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_WORK_RTX 1
117 #define MPTCP_FALLBACK_DONE 2
118 #define MPTCP_WORK_CLOSE_SUBFLOW 3
120 /* MPTCP socket release cb flags */
121 #define MPTCP_PUSH_PENDING 1
122 #define MPTCP_CLEAN_UNA 2
123 #define MPTCP_ERROR_REPORT 3
124 #define MPTCP_RETRANSMIT 4
125 #define MPTCP_FLUSH_JOIN_LIST 5
126 #define MPTCP_SYNC_STATE 6
127 #define MPTCP_SYNC_SNDBUF 7
129 struct mptcp_skb_cb {
136 #define MPTCP_SKB_CB(__skb) ((struct mptcp_skb_cb *)&((__skb)->cb[0]))
138 static inline bool before64(__u64 seq1, __u64 seq2)
140 return (__s64)(seq1 - seq2) < 0;
143 #define after64(seq2, seq1) before64(seq1, seq2)
145 struct mptcp_options_received {
170 u8 hmac[MPTCPOPT_HMAC_LEN];
171 struct mptcp_addr_info addr;
172 struct mptcp_rm_list rm_list;
177 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field)
179 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) |
180 ((nib & 0xF) << 8) | field);
183 enum mptcp_pm_status {
184 MPTCP_PM_ADD_ADDR_RECEIVED,
185 MPTCP_PM_ADD_ADDR_SEND_ACK,
186 MPTCP_PM_RM_ADDR_RECEIVED,
187 MPTCP_PM_ESTABLISHED,
188 MPTCP_PM_SUBFLOW_ESTABLISHED,
189 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */
190 MPTCP_PM_MPC_ENDPOINT_ACCOUNTED /* persistent status, set after MPC local address is
191 * accounted int id_avail_bitmap
196 MPTCP_PM_TYPE_KERNEL = 0,
197 MPTCP_PM_TYPE_USERSPACE,
200 __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1,
203 /* Status bits below MPTCP_PM_ALREADY_ESTABLISHED need pm worker actions */
204 #define MPTCP_PM_WORK_MASK ((1 << MPTCP_PM_ALREADY_ESTABLISHED) - 1)
206 enum mptcp_addr_signal_status {
207 MPTCP_ADD_ADDR_SIGNAL,
209 MPTCP_RM_ADDR_SIGNAL,
212 /* max value of mptcp_addr_info.id */
213 #define MPTCP_PM_MAX_ADDR_ID U8_MAX
215 struct mptcp_pm_data {
216 struct mptcp_addr_info local;
217 struct mptcp_addr_info remote;
218 struct list_head anno_list;
219 struct list_head userspace_pm_local_addr_list;
221 spinlock_t lock; /*protects the whole PM data */
228 bool remote_deny_join_id0;
229 u8 add_addr_signaled;
230 u8 add_addr_accepted;
235 DECLARE_BITMAP(id_avail_bitmap, MPTCP_PM_MAX_ADDR_ID + 1);
236 struct mptcp_rm_list rm_list_tx;
237 struct mptcp_rm_list rm_list_rx;
240 struct mptcp_pm_addr_entry {
241 struct list_head list;
242 struct mptcp_addr_info addr;
248 struct mptcp_data_frag {
249 struct list_head list;
258 /* MPTCP connection sock */
260 /* inet_connection_sock must be the first member */
261 struct inet_connection_sock sk;
262 u64 local_key; /* protected by the first subflow socket lock
263 * lockless access read
265 u64 remote_key; /* same as above */
271 atomic64_t rcv_wnd_sent;
272 u64 rcv_data_fin_seq;
278 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq;
279 * recovery related fields are under data_lock
285 unsigned long timer_ival;
289 unsigned long cb_flags;
290 bool recovery; /* closing subflow write queue reinjected */
292 bool fully_established;
294 bool snd_data_fin_enable;
296 bool use_64bit_ack; /* Set when we received a 64-bit DSN */
298 bool allow_infinite_fallback;
299 u8 pending_state; /* A subflow asked to set this sk_state,
300 * protected by the msk data lock
311 struct work_struct work;
312 struct sk_buff *ooo_last_skb;
313 struct rb_root out_of_order_queue;
314 struct sk_buff_head receive_queue;
315 struct list_head conn_list;
316 struct list_head rtx_queue;
317 struct mptcp_data_frag *first_pending;
318 struct list_head join_list;
319 struct sock *first; /* The mptcp ops can safely dereference, using suitable
320 * ONCE annotation, the subflow outside the socket
321 * lock as such sock is freed after close().
323 struct mptcp_pm_data pm;
324 struct mptcp_sched_ops *sched;
326 u32 space; /* bytes copied in last measurement window */
327 u32 copied; /* bytes copied in this measurement window */
328 u64 time; /* start time of measurement window */
329 u64 rtt_us; /* last maximum rtt of subflows */
335 char ca_name[TCP_CA_NAME_MAX];
338 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock)
339 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock)
341 #define mptcp_for_each_subflow(__msk, __subflow) \
342 list_for_each_entry(__subflow, &((__msk)->conn_list), node)
343 #define mptcp_for_each_subflow_safe(__msk, __subflow, __tmp) \
344 list_for_each_entry_safe(__subflow, __tmp, &((__msk)->conn_list), node)
346 extern struct genl_family mptcp_genl_family;
348 static inline void msk_owned_by_me(const struct mptcp_sock *msk)
350 sock_owned_by_me((const struct sock *)msk);
353 #ifdef CONFIG_DEBUG_NET
354 /* MPTCP-specific: we might (indirectly) call this helper with the wrong sk */
356 #define tcp_sk(ptr) ({ \
357 typeof(ptr) _ptr = (ptr); \
358 WARN_ON(_ptr->sk_protocol != IPPROTO_TCP); \
359 container_of_const(_ptr, struct tcp_sock, inet_conn.icsk_inet.sk); \
361 #define mptcp_sk(ptr) ({ \
362 typeof(ptr) _ptr = (ptr); \
363 WARN_ON(_ptr->sk_protocol != IPPROTO_MPTCP); \
364 container_of_const(_ptr, struct mptcp_sock, sk.icsk_inet.sk); \
367 #else /* !CONFIG_DEBUG_NET */
368 #define mptcp_sk(ptr) container_of_const(ptr, struct mptcp_sock, sk.icsk_inet.sk)
371 /* the msk socket don't use the backlog, also account for the bulk
374 static inline int __mptcp_rmem(const struct sock *sk)
376 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released);
379 static inline int mptcp_win_from_space(const struct sock *sk, int space)
381 return __tcp_win_from_space(mptcp_sk(sk)->scaling_ratio, space);
384 static inline int __mptcp_space(const struct sock *sk)
386 return mptcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk));
389 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk)
391 const struct mptcp_sock *msk = mptcp_sk(sk);
393 return READ_ONCE(msk->first_pending);
396 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk)
398 struct mptcp_sock *msk = mptcp_sk(sk);
399 struct mptcp_data_frag *cur;
401 cur = msk->first_pending;
402 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL :
403 list_next_entry(cur, list);
406 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk)
408 const struct mptcp_sock *msk = mptcp_sk(sk);
410 if (!msk->first_pending)
413 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue)))
416 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list);
419 static inline struct mptcp_data_frag *mptcp_rtx_head(struct sock *sk)
421 struct mptcp_sock *msk = mptcp_sk(sk);
423 if (msk->snd_una == msk->snd_nxt)
426 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list);
429 struct csum_pseudo_header {
436 struct mptcp_subflow_request_sock {
437 struct tcp_request_sock sk;
452 struct mptcp_sock *msk;
453 struct hlist_nulls_node token_node;
456 static inline struct mptcp_subflow_request_sock *
457 mptcp_subflow_rsk(const struct request_sock *rsk)
459 return (struct mptcp_subflow_request_sock *)rsk;
462 struct mptcp_delegated_action {
463 struct napi_struct napi;
464 struct list_head head;
467 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions);
469 #define MPTCP_DELEGATE_SCHEDULED 0
470 #define MPTCP_DELEGATE_SEND 1
471 #define MPTCP_DELEGATE_ACK 2
472 #define MPTCP_DELEGATE_SNDBUF 3
474 #define MPTCP_DELEGATE_ACTIONS_MASK (~BIT(MPTCP_DELEGATE_SCHEDULED))
475 /* MPTCP subflow context */
476 struct mptcp_subflow_context {
477 struct list_head node;/* conn_list of subflows */
481 unsigned long avg_pacing_rate; /* protected by msk socket lock */
492 __wsum map_data_csum;
494 u32 request_mptcp : 1, /* send MP_CAPABLE */
495 request_join : 1, /* send MP_JOIN */
497 mp_capable : 1, /* remote is MPTCP capable */
498 mp_join : 1, /* remote is JOINing */
499 fully_established : 1, /* path validated */
500 pm_notified : 1, /* PM hook called for established status */
510 send_infinite_map : 1,
511 remote_key_valid : 1, /* received the peer key from */
512 disposable : 1, /* ctx can be free at ulp release time */
513 stale : 1, /* unable to snd/rcv data, do not use for xmit */
514 valid_csum_seen : 1, /* at least one csum validated */
515 is_mptfo : 1, /* subflow is doing TFO */
524 u8 hmac[MPTCPOPT_HMAC_LEN]; /* MPJ subflow only */
525 u64 iasn; /* initial ack sequence number, MPC subflows only */
527 s16 local_id; /* if negative not initialized yet */
530 u8 reset_transient:1;
536 long delegated_status;
537 unsigned long fail_tout;
541 struct list_head delegated_node; /* link into delegated_action, protected by local BH */
544 u32 stale_rcv_tstamp;
545 int cached_sndbuf; /* sndbuf size when last synced with the msk sndbuf,
546 * protected by the msk socket lock
549 struct sock *tcp_sock; /* tcp sk backpointer */
550 struct sock *conn; /* parent mptcp_sock */
551 const struct inet_connection_sock_af_ops *icsk_af_ops;
552 void (*tcp_state_change)(struct sock *sk);
553 void (*tcp_error_report)(struct sock *sk);
558 static inline struct mptcp_subflow_context *
559 mptcp_subflow_ctx(const struct sock *sk)
561 struct inet_connection_sock *icsk = inet_csk(sk);
563 /* Use RCU on icsk_ulp_data only for sock diag code */
564 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data;
567 static inline struct sock *
568 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow)
570 return subflow->tcp_sock;
574 mptcp_subflow_ctx_reset(struct mptcp_subflow_context *subflow)
576 memset(&subflow->reset, 0, sizeof(subflow->reset));
577 subflow->request_mptcp = 1;
578 WRITE_ONCE(subflow->local_id, -1);
582 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow)
584 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq -
585 subflow->ssn_offset -
586 subflow->map_subflow_seq;
590 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow)
592 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow);
595 void mptcp_subflow_process_delegated(struct sock *ssk, long actions);
597 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow, int action)
599 long old, set_bits = BIT(MPTCP_DELEGATE_SCHEDULED) | BIT(action);
600 struct mptcp_delegated_action *delegated;
603 /* the caller held the subflow bh socket lock */
604 lockdep_assert_in_softirq();
606 /* The implied barrier pairs with tcp_release_cb_override()
607 * mptcp_napi_poll(), and ensures the below list check sees list
608 * updates done prior to delegated status bits changes
610 old = set_mask_bits(&subflow->delegated_status, 0, set_bits);
611 if (!(old & BIT(MPTCP_DELEGATE_SCHEDULED))) {
612 if (WARN_ON_ONCE(!list_empty(&subflow->delegated_node)))
615 delegated = this_cpu_ptr(&mptcp_delegated_actions);
616 schedule = list_empty(&delegated->head);
617 list_add_tail(&subflow->delegated_node, &delegated->head);
618 sock_hold(mptcp_subflow_tcp_sock(subflow));
620 napi_schedule(&delegated->napi);
624 static inline struct mptcp_subflow_context *
625 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated)
627 struct mptcp_subflow_context *ret;
629 if (list_empty(&delegated->head))
632 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node);
633 list_del_init(&ret->delegated_node);
637 int mptcp_is_enabled(const struct net *net);
638 unsigned int mptcp_get_add_addr_timeout(const struct net *net);
639 int mptcp_is_checksum_enabled(const struct net *net);
640 int mptcp_allow_join_id0(const struct net *net);
641 unsigned int mptcp_stale_loss_cnt(const struct net *net);
642 unsigned int mptcp_close_timeout(const struct sock *sk);
643 int mptcp_get_pm_type(const struct net *net);
644 const char *mptcp_get_scheduler(const struct net *net);
645 void __mptcp_subflow_fully_established(struct mptcp_sock *msk,
646 struct mptcp_subflow_context *subflow,
647 const struct mptcp_options_received *mp_opt);
648 bool __mptcp_retransmit_pending_data(struct sock *sk);
649 void mptcp_check_and_set_pending(struct sock *sk);
650 void __mptcp_push_pending(struct sock *sk, unsigned int flags);
651 bool mptcp_subflow_data_available(struct sock *sk);
652 void __init mptcp_subflow_init(void);
653 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how);
654 void mptcp_close_ssk(struct sock *sk, struct sock *ssk,
655 struct mptcp_subflow_context *subflow);
656 void __mptcp_subflow_send_ack(struct sock *ssk);
657 void mptcp_subflow_reset(struct sock *ssk);
658 void mptcp_subflow_queue_clean(struct sock *sk, struct sock *ssk);
659 void mptcp_sock_graft(struct sock *sk, struct socket *parent);
660 struct sock *__mptcp_nmpc_sk(struct mptcp_sock *msk);
661 bool __mptcp_close(struct sock *sk, long timeout);
662 void mptcp_cancel_work(struct sock *sk);
663 void __mptcp_unaccepted_force_close(struct sock *sk);
664 void mptcp_set_owner_r(struct sk_buff *skb, struct sock *sk);
665 void mptcp_set_state(struct sock *sk, int state);
667 bool mptcp_addresses_equal(const struct mptcp_addr_info *a,
668 const struct mptcp_addr_info *b, bool use_port);
669 void mptcp_local_address(const struct sock_common *skc, struct mptcp_addr_info *addr);
671 /* called with sk socket lock held */
672 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc,
673 const struct mptcp_addr_info *remote);
674 int mptcp_subflow_create_socket(struct sock *sk, unsigned short family,
675 struct socket **new_sock);
676 void mptcp_info2sockaddr(const struct mptcp_addr_info *info,
677 struct sockaddr_storage *addr,
678 unsigned short family);
679 struct mptcp_sched_ops *mptcp_sched_find(const char *name);
680 int mptcp_register_scheduler(struct mptcp_sched_ops *sched);
681 void mptcp_unregister_scheduler(struct mptcp_sched_ops *sched);
682 void mptcp_sched_init(void);
683 int mptcp_init_sched(struct mptcp_sock *msk,
684 struct mptcp_sched_ops *sched);
685 void mptcp_release_sched(struct mptcp_sock *msk);
686 void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow,
688 struct sock *mptcp_subflow_get_send(struct mptcp_sock *msk);
689 struct sock *mptcp_subflow_get_retrans(struct mptcp_sock *msk);
690 int mptcp_sched_get_send(struct mptcp_sock *msk);
691 int mptcp_sched_get_retrans(struct mptcp_sock *msk);
693 static inline u64 mptcp_data_avail(const struct mptcp_sock *msk)
695 return READ_ONCE(msk->bytes_received) - READ_ONCE(msk->bytes_consumed);
698 static inline bool mptcp_epollin_ready(const struct sock *sk)
700 /* mptcp doesn't have to deal with small skbs in the receive queue,
701 * at it can always coalesce them
703 return (mptcp_data_avail(mptcp_sk(sk)) >= sk->sk_rcvlowat) ||
704 (mem_cgroup_sockets_enabled && sk->sk_memcg &&
705 mem_cgroup_under_socket_pressure(sk->sk_memcg)) ||
706 READ_ONCE(tcp_memory_pressure);
709 int mptcp_set_rcvlowat(struct sock *sk, int val);
711 static inline bool __tcp_can_send(const struct sock *ssk)
713 /* only send if our side has not closed yet */
714 return ((1 << inet_sk_state_load(ssk)) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT));
717 static inline bool __mptcp_subflow_active(struct mptcp_subflow_context *subflow)
719 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */
720 if (subflow->request_join && !subflow->fully_established)
723 return __tcp_can_send(mptcp_subflow_tcp_sock(subflow));
726 void mptcp_subflow_set_active(struct mptcp_subflow_context *subflow);
728 bool mptcp_subflow_active(struct mptcp_subflow_context *subflow);
730 void mptcp_subflow_drop_ctx(struct sock *ssk);
732 static inline void mptcp_subflow_tcp_fallback(struct sock *sk,
733 struct mptcp_subflow_context *ctx)
735 sk->sk_data_ready = sock_def_readable;
736 sk->sk_state_change = ctx->tcp_state_change;
737 sk->sk_write_space = sk_stream_write_space;
738 sk->sk_error_report = ctx->tcp_error_report;
740 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops;
743 void __init mptcp_proto_init(void);
744 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
745 int __init mptcp_proto_v6_init(void);
748 struct sock *mptcp_sk_clone_init(const struct sock *sk,
749 const struct mptcp_options_received *mp_opt,
751 struct request_sock *req);
752 void mptcp_get_options(const struct sk_buff *skb,
753 struct mptcp_options_received *mp_opt);
755 void mptcp_finish_connect(struct sock *sk);
756 void __mptcp_sync_state(struct sock *sk, int state);
757 void mptcp_reset_tout_timer(struct mptcp_sock *msk, unsigned long fail_tout);
759 static inline void mptcp_stop_tout_timer(struct sock *sk)
761 if (!inet_csk(sk)->icsk_mtup.probe_timestamp)
764 sk_stop_timer(sk, &sk->sk_timer);
765 inet_csk(sk)->icsk_mtup.probe_timestamp = 0;
768 static inline void mptcp_set_close_tout(struct sock *sk, unsigned long tout)
770 /* avoid 0 timestamp, as that means no close timeout */
771 inet_csk(sk)->icsk_mtup.probe_timestamp = tout ? : 1;
774 static inline void mptcp_start_tout_timer(struct sock *sk)
776 mptcp_set_close_tout(sk, tcp_jiffies32);
777 mptcp_reset_tout_timer(mptcp_sk(sk), 0);
780 static inline bool mptcp_is_fully_established(struct sock *sk)
782 return inet_sk_state_load(sk) == TCP_ESTABLISHED &&
783 READ_ONCE(mptcp_sk(sk)->fully_established);
786 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk);
787 void mptcp_data_ready(struct sock *sk, struct sock *ssk);
788 bool mptcp_finish_join(struct sock *sk);
789 bool mptcp_schedule_work(struct sock *sk);
790 int mptcp_setsockopt(struct sock *sk, int level, int optname,
791 sockptr_t optval, unsigned int optlen);
792 int mptcp_getsockopt(struct sock *sk, int level, int optname,
793 char __user *optval, int __user *option);
795 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq);
796 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit)
801 return __mptcp_expand_seq(old_seq, cur_seq);
803 void __mptcp_check_push(struct sock *sk, struct sock *ssk);
804 void __mptcp_data_acked(struct sock *sk);
805 void __mptcp_error_report(struct sock *sk);
806 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit);
807 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk)
809 return READ_ONCE(msk->snd_data_fin_enable) &&
810 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt);
813 static inline u32 mptcp_notsent_lowat(const struct sock *sk)
815 struct net *net = sock_net(sk);
818 val = READ_ONCE(mptcp_sk(sk)->notsent_lowat);
819 return val ?: READ_ONCE(net->ipv4.sysctl_tcp_notsent_lowat);
822 static inline bool mptcp_stream_memory_free(const struct sock *sk, int wake)
824 const struct mptcp_sock *msk = mptcp_sk(sk);
827 notsent_bytes = READ_ONCE(msk->write_seq) - READ_ONCE(msk->snd_nxt);
828 return (notsent_bytes << wake) < mptcp_notsent_lowat(sk);
831 static inline bool __mptcp_stream_is_writeable(const struct sock *sk, int wake)
833 return mptcp_stream_memory_free(sk, wake) &&
834 __sk_stream_is_writeable(sk, wake);
837 static inline void mptcp_write_space(struct sock *sk)
839 /* pairs with memory barrier in mptcp_poll */
841 if (mptcp_stream_memory_free(sk, 1))
842 sk_stream_write_space(sk);
845 static inline void __mptcp_sync_sndbuf(struct sock *sk)
847 struct mptcp_subflow_context *subflow;
848 int ssk_sndbuf, new_sndbuf;
850 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
853 new_sndbuf = READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_wmem[0]);
854 mptcp_for_each_subflow(mptcp_sk(sk), subflow) {
855 ssk_sndbuf = READ_ONCE(mptcp_subflow_tcp_sock(subflow)->sk_sndbuf);
857 subflow->cached_sndbuf = ssk_sndbuf;
858 new_sndbuf += ssk_sndbuf;
861 /* the msk max wmem limit is <nr_subflows> * tcp wmem[2] */
862 WRITE_ONCE(sk->sk_sndbuf, new_sndbuf);
863 mptcp_write_space(sk);
866 /* The called held both the msk socket and the subflow socket locks,
869 static inline void __mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
871 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
873 if (READ_ONCE(ssk->sk_sndbuf) != subflow->cached_sndbuf)
874 __mptcp_sync_sndbuf(sk);
877 /* the caller held only the subflow socket lock, either in process or
878 * BH context. Additionally this can be called under the msk data lock,
879 * so we can't acquire such lock here: let the delegate action acquires
880 * the needed locks in suitable order.
882 static inline void mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk)
884 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
886 if (likely(READ_ONCE(ssk->sk_sndbuf) == subflow->cached_sndbuf))
890 mptcp_subflow_delegate(subflow, MPTCP_DELEGATE_SNDBUF);
894 void mptcp_destroy_common(struct mptcp_sock *msk, unsigned int flags);
896 #define MPTCP_TOKEN_MAX_RETRIES 4
898 void __init mptcp_token_init(void);
899 static inline void mptcp_token_init_request(struct request_sock *req)
901 mptcp_subflow_rsk(req)->token_node.pprev = NULL;
904 int mptcp_token_new_request(struct request_sock *req);
905 void mptcp_token_destroy_request(struct request_sock *req);
906 int mptcp_token_new_connect(struct sock *ssk);
907 void mptcp_token_accept(struct mptcp_subflow_request_sock *r,
908 struct mptcp_sock *msk);
909 bool mptcp_token_exists(u32 token);
910 struct mptcp_sock *mptcp_token_get_sock(struct net *net, u32 token);
911 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot,
913 void mptcp_token_destroy(struct mptcp_sock *msk);
915 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn);
917 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac);
918 __sum16 __mptcp_make_csum(u64 data_seq, u32 subflow_seq, u16 data_len, __wsum sum);
920 void __init mptcp_pm_init(void);
921 void mptcp_pm_data_init(struct mptcp_sock *msk);
922 void mptcp_pm_data_reset(struct mptcp_sock *msk);
923 int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
924 struct mptcp_addr_info *addr);
925 int mptcp_pm_parse_entry(struct nlattr *attr, struct genl_info *info,
927 struct mptcp_pm_addr_entry *entry);
928 bool mptcp_pm_addr_families_match(const struct sock *sk,
929 const struct mptcp_addr_info *loc,
930 const struct mptcp_addr_info *rem);
931 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
932 void mptcp_pm_nl_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk);
933 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side);
934 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk);
935 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk);
936 void mptcp_pm_connection_closed(struct mptcp_sock *msk);
937 void mptcp_pm_subflow_established(struct mptcp_sock *msk);
938 bool mptcp_pm_nl_check_work_pending(struct mptcp_sock *msk);
939 void mptcp_pm_subflow_check_next(struct mptcp_sock *msk,
940 const struct mptcp_subflow_context *subflow);
941 void mptcp_pm_add_addr_received(const struct sock *ssk,
942 const struct mptcp_addr_info *addr);
943 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk,
944 const struct mptcp_addr_info *addr);
945 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk);
946 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk);
947 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk,
948 const struct mptcp_rm_list *rm_list);
949 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup);
950 void mptcp_pm_mp_fail_received(struct sock *sk, u64 fail_seq);
951 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
952 struct mptcp_addr_info *addr,
953 struct mptcp_addr_info *rem,
955 bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
956 const struct mptcp_addr_info *addr);
957 void mptcp_pm_free_anno_list(struct mptcp_sock *msk);
958 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk);
959 struct mptcp_pm_add_entry *
960 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
961 const struct mptcp_addr_info *addr, bool check_id);
962 struct mptcp_pm_add_entry *
963 mptcp_lookup_anno_list_by_saddr(const struct mptcp_sock *msk,
964 const struct mptcp_addr_info *addr);
965 int mptcp_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
967 u8 *flags, int *ifindex);
968 int mptcp_pm_nl_get_flags_and_ifindex_by_id(struct mptcp_sock *msk, unsigned int id,
969 u8 *flags, int *ifindex);
970 int mptcp_userspace_pm_get_flags_and_ifindex_by_id(struct mptcp_sock *msk,
972 u8 *flags, int *ifindex);
973 int mptcp_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
974 int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info);
975 int mptcp_userspace_pm_set_flags(struct sk_buff *skb, struct genl_info *info);
976 int mptcp_pm_announce_addr(struct mptcp_sock *msk,
977 const struct mptcp_addr_info *addr,
979 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
980 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list);
981 void mptcp_pm_remove_addrs(struct mptcp_sock *msk, struct list_head *rm_list);
983 void mptcp_free_local_addr_list(struct mptcp_sock *msk);
985 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
986 const struct sock *ssk, gfp_t gfp);
987 void mptcp_event_addr_announced(const struct sock *ssk, const struct mptcp_addr_info *info);
988 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id);
989 void mptcp_event_pm_listener(const struct sock *ssk,
990 enum mptcp_event_type event);
991 bool mptcp_userspace_pm_active(const struct mptcp_sock *msk);
993 void __mptcp_fastopen_gen_msk_ackseq(struct mptcp_sock *msk, struct mptcp_subflow_context *subflow,
994 const struct mptcp_options_received *mp_opt);
995 void mptcp_fastopen_subflow_synack_set_params(struct mptcp_subflow_context *subflow,
996 struct request_sock *req);
997 int mptcp_nl_fill_addr(struct sk_buff *skb,
998 struct mptcp_pm_addr_entry *entry);
1000 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk)
1002 return READ_ONCE(msk->pm.addr_signal) &
1003 (BIT(MPTCP_ADD_ADDR_SIGNAL) | BIT(MPTCP_ADD_ADDR_ECHO));
1006 static inline bool mptcp_pm_should_add_signal_addr(struct mptcp_sock *msk)
1008 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL);
1011 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk)
1013 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO);
1016 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk)
1018 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL);
1021 static inline bool mptcp_pm_is_userspace(const struct mptcp_sock *msk)
1023 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_USERSPACE;
1026 static inline bool mptcp_pm_is_kernel(const struct mptcp_sock *msk)
1028 return READ_ONCE(msk->pm.pm_type) == MPTCP_PM_TYPE_KERNEL;
1031 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port)
1033 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE;
1035 if (family == AF_INET6)
1036 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE;
1038 len += MPTCPOPT_THMAC_LEN;
1039 /* account for 2 trailing 'nop' options */
1041 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN;
1046 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list)
1048 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX)
1051 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1;
1054 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, const struct sk_buff *skb,
1055 unsigned int opt_size, unsigned int remaining,
1056 struct mptcp_addr_info *addr, bool *echo,
1057 bool *drop_other_suboptions);
1058 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining,
1059 struct mptcp_rm_list *rm_list);
1060 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc);
1061 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1062 int mptcp_userspace_pm_get_local_id(struct mptcp_sock *msk, struct mptcp_addr_info *skc);
1063 int mptcp_pm_dump_addr(struct sk_buff *msg, struct netlink_callback *cb);
1064 int mptcp_pm_nl_dump_addr(struct sk_buff *msg,
1065 struct netlink_callback *cb);
1066 int mptcp_userspace_pm_dump_addr(struct sk_buff *msg,
1067 struct netlink_callback *cb);
1068 int mptcp_pm_get_addr(struct sk_buff *skb, struct genl_info *info);
1069 int mptcp_pm_nl_get_addr(struct sk_buff *skb, struct genl_info *info);
1070 int mptcp_userspace_pm_get_addr(struct sk_buff *skb,
1071 struct genl_info *info);
1073 static inline u8 subflow_get_local_id(const struct mptcp_subflow_context *subflow)
1075 int local_id = READ_ONCE(subflow->local_id);
1082 void __init mptcp_pm_nl_init(void);
1083 void mptcp_pm_nl_work(struct mptcp_sock *msk);
1084 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
1085 const struct mptcp_rm_list *rm_list);
1086 unsigned int mptcp_pm_get_add_addr_signal_max(const struct mptcp_sock *msk);
1087 unsigned int mptcp_pm_get_add_addr_accept_max(const struct mptcp_sock *msk);
1088 unsigned int mptcp_pm_get_subflows_max(const struct mptcp_sock *msk);
1089 unsigned int mptcp_pm_get_local_addr_max(const struct mptcp_sock *msk);
1091 /* called under PM lock */
1092 static inline void __mptcp_pm_close_subflow(struct mptcp_sock *msk)
1094 if (--msk->pm.subflows < mptcp_pm_get_subflows_max(msk))
1095 WRITE_ONCE(msk->pm.accept_subflow, true);
1098 static inline void mptcp_pm_close_subflow(struct mptcp_sock *msk)
1100 spin_lock_bh(&msk->pm.lock);
1101 __mptcp_pm_close_subflow(msk);
1102 spin_unlock_bh(&msk->pm.lock);
1105 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk);
1106 void mptcp_sockopt_sync_locked(struct mptcp_sock *msk, struct sock *ssk);
1108 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb)
1110 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP);
1113 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops);
1115 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk)
1117 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1120 static inline bool mptcp_check_fallback(const struct sock *sk)
1122 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1123 struct mptcp_sock *msk = mptcp_sk(subflow->conn);
1125 return __mptcp_check_fallback(msk);
1128 static inline void __mptcp_do_fallback(struct mptcp_sock *msk)
1130 if (__mptcp_check_fallback(msk)) {
1131 pr_debug("TCP fallback already done (msk=%p)", msk);
1134 set_bit(MPTCP_FALLBACK_DONE, &msk->flags);
1137 static inline bool __mptcp_has_initial_subflow(const struct mptcp_sock *msk)
1139 struct sock *ssk = READ_ONCE(msk->first);
1141 return ssk && ((1 << inet_sk_state_load(ssk)) &
1142 (TCPF_ESTABLISHED | TCPF_SYN_SENT |
1143 TCPF_SYN_RECV | TCPF_LISTEN));
1146 static inline void mptcp_do_fallback(struct sock *ssk)
1148 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(ssk);
1149 struct sock *sk = subflow->conn;
1150 struct mptcp_sock *msk;
1153 __mptcp_do_fallback(msk);
1154 if (READ_ONCE(msk->snd_data_fin_enable) && !(ssk->sk_shutdown & SEND_SHUTDOWN)) {
1155 gfp_t saved_allocation = ssk->sk_allocation;
1157 /* we are in a atomic (BH) scope, override ssk default for data
1160 ssk->sk_allocation = GFP_ATOMIC;
1161 ssk->sk_shutdown |= SEND_SHUTDOWN;
1162 tcp_shutdown(ssk, SEND_SHUTDOWN);
1163 ssk->sk_allocation = saved_allocation;
1167 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a)
1169 static inline bool mptcp_check_infinite_map(struct sk_buff *skb)
1171 struct mptcp_ext *mpext;
1173 mpext = skb ? mptcp_get_ext(skb) : NULL;
1174 if (mpext && mpext->infinite_map)
1180 static inline bool is_active_ssk(struct mptcp_subflow_context *subflow)
1182 return (subflow->request_mptcp || subflow->request_join);
1185 static inline bool subflow_simultaneous_connect(struct sock *sk)
1187 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk);
1189 return (1 << sk->sk_state) &
1190 (TCPF_ESTABLISHED | TCPF_FIN_WAIT1 | TCPF_FIN_WAIT2 | TCPF_CLOSING) &&
1191 is_active_ssk(subflow) &&
1192 !subflow->conn_finished;
1195 #ifdef CONFIG_SYN_COOKIES
1196 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1197 struct sk_buff *skb);
1198 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1199 struct sk_buff *skb);
1200 void __init mptcp_join_cookie_init(void);
1203 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req,
1204 struct sk_buff *skb) {}
1206 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req,
1207 struct sk_buff *skb)
1212 static inline void mptcp_join_cookie_init(void) {}
1215 #endif /* __MPTCP_PROTOCOL_H */