 1, /* at least one csum validated */
495 is_mptfo : 1, /* subflow is doing TFO */
504 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
505 u64 iasn; /* initial ack sequence number, MPC subflows only */
507 s16 local_id; /* if negative not initialized yet */
510 u8 reset_transient:1;
516 long delegated_status;
517 unsigned long fail_tout;
521 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
524 u32 stale_rcv_tstamp;
525 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
526 * protected by the msk socket lock
529 struct sock *tcp_sock; /* tcp sk backpointer */
530 struct sock *conn; /* parent mptcp_sock */
531 const struct inet_connection_sock_af_ops *icsk_af_ops;
532 void (*tcp_state_change)(struct sock *sk);
533 void (*tcp_error_report)(struct sock *sk);
538 static inline struct mptcp_subflow_context *
539 mptcp_subflow_ctx(const struct sock *sk)
541 struct inet_connection_sock *icsk = inet_csk(sk);
543 /* Use RCU on icsk_ulp_data only for sock diag code */
544 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
547 static inline struct sock *
548 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
550 return subflow->tcp_sock;
554 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
556 memset(&subflow->reset, 0, sizeof(subflow->reset));
557 subflow->request_mptcp = 1;
558 WRITE_ONCE(subflow->local_id, -1);
562 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
564 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
565 subflow->ssn_offset -
566 subflow->map_subflow_seq;
570 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
572 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
575 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
577 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
579 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
580 struct mptcp_delegated_action *delegated;
583 /* the caller held the subflow bh socket lock */
584 lockdep_assert_in_softirq();
586 /* The implied barrier pairs with tcp_release_cb_override()
587 * mptcp_napi_poll(), and ensures the below list check sees list
588 * updates done prior to delegated status bits changes
590 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
591 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
592 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
595 delegated = this_cpu_ptr(&mptcp_delegated_actions);
596 schedule = list_empty(&delegated->head);
597 list_add_tail(&subflow->delegated_node, &delegated->head);
598 sock_hold(mptcp_subflow_tcp_sock(subflow));
600 napi_schedule(&delegated->napi);
604 static inline struct mptcp_subflow_context *
605 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
607 struct mptcp_subflow_context *ret;
609 if (list_empty(&delegated->head))
612 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
613 list_del_init(&ret->delegated_node);
617 int mptcp_is_enabled(const struct net *net);
618 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
619 int mptcp_is_checksum_enabled(const struct net *net);
620 int mptcp_allow_join_id0(const struct net *net);
621 unsigned int mptcp_stale_loss_cnt(const struct net *net);
622 unsigned int mptcp_close_timeout(const struct sock *sk);
623 int mptcp_get_pm_type(const struct net *net);
624 const char *mptcp_get_scheduler(const struct net *net);
625 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
626 struct mptcp_subflow_context *subflow,
627 const struct mptcp_options_received *mp_opt);
628 bool __mptcp_retransmit_pending_data(struct sock *sk);
629 void mptcp_check_and_set_pending(struct sock *sk);
630 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
631 bool mptcp_subflow_data_available(struct sock *sk);
632 void __init mptcp_subflow_init(void);
633 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
634 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
635 struct mptcp_subflow_context *subflow);
636 void __mptcp_subflow_send_ack(struct sock *ssk);
637 void mptcp_subflow_reset(struct sock *ssk);
638 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
639 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
640 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
641 bool __mptcp_close(struct sock *sk, long timeout);
642 void mptcp_cancel_work(struct sock *sk);
643 void __mptcp_unaccepted_force_close(struct sock *sk);
644 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
645 void mptcp_set_state(struct sock *sk, int state);
647 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
648 const struct mptcp_addr_info *b, bool use_port);
649 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
651 /* called with sk socket lock held */
652 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
653 const struct mptcp_addr_info *remote);
654 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
655 struct socket **new_sock);
656 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
657 struct sockaddr_storage *addr,
658 unsigned short family);
659 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
660 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
661 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
662 void mptcp_sched_init(void);
663 int mptcp_init_sched(struct mptcp_sock *msk,
664 struct mptcp_sched_ops *sched);
665 void mptcp_release_sched(struct mptcp_sock *msk);
666 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
668 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
669 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
670 int mptcp_sched_get_send(struct mptcp_sock *msk);
671 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
673 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
675 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
678 static inline bool mptcp_epollin_ready(const struct sock *sk)
680 /* mptcp doesn't have to deal with small skbs in the receive queue,
681 * at it can always coalesce them
683 return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
684 (mem_cgroup_sockets_enabled && sk->sk_memcg &&
685 mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
686 READ_ONCE(tcp_memory_pressure);
689 int mptcp_set_rcvlowat(struct sock *sk, int val);
691 static inline bool __tcp_can_send(const struct sock *ssk)
693 /* only send if our side has not closed yet */
694 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
697 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
699 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
700 if (subflow->request_join && !subflow->fully_established)
703 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
706 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
708 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
710 void mptcp_subflow_drop_ctx(struct sock *ssk);
712 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
713 struct mptcp_subflow_context *ctx)
715 sk->sk_data_ready = sock_def_readable;
716 sk->sk_state_change = ctx->tcp_state_change;
717 sk->sk_write_space = sk_stream_write_space;
718 sk->sk_error_report = ctx->tcp_error_report;
720 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
723 void __init mptcp_proto_init(void);
724 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
725 int __init mptcp_proto_v6_init(void);
728 struct sock *mptcp_sk_clone_init(const struct sock *sk,
729 const struct mptcp_options_received *mp_opt,
731 struct request_sock *req);
732 void mptcp_get_options(const struct sk_buff *skb,
733 struct mptcp_options_received *mp_opt);
735 void mptcp_finish_connect(struct sock *sk);
736 void __mptcp_sync_state(struct sock *sk, int state);
737 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
739 static inline void mptcp_stop_tout_timer(struct sock *sk)
741 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
744 sk_stop_timer(sk, &sk->sk_timer);
745 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
748 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
750 /* avoid 0 timestamp, as that means no close timeout */
751 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
754 static inline void mptcp_start_tout_timer(struct sock *sk)
756 mptcp_set_close_tout(sk, tcp_jiffies32);
757 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
760 static inline bool mptcp_is_fully_established(struct sock *sk)
762 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
763 READ_ONCE(mptcp_sk(sk)->fully_established);
766 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
767 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
768 bool mptcp_finish_join(struct sock *sk);
769 bool mptcp_schedule_work(struct sock *sk);
770 int mptcp_setsockopt(struct sock *sk, int level, int optname,
771 sockptr_t optval, unsigned int optlen);
772 int mptcp_getsockopt(struct sock *sk, int level, int optname,
773 char __user *optval, int __user *option);
775 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
776 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
781 return __mptcp_expand_seq(old_seq, cur_seq);
783 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
784 void __mptcp_data_acked(struct sock *sk);
785 void __mptcp_error_report(struct sock *sk);
786 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
787 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
789 return READ_ONCE(msk->snd_data_fin_enable) &&
790 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
793 static inline void mptcp_write_space(struct sock *sk)
795 if (sk_stream_is_writeable(sk)) {
796 /* pairs with memory barrier in mptcp_poll */
798 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags))
799 sk_stream_write_space(sk);
803 static inline void __mptcp_sync_sndbuf(struct sock *sk)
805 struct mptcp_subflow_context *subflow;
806 int ssk_sndbuf, new_sndbuf;
808 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
811 new_sndbuf = sock_net(sk)->ipv4.sysctl_tcp_wmem[0];
812 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
813 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
815 subflow->cached_sndbuf = ssk_sndbuf;
816 new_sndbuf += ssk_sndbuf;
819 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
820 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
821 mptcp_write_space(sk);
824 /* The called held both the msk socket and the subflow socket locks,
827 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
829 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
831 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
832 __mptcp_sync_sndbuf(sk);
835 /* the caller held only the subflow socket lock, either in process or
836 * BH context. Additionally this can be called under the msk data lock,
837 * so we can't acquire such lock here: let the delegate action acquires
838 * the needed locks in suitable order.
840 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
842 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
844 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
848 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
852 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
854 #define MPTCP_TOKEN_MAX_RETRIES 4
856 void __init mptcp_token_init(void);
857 static inline void mptcp_token_init_request(struct request_sock *req)
859 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
862 int mptcp_token_new_request(struct request_sock *req);
863 void mptcp_token_destroy_request(struct request_sock *req);
864 int mptcp_token_new_connect(struct sock *ssk);
865 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
866 struct mptcp_sock *msk);
867 bool mptcp_token_exists(u32 token);
868 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
869 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
871 void mptcp_token_destroy(struct mptcp_sock *msk);
873 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
875 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
876 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
878 void __init mptcp_pm_init(void);
879 void mptcp_pm_data_init(struct mptcp_sock *msk);
880 void mptcp_pm_data_reset(struct mptcp_sock *msk);
881 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
882 struct mptcp_addr_info *addr);
883 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
885 struct mptcp_pm_addr_entry *entry);
886 bool mptcp_pm_addr_families_match(const struct sock *sk,
887 const struct mptcp_addr_info *loc,
888 const struct mptcp_addr_info *rem);
889 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
890 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
891 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
892 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
893 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
894 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
895 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
896 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
897 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
898 const struct mptcp_subflow_context *subflow);
899 void mptcp_pm_add_addr_received(const struct sock *ssk,
900 const struct mptcp_addr_info *addr);
901 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
902 const struct mptcp_addr_info *addr);
903 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
904 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
905 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
906 const struct mptcp_rm_list *rm_list);
907 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
908 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
909 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
910 struct mptcp_addr_info *addr,
911 struct mptcp_addr_info *rem,
913 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
914 const struct mptcp_addr_info *addr);
915 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
916 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
917 struct mptcp_pm_add_entry *
918 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
919 const struct mptcp_addr_info *addr, bool check_id);
920 struct mptcp_pm_add_entry *
921 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
922 const struct mptcp_addr_info *addr);
923 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
925 u8 *flags, int *ifindex);
926 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
927 u8 *flags, int *ifindex);
928 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
930 u8 *flags, int *ifindex);
931 int mptcp_pm_set_flags(struct net *net, struct nlattr *token,
932 struct mptcp_pm_addr_entry *loc,
933 struct mptcp_pm_addr_entry *rem, u8 bkup);
934 int mptcp_pm_nl_set_flags(struct net *net, struct mptcp_pm_addr_entry *addr, u8 bkup);
935 int mptcp_userspace_pm_set_flags(struct net *net, struct nlattr *token,
936 struct mptcp_pm_addr_entry *loc,
937 struct mptcp_pm_addr_entry *rem, u8 bkup);
938 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
939 const struct mptcp_addr_info *addr,
941 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
942 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
943 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
944 void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
945 struct list_head *rm_list);
947 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
949 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
950 const struct sock *ssk, gfp_t gfp);
951 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
952 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
953 void mptcp_event_pm_listener(const struct sock *ssk,
954 enum mptcp_event_type event);
955 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
957 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
958 const struct mptcp_options_received *mp_opt);
959 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
960 struct request_sock *req);
962 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
964 return READ_ONCE(msk->pm.addr_signal) &
965 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
968 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
970 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
973 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
975 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
978 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
980 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
983 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
985 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
988 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
990 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
993 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
995 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
997 if (family == AF_INET6)
998 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1000 len += MPTCPOPT_THMAC_LEN;
1001 /* account for 2 trailing 'nop' options */
1003 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1008 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1010 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1013 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1016 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1017 unsigned int opt_size, unsigned int remaining,
1018 struct mptcp_addr_info *addr, bool *echo,
1019 bool *drop_other_suboptions);
1020 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1021 struct mptcp_rm_list *rm_list);
1022 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1023 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1024 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1026 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1028 int local_id = READ_ONCE(subflow->local_id);
1035 void __init mptcp_pm_nl_init(void);
1036 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1037 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1038 const struct mptcp_rm_list *rm_list);
1039 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1040 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1041 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1042 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1044 /* called under PM lock */
1045 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1047 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1048 WRITE_ONCE(msk->pm.accept_subflow, true);
1051 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1053 spin_lock_bh(&msk->pm.lock);
1054 __mptcp_pm_close_subflow(msk);
1055 spin_unlock_bh(&msk->pm.lock);
1058 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1059 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1061 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1063 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1066 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1068 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1070 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1073 static inline bool mptcp_check_fallback(const struct sock *sk)
1075 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1076 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1078 return __mptcp_check_fallback(msk);
1081 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1083 if (__mptcp_check_fallback(msk)) {
1084 pr_debug("TCP fallback already done (msk=%p)", msk);
1087 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1090 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1092 struct sock *ssk = READ_ONCE(msk->first);
1094 return ssk && ((1 << inet_sk_state_load(ssk)) &
1095 (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1096 TCPF_SYN_RECV | TCPF_LISTEN));
1099 static inline void mptcp_do_fallback(struct sock *ssk)
1101 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1102 struct sock *sk = subflow->conn;
1103 struct mptcp_sock *msk;
1106 __mptcp_do_fallback(msk);
1107 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1108 gfp_t saved_allocation = ssk->sk_allocation;
1110 /* we are in a atomic (BH) scope, override ssk default for data
1113 ssk->sk_allocation = GFP_ATOMIC;
1114 ssk->sk_shutdown |= SEND_SHUTDOWN;
1115 tcp_shutdown(ssk, SEND_SHUTDOWN);
1116 ssk->sk_allocation = saved_allocation;
1120 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1122 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1124 struct mptcp_ext *mpext;
1126 mpext = skb ? mptcp_get_ext(skb) : NULL;
1127 if (mpext && mpext->infinite_map)
1133 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1135 return (subflow->request_mptcp || subflow->request_join);
1138 static inline bool subflow_simultaneous_connect(struct sock *sk)
1140 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1142 return (1 << sk->sk_state) &
1143 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1144 is_active_ssk(subflow) &&
1145 !subflow->conn_finished;
1148 #ifdef CONFIG_SYN_COOKIES
1149 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1150 struct sk_buff *skb);
1151 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1152 struct sk_buff *skb);
1153 void __init mptcp_join_cookie_init(void);
1156 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1157 struct sk_buff *skb) {}
1159 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1160 struct sk_buff *skb)
1165 static inline void mptcp_join_cookie_init(void) {}
1168 #endif /* __MPTCP_PROTOCOL_H */