1 /* SPDX-License-Identifier: GPL-2.0 */ 2 /* Multipath TCP 3 * 4 * Copyright (c) 2017 - 2019, Intel Corporation. 5 */ 6 7 #ifndef __MPTCP_PROTOCOL_H 8 #define __MPTCP_PROTOCOL_H 9 10 #include <linux/random.h> 11 #include <net/tcp.h> 12 #include <net/inet_connection_sock.h> 13 #include <uapi/linux/mptcp.h> 14 15 #define MPTCP_SUPPORTED_VERSION 1 16 17 /* MPTCP option bits */ 18 #define OPTION_MPTCP_MPC_SYN BIT(0) 19 #define OPTION_MPTCP_MPC_SYNACK BIT(1) 20 #define OPTION_MPTCP_MPC_ACK BIT(2) 21 #define OPTION_MPTCP_MPJ_SYN BIT(3) 22 #define OPTION_MPTCP_MPJ_SYNACK BIT(4) 23 #define OPTION_MPTCP_MPJ_ACK BIT(5) 24 #define OPTION_MPTCP_ADD_ADDR BIT(6) 25 #define OPTION_MPTCP_ADD_ADDR6 BIT(7) 26 #define OPTION_MPTCP_RM_ADDR BIT(8) 27 #define OPTION_MPTCP_FASTCLOSE BIT(9) 28 #define OPTION_MPTCP_PRIO BIT(10) 29 30 /* MPTCP option subtypes */ 31 #define MPTCPOPT_MP_CAPABLE 0 32 #define MPTCPOPT_MP_JOIN 1 33 #define MPTCPOPT_DSS 2 34 #define MPTCPOPT_ADD_ADDR 3 35 #define MPTCPOPT_RM_ADDR 4 36 #define MPTCPOPT_MP_PRIO 5 37 #define MPTCPOPT_MP_FAIL 6 38 #define MPTCPOPT_MP_FASTCLOSE 7 39 40 /* MPTCP suboption lengths */ 41 #define TCPOLEN_MPTCP_MPC_SYN 4 42 #define TCPOLEN_MPTCP_MPC_SYNACK 12 43 #define TCPOLEN_MPTCP_MPC_ACK 20 44 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 45 #define TCPOLEN_MPTCP_MPJ_SYN 12 46 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 47 #define TCPOLEN_MPTCP_MPJ_ACK 24 48 #define TCPOLEN_MPTCP_DSS_BASE 4 49 #define TCPOLEN_MPTCP_DSS_ACK32 4 50 #define TCPOLEN_MPTCP_DSS_ACK64 8 51 #define TCPOLEN_MPTCP_DSS_MAP32 10 52 #define TCPOLEN_MPTCP_DSS_MAP64 14 53 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 54 #define TCPOLEN_MPTCP_ADD_ADDR 16 55 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 56 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 57 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 58 #define TCPOLEN_MPTCP_ADD_ADDR6 28 59 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 60 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 61 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 62 #define TCPOLEN_MPTCP_PORT_LEN 2 63 #define TCPOLEN_MPTCP_PORT_ALIGN 2 64 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 65 #define TCPOLEN_MPTCP_PRIO 3 66 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 67 #define TCPOLEN_MPTCP_FASTCLOSE 12 68 69 /* MPTCP MP_JOIN flags */ 70 #define MPTCPOPT_BACKUP BIT(0) 71 #define MPTCPOPT_HMAC_LEN 20 72 #define MPTCPOPT_THMAC_LEN 8 73 74 /* MPTCP MP_CAPABLE flags */ 75 #define MPTCP_VERSION_MASK (0x0F) 76 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 77 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 78 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 79 #define MPTCP_CAP_FLAG_MASK (0x3F) 80 81 /* MPTCP DSS flags */ 82 #define MPTCP_DSS_DATA_FIN BIT(4) 83 #define MPTCP_DSS_DSN64 BIT(3) 84 #define MPTCP_DSS_HAS_MAP BIT(2) 85 #define MPTCP_DSS_ACK64 BIT(1) 86 #define MPTCP_DSS_HAS_ACK BIT(0) 87 #define MPTCP_DSS_FLAG_MASK (0x1F) 88 89 /* MPTCP ADD_ADDR flags */ 90 #define MPTCP_ADDR_ECHO BIT(0) 91 #define MPTCP_ADDR_IPVERSION_4 4 92 #define MPTCP_ADDR_IPVERSION_6 6 93 94 /* MPTCP MP_PRIO flags */ 95 #define MPTCP_PRIO_BKUP BIT(0) 96 97 /* MPTCP socket flags */ 98 #define MPTCP_DATA_READY 0 99 #define MPTCP_NOSPACE 1 100 #define MPTCP_WORK_RTX 2 101 #define MPTCP_WORK_EOF 3 102 #define MPTCP_FALLBACK_DONE 4 103 #define MPTCP_WORK_CLOSE_SUBFLOW 5 104 #define MPTCP_PUSH_PENDING 6 105 #define MPTCP_CLEAN_UNA 7 106 #define MPTCP_ERROR_REPORT 8 107 #define MPTCP_RETRANSMIT 9 108 109 static inline bool before64(__u64 seq1, __u64 seq2) 110 { 111 return (__s64)(seq1 - seq2) < 0; 112 } 113 114 #define after64(seq2, seq1) before64(seq1, seq2) 115 116 struct mptcp_options_received { 117 u64 sndr_key; 118 u64 rcvr_key; 119 u64 data_ack; 120 u64 data_seq; 121 u32 subflow_seq; 122 u16 data_len; 123 u16 mp_capable : 1, 124 mp_join : 1, 125 fastclose : 1, 126 dss : 1, 127 add_addr : 1, 128 rm_addr : 1, 129 mp_prio : 1, 130 family : 4, 131 echo : 1, 132 backup : 1; 133 u32 token; 134 u32 nonce; 135 u64 thmac; 136 u8 hmac[MPTCPOPT_HMAC_LEN]; 137 u8 join_id; 138 u8 use_map:1, 139 dsn64:1, 140 data_fin:1, 141 use_ack:1, 142 ack64:1, 143 mpc_map:1, 144 __unused:2; 145 u8 addr_id; 146 struct mptcp_rm_list rm_list; 147 union { 148 struct in_addr addr; 149 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 150 struct in6_addr addr6; 151 #endif 152 }; 153 u64 ahmac; 154 u16 port; 155 }; 156 157 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 158 { 159 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 160 ((nib & 0xF) << 8) | field); 161 } 162 163 struct mptcp_addr_info { 164 sa_family_t family; 165 __be16 port; 166 u8 id; 167 u8 flags; 168 int ifindex; 169 union { 170 struct in_addr addr; 171 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 172 struct in6_addr addr6; 173 #endif 174 }; 175 }; 176 177 enum mptcp_pm_status { 178 MPTCP_PM_ADD_ADDR_RECEIVED, 179 MPTCP_PM_ADD_ADDR_SEND_ACK, 180 MPTCP_PM_RM_ADDR_RECEIVED, 181 MPTCP_PM_ESTABLISHED, 182 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 183 MPTCP_PM_SUBFLOW_ESTABLISHED, 184 }; 185 186 enum mptcp_addr_signal_status { 187 MPTCP_ADD_ADDR_SIGNAL, 188 MPTCP_ADD_ADDR_ECHO, 189 MPTCP_ADD_ADDR_IPV6, 190 MPTCP_ADD_ADDR_PORT, 191 MPTCP_RM_ADDR_SIGNAL, 192 }; 193 194 struct mptcp_pm_data { 195 struct mptcp_addr_info local; 196 struct mptcp_addr_info remote; 197 struct list_head anno_list; 198 199 spinlock_t lock; /*protects the whole PM data */ 200 201 u8 addr_signal; 202 bool server_side; 203 bool work_pending; 204 bool accept_addr; 205 bool accept_subflow; 206 u8 add_addr_signaled; 207 u8 add_addr_accepted; 208 u8 local_addr_used; 209 u8 subflows; 210 u8 status; 211 struct mptcp_rm_list rm_list_tx; 212 struct mptcp_rm_list rm_list_rx; 213 }; 214 215 struct mptcp_data_frag { 216 struct list_head list; 217 u64 data_seq; 218 u16 data_len; 219 u16 offset; 220 u16 overhead; 221 u16 already_sent; 222 struct page *page; 223 }; 224 225 /* MPTCP connection sock */ 226 struct mptcp_sock { 227 /* inet_connection_sock must be the first member */ 228 struct inet_connection_sock sk; 229 u64 local_key; 230 u64 remote_key; 231 u64 write_seq; 232 u64 snd_nxt; 233 u64 ack_seq; 234 u64 rcv_wnd_sent; 235 u64 rcv_data_fin_seq; 236 int wmem_reserved; 237 struct sock *last_snd; 238 int snd_burst; 239 int old_wspace; 240 u64 snd_una; 241 u64 wnd_end; 242 unsigned long timer_ival; 243 u32 token; 244 int rmem_released; 245 unsigned long flags; 246 bool can_ack; 247 bool fully_established; 248 bool rcv_data_fin; 249 bool snd_data_fin_enable; 250 bool rcv_fastclose; 251 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 252 spinlock_t join_list_lock; 253 struct sock *ack_hint; 254 struct work_struct work; 255 struct sk_buff *ooo_last_skb; 256 struct rb_root out_of_order_queue; 257 struct sk_buff_head receive_queue; 258 struct sk_buff_head skb_tx_cache; /* this is wmem accounted */ 259 int tx_pending_data; 260 int size_goal_cache; 261 struct list_head conn_list; 262 struct list_head rtx_queue; 263 struct mptcp_data_frag *first_pending; 264 struct list_head join_list; 265 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 266 struct sock *first; 267 struct mptcp_pm_data pm; 268 struct { 269 u32 space; /* bytes copied in last measurement window */ 270 u32 copied; /* bytes copied in this measurement window */ 271 u64 time; /* start time of measurement window */ 272 u64 rtt_us; /* last maximum rtt of subflows */ 273 } rcvq_space; 274 }; 275 276 #define mptcp_lock_sock(___sk, cb) do { \ 277 struct sock *__sk = (___sk); /* silence macro reuse warning */ \ 278 might_sleep(); \ 279 spin_lock_bh(&__sk->sk_lock.slock); \ 280 if (__sk->sk_lock.owned) \ 281 __lock_sock(__sk); \ 282 cb; \ 283 __sk->sk_lock.owned = 1; \ 284 spin_unlock(&__sk->sk_lock.slock); \ 285 mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_); \ 286 local_bh_enable(); \ 287 } while (0) 288 289 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 290 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 291 292 #define mptcp_for_each_subflow(__msk, __subflow) \ 293 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 294 295 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 296 { 297 sock_owned_by_me((const struct sock *)msk); 298 } 299 300 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 301 { 302 return (struct mptcp_sock *)sk; 303 } 304 305 static inline int __mptcp_space(const struct sock *sk) 306 { 307 return tcp_space(sk) + READ_ONCE(mptcp_sk(sk)->rmem_released); 308 } 309 310 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 311 { 312 const struct mptcp_sock *msk = mptcp_sk(sk); 313 314 return READ_ONCE(msk->first_pending); 315 } 316 317 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 318 { 319 struct mptcp_sock *msk = mptcp_sk(sk); 320 struct mptcp_data_frag *cur; 321 322 cur = msk->first_pending; 323 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 324 list_next_entry(cur, list); 325 } 326 327 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 328 { 329 struct mptcp_sock *msk = mptcp_sk(sk); 330 331 if (!msk->first_pending) 332 return NULL; 333 334 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 335 return NULL; 336 337 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 338 } 339 340 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 341 { 342 struct mptcp_sock *msk = mptcp_sk(sk); 343 344 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 345 return NULL; 346 347 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 348 } 349 350 struct mptcp_subflow_request_sock { 351 struct tcp_request_sock sk; 352 u16 mp_capable : 1, 353 mp_join : 1, 354 backup : 1; 355 u8 local_id; 356 u8 remote_id; 357 u64 local_key; 358 u64 idsn; 359 u32 token; 360 u32 ssn_offset; 361 u64 thmac; 362 u32 local_nonce; 363 u32 remote_nonce; 364 struct mptcp_sock *msk; 365 struct hlist_nulls_node token_node; 366 }; 367 368 static inline struct mptcp_subflow_request_sock * 369 mptcp_subflow_rsk(const struct request_sock *rsk) 370 { 371 return (struct mptcp_subflow_request_sock *)rsk; 372 } 373 374 enum mptcp_data_avail { 375 MPTCP_SUBFLOW_NODATA, 376 MPTCP_SUBFLOW_DATA_AVAIL, 377 MPTCP_SUBFLOW_OOO_DATA 378 }; 379 380 struct mptcp_delegated_action { 381 struct napi_struct napi; 382 struct list_head head; 383 }; 384 385 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 386 387 #define MPTCP_DELEGATE_SEND 0 388 389 /* MPTCP subflow context */ 390 struct mptcp_subflow_context { 391 struct list_head node;/* conn_list of subflows */ 392 u64 local_key; 393 u64 remote_key; 394 u64 idsn; 395 u64 map_seq; 396 u32 snd_isn; 397 u32 token; 398 u32 rel_write_seq; 399 u32 map_subflow_seq; 400 u32 ssn_offset; 401 u32 map_data_len; 402 u32 request_mptcp : 1, /* send MP_CAPABLE */ 403 request_join : 1, /* send MP_JOIN */ 404 request_bkup : 1, 405 mp_capable : 1, /* remote is MPTCP capable */ 406 mp_join : 1, /* remote is JOINing */ 407 fully_established : 1, /* path validated */ 408 pm_notified : 1, /* PM hook called for established status */ 409 conn_finished : 1, 410 map_valid : 1, 411 mpc_map : 1, 412 backup : 1, 413 send_mp_prio : 1, 414 rx_eof : 1, 415 can_ack : 1, /* only after processing the remote a key */ 416 disposable : 1; /* ctx can be free at ulp release time */ 417 enum mptcp_data_avail data_avail; 418 u32 remote_nonce; 419 u64 thmac; 420 u32 local_nonce; 421 u32 remote_token; 422 u8 hmac[MPTCPOPT_HMAC_LEN]; 423 u8 local_id; 424 u8 remote_id; 425 426 long delegated_status; 427 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 428 429 struct sock *tcp_sock; /* tcp sk backpointer */ 430 struct sock *conn; /* parent mptcp_sock */ 431 const struct inet_connection_sock_af_ops *icsk_af_ops; 432 void (*tcp_data_ready)(struct sock *sk); 433 void (*tcp_state_change)(struct sock *sk); 434 void (*tcp_write_space)(struct sock *sk); 435 void (*tcp_error_report)(struct sock *sk); 436 437 struct rcu_head rcu; 438 }; 439 440 static inline struct mptcp_subflow_context * 441 mptcp_subflow_ctx(const struct sock *sk) 442 { 443 struct inet_connection_sock *icsk = inet_csk(sk); 444 445 /* Use RCU on icsk_ulp_data only for sock diag code */ 446 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 447 } 448 449 static inline struct sock * 450 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 451 { 452 return subflow->tcp_sock; 453 } 454 455 static inline u64 456 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 457 { 458 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 459 subflow->ssn_offset - 460 subflow->map_subflow_seq; 461 } 462 463 static inline u64 464 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 465 { 466 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 467 } 468 469 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk, 470 struct mptcp_subflow_context *subflow) 471 { 472 sock_hold(mptcp_subflow_tcp_sock(subflow)); 473 spin_lock_bh(&msk->join_list_lock); 474 list_add_tail(&subflow->node, &msk->join_list); 475 spin_unlock_bh(&msk->join_list_lock); 476 } 477 478 void mptcp_subflow_process_delegated(struct sock *ssk); 479 480 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow) 481 { 482 struct mptcp_delegated_action *delegated; 483 bool schedule; 484 485 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 486 * ensures the below list check sees list updates done prior to status 487 * bit changes 488 */ 489 if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) { 490 /* still on delegated list from previous scheduling */ 491 if (!list_empty(&subflow->delegated_node)) 492 return; 493 494 /* the caller held the subflow bh socket lock */ 495 lockdep_assert_in_softirq(); 496 497 delegated = this_cpu_ptr(&mptcp_delegated_actions); 498 schedule = list_empty(&delegated->head); 499 list_add_tail(&subflow->delegated_node, &delegated->head); 500 sock_hold(mptcp_subflow_tcp_sock(subflow)); 501 if (schedule) 502 napi_schedule(&delegated->napi); 503 } 504 } 505 506 static inline struct mptcp_subflow_context * 507 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 508 { 509 struct mptcp_subflow_context *ret; 510 511 if (list_empty(&delegated->head)) 512 return NULL; 513 514 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 515 list_del_init(&ret->delegated_node); 516 return ret; 517 } 518 519 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 520 { 521 return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 522 } 523 524 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow) 525 { 526 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 527 * touching the status bit 528 */ 529 smp_wmb(); 530 clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 531 } 532 533 int mptcp_is_enabled(struct net *net); 534 unsigned int mptcp_get_add_addr_timeout(struct net *net); 535 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 536 struct mptcp_options_received *mp_opt); 537 bool mptcp_subflow_data_available(struct sock *sk); 538 void __init mptcp_subflow_init(void); 539 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 540 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 541 struct mptcp_subflow_context *subflow); 542 void mptcp_subflow_reset(struct sock *ssk); 543 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 544 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 545 546 /* called with sk socket lock held */ 547 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 548 const struct mptcp_addr_info *remote); 549 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 550 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 551 struct sockaddr_storage *addr, 552 unsigned short family); 553 554 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 555 struct mptcp_subflow_context *ctx) 556 { 557 sk->sk_data_ready = ctx->tcp_data_ready; 558 sk->sk_state_change = ctx->tcp_state_change; 559 sk->sk_write_space = ctx->tcp_write_space; 560 sk->sk_error_report = ctx->tcp_error_report; 561 562 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 563 } 564 565 void __init mptcp_proto_init(void); 566 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 567 int __init mptcp_proto_v6_init(void); 568 #endif 569 570 struct sock *mptcp_sk_clone(const struct sock *sk, 571 const struct mptcp_options_received *mp_opt, 572 struct request_sock *req); 573 void mptcp_get_options(const struct sk_buff *skb, 574 struct mptcp_options_received *mp_opt); 575 576 void mptcp_finish_connect(struct sock *sk); 577 static inline bool mptcp_is_fully_established(struct sock *sk) 578 { 579 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 580 READ_ONCE(mptcp_sk(sk)->fully_established); 581 } 582 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 583 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 584 bool mptcp_finish_join(struct sock *sk); 585 bool mptcp_schedule_work(struct sock *sk); 586 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 587 void __mptcp_data_acked(struct sock *sk); 588 void __mptcp_error_report(struct sock *sk); 589 void mptcp_subflow_eof(struct sock *sk); 590 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 591 void __mptcp_flush_join_list(struct mptcp_sock *msk); 592 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 593 { 594 return READ_ONCE(msk->snd_data_fin_enable) && 595 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 596 } 597 598 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 599 { 600 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 601 return false; 602 603 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 604 return true; 605 } 606 607 static inline void mptcp_write_space(struct sock *sk) 608 { 609 if (sk_stream_is_writeable(sk)) { 610 /* pairs with memory barrier in mptcp_poll */ 611 smp_mb(); 612 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 613 sk_stream_write_space(sk); 614 } 615 } 616 617 void mptcp_destroy_common(struct mptcp_sock *msk); 618 619 void __init mptcp_token_init(void); 620 static inline void mptcp_token_init_request(struct request_sock *req) 621 { 622 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 623 } 624 625 int mptcp_token_new_request(struct request_sock *req); 626 void mptcp_token_destroy_request(struct request_sock *req); 627 int mptcp_token_new_connect(struct sock *sk); 628 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 629 struct mptcp_sock *msk); 630 bool mptcp_token_exists(u32 token); 631 struct mptcp_sock *mptcp_token_get_sock(u32 token); 632 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 633 long *s_num); 634 void mptcp_token_destroy(struct mptcp_sock *msk); 635 636 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 637 638 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 639 640 void __init mptcp_pm_init(void); 641 void mptcp_pm_data_init(struct mptcp_sock *msk); 642 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 643 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 644 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 645 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 646 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 647 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id); 648 void mptcp_pm_add_addr_received(struct mptcp_sock *msk, 649 const struct mptcp_addr_info *addr); 650 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 651 struct mptcp_addr_info *addr); 652 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 653 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 654 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 655 const struct mptcp_rm_list *rm_list); 656 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 657 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 658 struct mptcp_addr_info *addr, 659 u8 bkup); 660 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 661 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 662 struct mptcp_pm_add_entry * 663 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 664 struct mptcp_addr_info *addr); 665 struct mptcp_pm_add_entry * 666 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk, 667 struct mptcp_addr_info *addr); 668 669 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 670 const struct mptcp_addr_info *addr, 671 bool echo); 672 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 673 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 674 675 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 676 const struct sock *ssk, gfp_t gfp); 677 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info); 678 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 679 680 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 681 { 682 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 683 } 684 685 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 686 { 687 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 688 } 689 690 static inline bool mptcp_pm_should_add_signal_ipv6(struct mptcp_sock *msk) 691 { 692 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_IPV6); 693 } 694 695 static inline bool mptcp_pm_should_add_signal_port(struct mptcp_sock *msk) 696 { 697 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_PORT); 698 } 699 700 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 701 { 702 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 703 } 704 705 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 706 { 707 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 708 709 if (family == AF_INET6) 710 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 711 if (!echo) 712 len += MPTCPOPT_THMAC_LEN; 713 /* account for 2 trailing 'nop' options */ 714 if (port) 715 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 716 717 return len; 718 } 719 720 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 721 { 722 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 723 return -EINVAL; 724 725 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 726 } 727 728 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 729 struct mptcp_addr_info *saddr, bool *echo, bool *port); 730 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 731 struct mptcp_rm_list *rm_list); 732 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 733 734 void __init mptcp_pm_nl_init(void); 735 void mptcp_pm_nl_data_init(struct mptcp_sock *msk); 736 void mptcp_pm_nl_work(struct mptcp_sock *msk); 737 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 738 const struct mptcp_rm_list *rm_list); 739 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 740 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk); 741 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk); 742 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk); 743 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk); 744 745 static inline struct mptcp_ext *mptcp_get_ext(struct sk_buff *skb) 746 { 747 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 748 } 749 750 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 751 752 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 753 { 754 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 755 } 756 757 static inline bool mptcp_check_fallback(const struct sock *sk) 758 { 759 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 760 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 761 762 return __mptcp_check_fallback(msk); 763 } 764 765 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 766 { 767 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 768 pr_debug("TCP fallback already done (msk=%p)", msk); 769 return; 770 } 771 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 772 } 773 774 static inline void mptcp_do_fallback(struct sock *sk) 775 { 776 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 777 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 778 779 __mptcp_do_fallback(msk); 780 } 781 782 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 783 784 static inline bool subflow_simultaneous_connect(struct sock *sk) 785 { 786 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 787 struct sock *parent = subflow->conn; 788 789 return sk->sk_state == TCP_ESTABLISHED && 790 !mptcp_sk(parent)->pm.server_side && 791 !subflow->conn_finished; 792 } 793 794 #ifdef CONFIG_SYN_COOKIES 795 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 796 struct sk_buff *skb); 797 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 798 struct sk_buff *skb); 799 void __init mptcp_join_cookie_init(void); 800 #else 801 static inline void 802 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 803 struct sk_buff *skb) {} 804 static inline bool 805 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 806 struct sk_buff *skb) 807 { 808 return false; 809 } 810 811 static inline void mptcp_join_cookie_init(void) {} 812 #endif 813 814 #endif /* __MPTCP_PROTOCOL_H */ 815