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_RM_ADDR BIT(7) 26 #define OPTION_MPTCP_FASTCLOSE BIT(8) 27 #define OPTION_MPTCP_PRIO BIT(9) 28 #define OPTION_MPTCP_RST 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 #define MPTCPOPT_RST 8 40 41 /* MPTCP suboption lengths */ 42 #define TCPOLEN_MPTCP_MPC_SYN 4 43 #define TCPOLEN_MPTCP_MPC_SYNACK 12 44 #define TCPOLEN_MPTCP_MPC_ACK 20 45 #define TCPOLEN_MPTCP_MPC_ACK_DATA 22 46 #define TCPOLEN_MPTCP_MPJ_SYN 12 47 #define TCPOLEN_MPTCP_MPJ_SYNACK 16 48 #define TCPOLEN_MPTCP_MPJ_ACK 24 49 #define TCPOLEN_MPTCP_DSS_BASE 4 50 #define TCPOLEN_MPTCP_DSS_ACK32 4 51 #define TCPOLEN_MPTCP_DSS_ACK64 8 52 #define TCPOLEN_MPTCP_DSS_MAP32 10 53 #define TCPOLEN_MPTCP_DSS_MAP64 14 54 #define TCPOLEN_MPTCP_DSS_CHECKSUM 2 55 #define TCPOLEN_MPTCP_ADD_ADDR 16 56 #define TCPOLEN_MPTCP_ADD_ADDR_PORT 18 57 #define TCPOLEN_MPTCP_ADD_ADDR_BASE 8 58 #define TCPOLEN_MPTCP_ADD_ADDR_BASE_PORT 10 59 #define TCPOLEN_MPTCP_ADD_ADDR6 28 60 #define TCPOLEN_MPTCP_ADD_ADDR6_PORT 30 61 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE 20 62 #define TCPOLEN_MPTCP_ADD_ADDR6_BASE_PORT 22 63 #define TCPOLEN_MPTCP_PORT_LEN 2 64 #define TCPOLEN_MPTCP_PORT_ALIGN 2 65 #define TCPOLEN_MPTCP_RM_ADDR_BASE 3 66 #define TCPOLEN_MPTCP_PRIO 3 67 #define TCPOLEN_MPTCP_PRIO_ALIGN 4 68 #define TCPOLEN_MPTCP_FASTCLOSE 12 69 #define TCPOLEN_MPTCP_RST 4 70 71 #define TCPOLEN_MPTCP_MPC_ACK_DATA_CSUM (TCPOLEN_MPTCP_DSS_CHECKSUM + TCPOLEN_MPTCP_MPC_ACK_DATA) 72 73 /* MPTCP MP_JOIN flags */ 74 #define MPTCPOPT_BACKUP BIT(0) 75 #define MPTCPOPT_HMAC_LEN 20 76 #define MPTCPOPT_THMAC_LEN 8 77 78 /* MPTCP MP_CAPABLE flags */ 79 #define MPTCP_VERSION_MASK (0x0F) 80 #define MPTCP_CAP_CHECKSUM_REQD BIT(7) 81 #define MPTCP_CAP_EXTENSIBILITY BIT(6) 82 #define MPTCP_CAP_DENY_JOIN_ID0 BIT(5) 83 #define MPTCP_CAP_HMAC_SHA256 BIT(0) 84 #define MPTCP_CAP_FLAG_MASK (0x1F) 85 86 /* MPTCP DSS flags */ 87 #define MPTCP_DSS_DATA_FIN BIT(4) 88 #define MPTCP_DSS_DSN64 BIT(3) 89 #define MPTCP_DSS_HAS_MAP BIT(2) 90 #define MPTCP_DSS_ACK64 BIT(1) 91 #define MPTCP_DSS_HAS_ACK BIT(0) 92 #define MPTCP_DSS_FLAG_MASK (0x1F) 93 94 /* MPTCP ADD_ADDR flags */ 95 #define MPTCP_ADDR_ECHO BIT(0) 96 97 /* MPTCP MP_PRIO flags */ 98 #define MPTCP_PRIO_BKUP BIT(0) 99 100 /* MPTCP TCPRST flags */ 101 #define MPTCP_RST_TRANSIENT BIT(0) 102 103 /* MPTCP socket flags */ 104 #define MPTCP_DATA_READY 0 105 #define MPTCP_NOSPACE 1 106 #define MPTCP_WORK_RTX 2 107 #define MPTCP_WORK_EOF 3 108 #define MPTCP_FALLBACK_DONE 4 109 #define MPTCP_WORK_CLOSE_SUBFLOW 5 110 #define MPTCP_PUSH_PENDING 6 111 #define MPTCP_CLEAN_UNA 7 112 #define MPTCP_ERROR_REPORT 8 113 #define MPTCP_RETRANSMIT 9 114 #define MPTCP_WORK_SYNC_SETSOCKOPT 10 115 #define MPTCP_CONNECTED 11 116 117 static inline bool before64(__u64 seq1, __u64 seq2) 118 { 119 return (__s64)(seq1 - seq2) < 0; 120 } 121 122 #define after64(seq2, seq1) before64(seq1, seq2) 123 124 struct mptcp_options_received { 125 u64 sndr_key; 126 u64 rcvr_key; 127 u64 data_ack; 128 u64 data_seq; 129 u32 subflow_seq; 130 u16 data_len; 131 __sum16 csum; 132 u16 mp_capable : 1, 133 mp_join : 1, 134 fastclose : 1, 135 reset : 1, 136 dss : 1, 137 add_addr : 1, 138 rm_addr : 1, 139 mp_prio : 1, 140 echo : 1, 141 csum_reqd : 1, 142 backup : 1, 143 deny_join_id0 : 1; 144 u32 token; 145 u32 nonce; 146 u64 thmac; 147 u8 hmac[MPTCPOPT_HMAC_LEN]; 148 u8 join_id; 149 u8 use_map:1, 150 dsn64:1, 151 data_fin:1, 152 use_ack:1, 153 ack64:1, 154 mpc_map:1, 155 __unused:2; 156 struct mptcp_addr_info addr; 157 struct mptcp_rm_list rm_list; 158 u64 ahmac; 159 u8 reset_reason:4; 160 u8 reset_transient:1; 161 }; 162 163 static inline __be32 mptcp_option(u8 subopt, u8 len, u8 nib, u8 field) 164 { 165 return htonl((TCPOPT_MPTCP << 24) | (len << 16) | (subopt << 12) | 166 ((nib & 0xF) << 8) | field); 167 } 168 169 enum mptcp_pm_status { 170 MPTCP_PM_ADD_ADDR_RECEIVED, 171 MPTCP_PM_ADD_ADDR_SEND_ACK, 172 MPTCP_PM_RM_ADDR_RECEIVED, 173 MPTCP_PM_ESTABLISHED, 174 MPTCP_PM_ALREADY_ESTABLISHED, /* persistent status, set after ESTABLISHED event */ 175 MPTCP_PM_SUBFLOW_ESTABLISHED, 176 }; 177 178 enum mptcp_addr_signal_status { 179 MPTCP_ADD_ADDR_SIGNAL, 180 MPTCP_ADD_ADDR_ECHO, 181 MPTCP_ADD_ADDR_IPV6, 182 MPTCP_ADD_ADDR_PORT, 183 MPTCP_RM_ADDR_SIGNAL, 184 }; 185 186 struct mptcp_pm_data { 187 struct mptcp_addr_info local; 188 struct mptcp_addr_info remote; 189 struct list_head anno_list; 190 191 spinlock_t lock; /*protects the whole PM data */ 192 193 u8 addr_signal; 194 bool server_side; 195 bool work_pending; 196 bool accept_addr; 197 bool accept_subflow; 198 bool remote_deny_join_id0; 199 u8 add_addr_signaled; 200 u8 add_addr_accepted; 201 u8 local_addr_used; 202 u8 subflows; 203 u8 status; 204 struct mptcp_rm_list rm_list_tx; 205 struct mptcp_rm_list rm_list_rx; 206 }; 207 208 struct mptcp_data_frag { 209 struct list_head list; 210 u64 data_seq; 211 u16 data_len; 212 u16 offset; 213 u16 overhead; 214 u16 already_sent; 215 struct page *page; 216 }; 217 218 /* MPTCP connection sock */ 219 struct mptcp_sock { 220 /* inet_connection_sock must be the first member */ 221 struct inet_connection_sock sk; 222 u64 local_key; 223 u64 remote_key; 224 u64 write_seq; 225 u64 snd_nxt; 226 u64 ack_seq; 227 u64 rcv_wnd_sent; 228 u64 rcv_data_fin_seq; 229 int wmem_reserved; 230 struct sock *last_snd; 231 int snd_burst; 232 int old_wspace; 233 u64 recovery_snd_nxt; /* in recovery mode accept up to this seq; 234 * recovery related fields are under data_lock 235 * protection 236 */ 237 u64 snd_una; 238 u64 wnd_end; 239 unsigned long timer_ival; 240 u32 token; 241 int rmem_released; 242 unsigned long flags; 243 bool recovery; /* closing subflow write queue reinjected */ 244 bool can_ack; 245 bool fully_established; 246 bool rcv_data_fin; 247 bool snd_data_fin_enable; 248 bool rcv_fastclose; 249 bool use_64bit_ack; /* Set when we received a 64-bit DSN */ 250 bool csum_enabled; 251 spinlock_t join_list_lock; 252 struct work_struct work; 253 struct sk_buff *ooo_last_skb; 254 struct rb_root out_of_order_queue; 255 struct sk_buff_head receive_queue; 256 int tx_pending_data; 257 struct list_head conn_list; 258 struct list_head rtx_queue; 259 struct mptcp_data_frag *first_pending; 260 struct list_head join_list; 261 struct socket *subflow; /* outgoing connect/listener/!mp_capable */ 262 struct sock *first; 263 struct mptcp_pm_data pm; 264 struct { 265 u32 space; /* bytes copied in last measurement window */ 266 u32 copied; /* bytes copied in this measurement window */ 267 u64 time; /* start time of measurement window */ 268 u64 rtt_us; /* last maximum rtt of subflows */ 269 } rcvq_space; 270 271 u32 setsockopt_seq; 272 char ca_name[TCP_CA_NAME_MAX]; 273 }; 274 275 #define mptcp_lock_sock(___sk, cb) do { \ 276 struct sock *__sk = (___sk); /* silence macro reuse warning */ \ 277 might_sleep(); \ 278 spin_lock_bh(&__sk->sk_lock.slock); \ 279 if (__sk->sk_lock.owned) \ 280 __lock_sock(__sk); \ 281 cb; \ 282 __sk->sk_lock.owned = 1; \ 283 spin_unlock(&__sk->sk_lock.slock); \ 284 mutex_acquire(&__sk->sk_lock.dep_map, 0, 0, _RET_IP_); \ 285 local_bh_enable(); \ 286 } while (0) 287 288 #define mptcp_data_lock(sk) spin_lock_bh(&(sk)->sk_lock.slock) 289 #define mptcp_data_unlock(sk) spin_unlock_bh(&(sk)->sk_lock.slock) 290 291 #define mptcp_for_each_subflow(__msk, __subflow) \ 292 list_for_each_entry(__subflow, &((__msk)->conn_list), node) 293 294 static inline void msk_owned_by_me(const struct mptcp_sock *msk) 295 { 296 sock_owned_by_me((const struct sock *)msk); 297 } 298 299 static inline struct mptcp_sock *mptcp_sk(const struct sock *sk) 300 { 301 return (struct mptcp_sock *)sk; 302 } 303 304 /* the msk socket don't use the backlog, also account for the bulk 305 * free memory 306 */ 307 static inline int __mptcp_rmem(const struct sock *sk) 308 { 309 return atomic_read(&sk->sk_rmem_alloc) - READ_ONCE(mptcp_sk(sk)->rmem_released); 310 } 311 312 static inline int __mptcp_space(const struct sock *sk) 313 { 314 return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) - __mptcp_rmem(sk)); 315 } 316 317 static inline struct mptcp_data_frag *mptcp_send_head(const struct sock *sk) 318 { 319 const struct mptcp_sock *msk = mptcp_sk(sk); 320 321 return READ_ONCE(msk->first_pending); 322 } 323 324 static inline struct mptcp_data_frag *mptcp_send_next(struct sock *sk) 325 { 326 struct mptcp_sock *msk = mptcp_sk(sk); 327 struct mptcp_data_frag *cur; 328 329 cur = msk->first_pending; 330 return list_is_last(&cur->list, &msk->rtx_queue) ? NULL : 331 list_next_entry(cur, list); 332 } 333 334 static inline struct mptcp_data_frag *mptcp_pending_tail(const struct sock *sk) 335 { 336 struct mptcp_sock *msk = mptcp_sk(sk); 337 338 if (!msk->first_pending) 339 return NULL; 340 341 if (WARN_ON_ONCE(list_empty(&msk->rtx_queue))) 342 return NULL; 343 344 return list_last_entry(&msk->rtx_queue, struct mptcp_data_frag, list); 345 } 346 347 static inline struct mptcp_data_frag *mptcp_rtx_head(const struct sock *sk) 348 { 349 struct mptcp_sock *msk = mptcp_sk(sk); 350 351 if (msk->snd_una == READ_ONCE(msk->snd_nxt)) 352 return NULL; 353 354 return list_first_entry_or_null(&msk->rtx_queue, struct mptcp_data_frag, list); 355 } 356 357 struct csum_pseudo_header { 358 __be64 data_seq; 359 __be32 subflow_seq; 360 __be16 data_len; 361 __sum16 csum; 362 }; 363 364 struct mptcp_subflow_request_sock { 365 struct tcp_request_sock sk; 366 u16 mp_capable : 1, 367 mp_join : 1, 368 backup : 1, 369 csum_reqd : 1, 370 allow_join_id0 : 1; 371 u8 local_id; 372 u8 remote_id; 373 u64 local_key; 374 u64 idsn; 375 u32 token; 376 u32 ssn_offset; 377 u64 thmac; 378 u32 local_nonce; 379 u32 remote_nonce; 380 struct mptcp_sock *msk; 381 struct hlist_nulls_node token_node; 382 }; 383 384 static inline struct mptcp_subflow_request_sock * 385 mptcp_subflow_rsk(const struct request_sock *rsk) 386 { 387 return (struct mptcp_subflow_request_sock *)rsk; 388 } 389 390 enum mptcp_data_avail { 391 MPTCP_SUBFLOW_NODATA, 392 MPTCP_SUBFLOW_DATA_AVAIL, 393 }; 394 395 struct mptcp_delegated_action { 396 struct napi_struct napi; 397 struct list_head head; 398 }; 399 400 DECLARE_PER_CPU(struct mptcp_delegated_action, mptcp_delegated_actions); 401 402 #define MPTCP_DELEGATE_SEND 0 403 404 /* MPTCP subflow context */ 405 struct mptcp_subflow_context { 406 struct list_head node;/* conn_list of subflows */ 407 u64 local_key; 408 u64 remote_key; 409 u64 idsn; 410 u64 map_seq; 411 u32 snd_isn; 412 u32 token; 413 u32 rel_write_seq; 414 u32 map_subflow_seq; 415 u32 ssn_offset; 416 u32 map_data_len; 417 __wsum map_data_csum; 418 u32 map_csum_len; 419 u32 request_mptcp : 1, /* send MP_CAPABLE */ 420 request_join : 1, /* send MP_JOIN */ 421 request_bkup : 1, 422 mp_capable : 1, /* remote is MPTCP capable */ 423 mp_join : 1, /* remote is JOINing */ 424 fully_established : 1, /* path validated */ 425 pm_notified : 1, /* PM hook called for established status */ 426 conn_finished : 1, 427 map_valid : 1, 428 map_csum_reqd : 1, 429 map_data_fin : 1, 430 mpc_map : 1, 431 backup : 1, 432 send_mp_prio : 1, 433 rx_eof : 1, 434 can_ack : 1, /* only after processing the remote a key */ 435 disposable : 1; /* ctx can be free at ulp release time */ 436 enum mptcp_data_avail data_avail; 437 u32 remote_nonce; 438 u64 thmac; 439 u32 local_nonce; 440 u32 remote_token; 441 u8 hmac[MPTCPOPT_HMAC_LEN]; 442 u8 local_id; 443 u8 remote_id; 444 u8 reset_seen:1; 445 u8 reset_transient:1; 446 u8 reset_reason:4; 447 u8 stale_count; 448 449 long delegated_status; 450 struct list_head delegated_node; /* link into delegated_action, protected by local BH */ 451 452 u32 setsockopt_seq; 453 u32 stale_rcv_tstamp; 454 455 struct sock *tcp_sock; /* tcp sk backpointer */ 456 struct sock *conn; /* parent mptcp_sock */ 457 const struct inet_connection_sock_af_ops *icsk_af_ops; 458 void (*tcp_data_ready)(struct sock *sk); 459 void (*tcp_state_change)(struct sock *sk); 460 void (*tcp_write_space)(struct sock *sk); 461 void (*tcp_error_report)(struct sock *sk); 462 463 struct rcu_head rcu; 464 }; 465 466 static inline struct mptcp_subflow_context * 467 mptcp_subflow_ctx(const struct sock *sk) 468 { 469 struct inet_connection_sock *icsk = inet_csk(sk); 470 471 /* Use RCU on icsk_ulp_data only for sock diag code */ 472 return (__force struct mptcp_subflow_context *)icsk->icsk_ulp_data; 473 } 474 475 static inline struct sock * 476 mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) 477 { 478 return subflow->tcp_sock; 479 } 480 481 static inline u64 482 mptcp_subflow_get_map_offset(const struct mptcp_subflow_context *subflow) 483 { 484 return tcp_sk(mptcp_subflow_tcp_sock(subflow))->copied_seq - 485 subflow->ssn_offset - 486 subflow->map_subflow_seq; 487 } 488 489 static inline u64 490 mptcp_subflow_get_mapped_dsn(const struct mptcp_subflow_context *subflow) 491 { 492 return subflow->map_seq + mptcp_subflow_get_map_offset(subflow); 493 } 494 495 static inline void mptcp_add_pending_subflow(struct mptcp_sock *msk, 496 struct mptcp_subflow_context *subflow) 497 { 498 sock_hold(mptcp_subflow_tcp_sock(subflow)); 499 spin_lock_bh(&msk->join_list_lock); 500 list_add_tail(&subflow->node, &msk->join_list); 501 spin_unlock_bh(&msk->join_list_lock); 502 } 503 504 void mptcp_subflow_process_delegated(struct sock *ssk); 505 506 static inline void mptcp_subflow_delegate(struct mptcp_subflow_context *subflow) 507 { 508 struct mptcp_delegated_action *delegated; 509 bool schedule; 510 511 /* The implied barrier pairs with mptcp_subflow_delegated_done(), and 512 * ensures the below list check sees list updates done prior to status 513 * bit changes 514 */ 515 if (!test_and_set_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status)) { 516 /* still on delegated list from previous scheduling */ 517 if (!list_empty(&subflow->delegated_node)) 518 return; 519 520 /* the caller held the subflow bh socket lock */ 521 lockdep_assert_in_softirq(); 522 523 delegated = this_cpu_ptr(&mptcp_delegated_actions); 524 schedule = list_empty(&delegated->head); 525 list_add_tail(&subflow->delegated_node, &delegated->head); 526 sock_hold(mptcp_subflow_tcp_sock(subflow)); 527 if (schedule) 528 napi_schedule(&delegated->napi); 529 } 530 } 531 532 static inline struct mptcp_subflow_context * 533 mptcp_subflow_delegated_next(struct mptcp_delegated_action *delegated) 534 { 535 struct mptcp_subflow_context *ret; 536 537 if (list_empty(&delegated->head)) 538 return NULL; 539 540 ret = list_first_entry(&delegated->head, struct mptcp_subflow_context, delegated_node); 541 list_del_init(&ret->delegated_node); 542 return ret; 543 } 544 545 static inline bool mptcp_subflow_has_delegated_action(const struct mptcp_subflow_context *subflow) 546 { 547 return test_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 548 } 549 550 static inline void mptcp_subflow_delegated_done(struct mptcp_subflow_context *subflow) 551 { 552 /* pairs with mptcp_subflow_delegate, ensures delegate_node is updated before 553 * touching the status bit 554 */ 555 smp_wmb(); 556 clear_bit(MPTCP_DELEGATE_SEND, &subflow->delegated_status); 557 } 558 559 int mptcp_is_enabled(const struct net *net); 560 unsigned int mptcp_get_add_addr_timeout(const struct net *net); 561 int mptcp_is_checksum_enabled(const struct net *net); 562 int mptcp_allow_join_id0(const struct net *net); 563 void mptcp_subflow_fully_established(struct mptcp_subflow_context *subflow, 564 struct mptcp_options_received *mp_opt); 565 bool __mptcp_retransmit_pending_data(struct sock *sk); 566 bool mptcp_subflow_data_available(struct sock *sk); 567 void __init mptcp_subflow_init(void); 568 void mptcp_subflow_shutdown(struct sock *sk, struct sock *ssk, int how); 569 void mptcp_close_ssk(struct sock *sk, struct sock *ssk, 570 struct mptcp_subflow_context *subflow); 571 void mptcp_subflow_reset(struct sock *ssk); 572 void mptcp_sock_graft(struct sock *sk, struct socket *parent); 573 struct socket *__mptcp_nmpc_socket(const struct mptcp_sock *msk); 574 575 /* called with sk socket lock held */ 576 int __mptcp_subflow_connect(struct sock *sk, const struct mptcp_addr_info *loc, 577 const struct mptcp_addr_info *remote, 578 u8 flags, int ifindex); 579 int mptcp_subflow_create_socket(struct sock *sk, struct socket **new_sock); 580 void mptcp_info2sockaddr(const struct mptcp_addr_info *info, 581 struct sockaddr_storage *addr, 582 unsigned short family); 583 584 static inline bool mptcp_subflow_active(struct mptcp_subflow_context *subflow) 585 { 586 struct sock *ssk = mptcp_subflow_tcp_sock(subflow); 587 588 /* can't send if JOIN hasn't completed yet (i.e. is usable for mptcp) */ 589 if (subflow->request_join && !subflow->fully_established) 590 return false; 591 592 /* only send if our side has not closed yet */ 593 return ((1 << ssk->sk_state) & (TCPF_ESTABLISHED | TCPF_CLOSE_WAIT)); 594 } 595 596 static inline void mptcp_subflow_tcp_fallback(struct sock *sk, 597 struct mptcp_subflow_context *ctx) 598 { 599 sk->sk_data_ready = ctx->tcp_data_ready; 600 sk->sk_state_change = ctx->tcp_state_change; 601 sk->sk_write_space = ctx->tcp_write_space; 602 sk->sk_error_report = ctx->tcp_error_report; 603 604 inet_csk(sk)->icsk_af_ops = ctx->icsk_af_ops; 605 } 606 607 void __init mptcp_proto_init(void); 608 #if IS_ENABLED(CONFIG_MPTCP_IPV6) 609 int __init mptcp_proto_v6_init(void); 610 #endif 611 612 struct sock *mptcp_sk_clone(const struct sock *sk, 613 const struct mptcp_options_received *mp_opt, 614 struct request_sock *req); 615 void mptcp_get_options(const struct sock *sk, 616 const struct sk_buff *skb, 617 struct mptcp_options_received *mp_opt); 618 619 void mptcp_finish_connect(struct sock *sk); 620 void __mptcp_set_connected(struct sock *sk); 621 static inline bool mptcp_is_fully_established(struct sock *sk) 622 { 623 return inet_sk_state_load(sk) == TCP_ESTABLISHED && 624 READ_ONCE(mptcp_sk(sk)->fully_established); 625 } 626 void mptcp_rcv_space_init(struct mptcp_sock *msk, const struct sock *ssk); 627 void mptcp_data_ready(struct sock *sk, struct sock *ssk); 628 bool mptcp_finish_join(struct sock *sk); 629 bool mptcp_schedule_work(struct sock *sk); 630 int mptcp_setsockopt(struct sock *sk, int level, int optname, 631 sockptr_t optval, unsigned int optlen); 632 int mptcp_getsockopt(struct sock *sk, int level, int optname, 633 char __user *optval, int __user *option); 634 635 u64 __mptcp_expand_seq(u64 old_seq, u64 cur_seq); 636 static inline u64 mptcp_expand_seq(u64 old_seq, u64 cur_seq, bool use_64bit) 637 { 638 if (use_64bit) 639 return cur_seq; 640 641 return __mptcp_expand_seq(old_seq, cur_seq); 642 } 643 void __mptcp_check_push(struct sock *sk, struct sock *ssk); 644 void __mptcp_data_acked(struct sock *sk); 645 void __mptcp_error_report(struct sock *sk); 646 void mptcp_subflow_eof(struct sock *sk); 647 bool mptcp_update_rcv_data_fin(struct mptcp_sock *msk, u64 data_fin_seq, bool use_64bit); 648 void __mptcp_flush_join_list(struct mptcp_sock *msk); 649 static inline bool mptcp_data_fin_enabled(const struct mptcp_sock *msk) 650 { 651 return READ_ONCE(msk->snd_data_fin_enable) && 652 READ_ONCE(msk->write_seq) == READ_ONCE(msk->snd_nxt); 653 } 654 655 static inline bool mptcp_propagate_sndbuf(struct sock *sk, struct sock *ssk) 656 { 657 if ((sk->sk_userlocks & SOCK_SNDBUF_LOCK) || ssk->sk_sndbuf <= READ_ONCE(sk->sk_sndbuf)) 658 return false; 659 660 WRITE_ONCE(sk->sk_sndbuf, ssk->sk_sndbuf); 661 return true; 662 } 663 664 static inline void mptcp_write_space(struct sock *sk) 665 { 666 if (sk_stream_is_writeable(sk)) { 667 /* pairs with memory barrier in mptcp_poll */ 668 smp_mb(); 669 if (test_and_clear_bit(MPTCP_NOSPACE, &mptcp_sk(sk)->flags)) 670 sk_stream_write_space(sk); 671 } 672 } 673 674 void mptcp_destroy_common(struct mptcp_sock *msk); 675 676 #define MPTCP_TOKEN_MAX_RETRIES 4 677 678 void __init mptcp_token_init(void); 679 static inline void mptcp_token_init_request(struct request_sock *req) 680 { 681 mptcp_subflow_rsk(req)->token_node.pprev = NULL; 682 } 683 684 int mptcp_token_new_request(struct request_sock *req); 685 void mptcp_token_destroy_request(struct request_sock *req); 686 int mptcp_token_new_connect(struct sock *sk); 687 void mptcp_token_accept(struct mptcp_subflow_request_sock *r, 688 struct mptcp_sock *msk); 689 bool mptcp_token_exists(u32 token); 690 struct mptcp_sock *mptcp_token_get_sock(u32 token); 691 struct mptcp_sock *mptcp_token_iter_next(const struct net *net, long *s_slot, 692 long *s_num); 693 void mptcp_token_destroy(struct mptcp_sock *msk); 694 695 void mptcp_crypto_key_sha(u64 key, u32 *token, u64 *idsn); 696 697 void mptcp_crypto_hmac_sha(u64 key1, u64 key2, u8 *msg, int len, void *hmac); 698 699 void __init mptcp_pm_init(void); 700 void mptcp_pm_data_init(struct mptcp_sock *msk); 701 void mptcp_pm_subflow_chk_stale(const struct mptcp_sock *msk, struct sock *ssk); 702 void mptcp_pm_new_connection(struct mptcp_sock *msk, const struct sock *ssk, int server_side); 703 void mptcp_pm_fully_established(struct mptcp_sock *msk, const struct sock *ssk, gfp_t gfp); 704 bool mptcp_pm_allow_new_subflow(struct mptcp_sock *msk); 705 void mptcp_pm_connection_closed(struct mptcp_sock *msk); 706 void mptcp_pm_subflow_established(struct mptcp_sock *msk); 707 void mptcp_pm_subflow_closed(struct mptcp_sock *msk, u8 id); 708 void mptcp_pm_add_addr_received(struct mptcp_sock *msk, 709 const struct mptcp_addr_info *addr); 710 void mptcp_pm_add_addr_echoed(struct mptcp_sock *msk, 711 struct mptcp_addr_info *addr); 712 void mptcp_pm_add_addr_send_ack(struct mptcp_sock *msk); 713 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk); 714 void mptcp_pm_rm_addr_received(struct mptcp_sock *msk, 715 const struct mptcp_rm_list *rm_list); 716 void mptcp_pm_mp_prio_received(struct sock *sk, u8 bkup); 717 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk, 718 struct mptcp_addr_info *addr, 719 u8 bkup); 720 void mptcp_pm_free_anno_list(struct mptcp_sock *msk); 721 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk); 722 struct mptcp_pm_add_entry * 723 mptcp_pm_del_add_timer(struct mptcp_sock *msk, 724 struct mptcp_addr_info *addr, bool check_id); 725 struct mptcp_pm_add_entry * 726 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk, 727 struct mptcp_addr_info *addr); 728 729 int mptcp_pm_announce_addr(struct mptcp_sock *msk, 730 const struct mptcp_addr_info *addr, 731 bool echo); 732 int mptcp_pm_remove_addr(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 733 int mptcp_pm_remove_subflow(struct mptcp_sock *msk, const struct mptcp_rm_list *rm_list); 734 735 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk, 736 const struct sock *ssk, gfp_t gfp); 737 void mptcp_event_addr_announced(const struct mptcp_sock *msk, const struct mptcp_addr_info *info); 738 void mptcp_event_addr_removed(const struct mptcp_sock *msk, u8 id); 739 740 static inline bool mptcp_pm_should_add_signal(struct mptcp_sock *msk) 741 { 742 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_SIGNAL); 743 } 744 745 static inline bool mptcp_pm_should_add_signal_echo(struct mptcp_sock *msk) 746 { 747 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_ECHO); 748 } 749 750 static inline bool mptcp_pm_should_add_signal_ipv6(struct mptcp_sock *msk) 751 { 752 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_IPV6); 753 } 754 755 static inline bool mptcp_pm_should_add_signal_port(struct mptcp_sock *msk) 756 { 757 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_ADD_ADDR_PORT); 758 } 759 760 static inline bool mptcp_pm_should_rm_signal(struct mptcp_sock *msk) 761 { 762 return READ_ONCE(msk->pm.addr_signal) & BIT(MPTCP_RM_ADDR_SIGNAL); 763 } 764 765 static inline unsigned int mptcp_add_addr_len(int family, bool echo, bool port) 766 { 767 u8 len = TCPOLEN_MPTCP_ADD_ADDR_BASE; 768 769 if (family == AF_INET6) 770 len = TCPOLEN_MPTCP_ADD_ADDR6_BASE; 771 if (!echo) 772 len += MPTCPOPT_THMAC_LEN; 773 /* account for 2 trailing 'nop' options */ 774 if (port) 775 len += TCPOLEN_MPTCP_PORT_LEN + TCPOLEN_MPTCP_PORT_ALIGN; 776 777 return len; 778 } 779 780 static inline int mptcp_rm_addr_len(const struct mptcp_rm_list *rm_list) 781 { 782 if (rm_list->nr == 0 || rm_list->nr > MPTCP_RM_IDS_MAX) 783 return -EINVAL; 784 785 return TCPOLEN_MPTCP_RM_ADDR_BASE + roundup(rm_list->nr - 1, 4) + 1; 786 } 787 788 bool mptcp_pm_add_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 789 struct mptcp_addr_info *saddr, bool *echo, bool *port); 790 bool mptcp_pm_rm_addr_signal(struct mptcp_sock *msk, unsigned int remaining, 791 struct mptcp_rm_list *rm_list); 792 int mptcp_pm_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 793 794 void __init mptcp_pm_nl_init(void); 795 void mptcp_pm_nl_data_init(struct mptcp_sock *msk); 796 void mptcp_pm_nl_work(struct mptcp_sock *msk); 797 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk, 798 const struct mptcp_rm_list *rm_list); 799 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc); 800 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk); 801 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk); 802 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk); 803 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk); 804 805 void mptcp_sockopt_sync(struct mptcp_sock *msk, struct sock *ssk); 806 void mptcp_sockopt_sync_all(struct mptcp_sock *msk); 807 808 static inline struct mptcp_ext *mptcp_get_ext(const struct sk_buff *skb) 809 { 810 return (struct mptcp_ext *)skb_ext_find(skb, SKB_EXT_MPTCP); 811 } 812 813 void mptcp_diag_subflow_init(struct tcp_ulp_ops *ops); 814 815 static inline bool __mptcp_check_fallback(const struct mptcp_sock *msk) 816 { 817 return test_bit(MPTCP_FALLBACK_DONE, &msk->flags); 818 } 819 820 static inline bool mptcp_check_fallback(const struct sock *sk) 821 { 822 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 823 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 824 825 return __mptcp_check_fallback(msk); 826 } 827 828 static inline void __mptcp_do_fallback(struct mptcp_sock *msk) 829 { 830 if (test_bit(MPTCP_FALLBACK_DONE, &msk->flags)) { 831 pr_debug("TCP fallback already done (msk=%p)", msk); 832 return; 833 } 834 set_bit(MPTCP_FALLBACK_DONE, &msk->flags); 835 } 836 837 static inline void mptcp_do_fallback(struct sock *sk) 838 { 839 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 840 struct mptcp_sock *msk = mptcp_sk(subflow->conn); 841 842 __mptcp_do_fallback(msk); 843 } 844 845 #define pr_fallback(a) pr_debug("%s:fallback to TCP (msk=%p)", __func__, a) 846 847 static inline bool subflow_simultaneous_connect(struct sock *sk) 848 { 849 struct mptcp_subflow_context *subflow = mptcp_subflow_ctx(sk); 850 struct sock *parent = subflow->conn; 851 852 return sk->sk_state == TCP_ESTABLISHED && 853 !mptcp_sk(parent)->pm.server_side && 854 !subflow->conn_finished; 855 } 856 857 #ifdef CONFIG_SYN_COOKIES 858 void subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 859 struct sk_buff *skb); 860 bool mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 861 struct sk_buff *skb); 862 void __init mptcp_join_cookie_init(void); 863 #else 864 static inline void 865 subflow_init_req_cookie_join_save(const struct mptcp_subflow_request_sock *subflow_req, 866 struct sk_buff *skb) {} 867 static inline bool 868 mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_req, 869 struct sk_buff *skb) 870 { 871 return false; 872 } 873 874 static inline void mptcp_join_cookie_init(void) {} 875 #endif 876 877 #endif /* __MPTCP_PROTOCOL_H */ 878